From 6b69a9e218f2ca4e071056c73705c6f91b035893 Mon Sep 17 00:00:00 2001
From: Emily Miller <ejm714@gmail.com>
Date: Mon, 25 Oct 2021 16:44:06 -0700
Subject: [PATCH] 496 docs review (#150)

* add denspose and update save dir

* commit two autogenerated files

* changelog edits

* contribute edits

* WIP edits for models

* finish edits for models and separate out densepose

* full list of parks

* config options

* vscode format

* move config guide into tutorials

* extra options

* windows not tested

* tutorial edits

* remove extra nb

* put save path back

* quickstart edits

* vscode formatting

* finish quickstart edits

* train tutorial

* add template section

* logging

* ffpmeg install in readme

* finetuning

* capitalization

* remove ffmpmeg

* date

* add densepose

* tweak

* simplify

* fix densepose video link

* caps and tensorboard

* add help

* remove pythong piece since this is focused on yaml files

* more tweaks

* edit history not change log

* copy edits

* fix changelog

* table

* table

* typo

* alphabetize

* table bug

* table edits

* Simplify save_dir and some directory -> dir renames (#151)

* wip renames

* renames in docs

* readme

* data dir renamme in docs

* rename in code from data_directory to data_dir

* maintaining update

* fix capitalization

* further updates

* tweak

* do not overwrite

* add overwrite save dir

* add overwrite save dir to config

* update configs with all info

* use full train configuration

* only upload if does not exist

* tests for save

* overwrite param

* better set up and test for overwrite

* docs

* update docs with overwrite

* from overwrite_save_dir to overwrite

* missed rename

* remove machine specific from vlc

* unindent so test actually runs

* check for local and cached checkpoints

* should be and

* write out predict config before preds start like we do for train config

* update all configs and use only first 10 digits of hash

* dry run check after save is configured; more robust test

* reorder

* show save directory

* copy edits

* update template

* fix test

* lower case for consistency

* fix test
---
 .github/MAINTAINING.md                        |   2 +-
 HISTORY.md                                    |  14 +-
 Makefile                                      |   2 +-
 README.md                                     |  15 +-
 docs/docs/api-reference/densepose_config.md   |   3 +
 docs/docs/api-reference/densepose_manager.md  |   3 +
 docs/docs/changelog/index.md                  |   0
 docs/docs/configurations.md                   | 120 ++---
 docs/docs/contribute/index.md                 |  16 +-
 docs/docs/debugging.md                        |  14 +-
 docs/docs/extra-options.md                    |  40 +-
 docs/docs/install.md                          |  10 +-
 .../zamba_config_diagram.png}                 | Bin
 docs/docs/models/denspose.md                  | 101 +++++
 docs/docs/models/index.md                     | 334 --------------
 docs/docs/models/species-detection.md         | 318 +++++++++++++
 docs/docs/predict-tutorial.md                 |  41 +-
 docs/docs/python-tutorial.ipynb               | 423 ------------------
 docs/docs/quickstart.md                       |  51 ++-
 docs/docs/train-tutorial.md                   |  83 ++--
 docs/docs/yaml-config.md                      | 116 +----
 docs/mkdocs.yml                               |  24 +-
 templates/european.yaml                       |   6 +-
 templates/slowfast.yaml                       |   4 +-
 templates/time_distributed.yaml               |  12 +-
 tests/assets/sample_predict_config.yaml       |   2 +-
 tests/assets/sample_train_config.yaml         |   2 +-
 tests/conftest.py                             |   4 +-
 tests/test_cli.py                             |  23 +-
 tests/test_config.py                          | 105 +++--
 tests/test_densepose.py                       |  10 +-
 tests/test_model_manager.py                   |  20 +-
 zamba/cli.py                                  |  47 +-
 zamba/models/config.py                        | 157 +++----
 zamba/models/densepose/config.py              |  20 +-
 zamba/models/model_manager.py                 |  35 +-
 .../official_models/european/config.yaml      |  29 +-
 .../european/predict_configuration.yaml       |  12 +-
 .../european/train_configuration.yaml         |  10 +-
 .../official_models/slowfast/config.yaml      |  42 +-
 .../slowfast/predict_configuration.yaml       |   6 +-
 .../slowfast/train_configuration.yaml         |   6 +-
 .../time_distributed/config.yaml              |  42 +-
 .../predict_configuration.yaml                |  12 +-
 .../time_distributed/train_configuration.yaml |  10 +-
 zamba/models/publish_models.py                |  62 ++-
 zamba/models/utils.py                         |   7 +-
 47 files changed, 1071 insertions(+), 1344 deletions(-)
 create mode 100644 docs/docs/api-reference/densepose_config.md
 create mode 100644 docs/docs/api-reference/densepose_manager.md
 delete mode 100644 docs/docs/changelog/index.md
 rename docs/docs/{config_diagram.png => media/zamba_config_diagram.png} (100%)
 create mode 100644 docs/docs/models/denspose.md
 delete mode 100644 docs/docs/models/index.md
 create mode 100644 docs/docs/models/species-detection.md
 delete mode 100644 docs/docs/python-tutorial.ipynb

diff --git a/.github/MAINTAINING.md b/.github/MAINTAINING.md
index 88b2bdda..657f6f8c 100644
--- a/.github/MAINTAINING.md
+++ b/.github/MAINTAINING.md
@@ -113,7 +113,7 @@ make publish_models
 
 This will generate a public file name for each model based on the config hash and upload the model weights to the three DrivenData public s3 buckets. This will generate a folder in `zamba/models/official_models/{your_name_name}` that contains the official config as well as reference yaml and json files. You should PR everything in this folder.
 
-Lastly, you need to update the template in `templates`. The template should contain all the same info as the model's `config.yaml`, plus placeholders for `data_directory` and `labels` in `train_config`, and `data_directory`, `filepaths`, and `checkpoint` in `predict_config`.
+Lastly, you need to update the template in `templates`. The template should contain all the same info as the model's `config.yaml`, plus placeholders for `data_dir` and `labels` in `train_config`, and `data_dir`, `filepaths`, and `checkpoint` in `predict_config`.
 
 ### New model checklist
 
diff --git a/HISTORY.md b/HISTORY.md
index 218bb272..9912d643 100644
--- a/HISTORY.md
+++ b/HISTORY.md
@@ -1,8 +1,8 @@
-# zamba Changelog
+# `zamba` changelog
 
-## v2 <!-- TODO: add release date as, eg, (2021-10-22)>
+## v2 (2021-10-22)
 
-### Previous Model - Machine Learning Competition
+### Previous model: Machine learning competition
 
 The algorithms used by `zamba` v1 were based on the winning solution from the
 [Pri-matrix Factorization](https://www.drivendata.org/competitions/49/deep-learning-camera-trap-animals/) machine learning
@@ -12,14 +12,14 @@ The core algorithm in `zamba` v1 was a [stacked ensemble](https://en.wikipedia.o
 learning models, whose individual predictions were combined in the second level
 of the stack to form the final prediction.
 
-In v2, the stacked ensemble algorithm from v1 is replaced with three more powerful [single-model options](https://zamba.drivendata.org/docs/models/index.md): `time_distributed`, `slowfast`, and `european`. The new models utilize state-of-the-art image and video classification architectures, and are able to outperform the much more computationally intensive stacked ensemble model.
+In v2, the stacked ensemble algorithm from v1 is replaced with three more powerful [single-model options](../models/index.md): `time_distributed`, `slowfast`, and `european`. The new models utilize state-of-the-art image and video classification architectures, and are able to outperform the much more computationally intensive stacked ensemble model.
 
 ### New geographies and species
 
-`zamba` v2 incorporates data from western Europe (Germany) in additional to locations in central and west Africa. The new data is packaged in the pretrained `european` model, which can predict 11 common European species not present in `zamba` v1.
+`zamba` v2 incorporates data from western Europe (Germany). The new data is packaged in the pretrained `european` model, which can predict 11 common European species not present in `zamba` v1.
 
-`zamba` v2 also incorporates new training data for central and west Africa. `zamba` v1 was primarily focused on species commonly found on savannas. v2 incorporates data from camera traps in jungle ecosystems, adding 13 additional species to the pretrained models for central and west Africa.
+`zamba` v2 also incorporates new training data from 15 countries in central and west Africa, and adds 12 additional species to the pretrained African models.
 
 ### Retraining flexibility
 
-Model training is easier to reproduce in `zamba` v2, so users can finetune a pretrained model using their own data. `zamba` v2 also allows users to retrain a model on completely new labels.
\ No newline at end of file
+Model training is made available `zamba` v2, so users can finetune a pretrained model using their own data to improve performance for a specific ecology or set of sites. `zamba` v2 also allows users to retrain a model on completely new species labels.
diff --git a/Makefile b/Makefile
index e7d88579..9005298c 100644
--- a/Makefile
+++ b/Makefile
@@ -85,7 +85,7 @@ docs-setup:
 	| sed 's|https://zamba.drivendata.org/docs/||g' \
 	> docs/docs/index.md
 
-	sed 's|https://zamba.drivendata.org/docs/|../|g' HISTORY.md > docs/docs/changelog/index.md
+	sed 's|https://zamba.drivendata.org/docs/|../|g' HISTORY.md > docs/docs/changelog.md
 
 ## Build the static version of the docs
 docs: docs-setup
diff --git a/README.md b/README.md
index bc826fd6..69ff9cce 100644
--- a/README.md
+++ b/README.md
@@ -11,10 +11,10 @@ https://user-images.githubusercontent.com/46792169/138346340-98ee196a-5ecd-4753-
 
 **`zamba` is a tool built in Python that uses machine learning and computer vision to automatically detect and classify animals in camera trap videos.** You can use `zamba` to:
 
-- Filter out blank videos
 - Identify which species appear in each video
+- Filter out blank videos
 
-The tool is already trained to identify 42 species common to Africa and Europe (as well as blank, or "no species present"). Users can also input their own labeled videos to finetune a model and make predictions for new species or new contexts.
+The models in `zamba` can identify blank videos (where no animal is present) along with 32 species common to Africa and 11 species commmon to Europe. Users can also finetune models using their own labeled videos to then make predictions for new species and/or new ecologies.
 
 `zamba` can be used both as a command-line tool and as a Python package. It is also available as a user-friendly website application, [Zamba Cloud](https://www.zambacloud.com/).
 
@@ -62,8 +62,9 @@ Options:
   --help                Show this message and exit.
 
 Commands:
-  predict  Identify species in a video.
-  train    Train a model on your labeled data.
+  densepose  Run densepose algorithm on videos.
+  predict    Identify species in a video.
+  train      Train a model on your labeled data.
 ```
 
 `zamba` can be used "out of the box" to generate predictions or train a model using your own videos. `zamba` supports the same video formats as FFmpeg, [which are listed here](https://www.ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features). Any videos that fail a set of FFmpeg checks will be skipped during inference or training.
@@ -81,10 +82,10 @@ See the [Quickstart](https://zamba.drivendata.org/docs/quickstart/) page or the
 ### Training a model
 
 ```console
-$ zamba train --data-dir path/to/videos --labels path_to_labels.csv
+$ zamba train --data-dir path/to/videos --labels path_to_labels.csv --save_dir my_trained_model
 ```
 
-The newly trained model will be saved to a folder in the current working directory called `zamba_{model_name}`. For example, a model finetuned from the pretrained `time_distributed` model (the default) will be saved in `zamba_time_distributed`. The folder will contain a model checkpoint as well as training configuration, model hyperparameters, and test and validation metrics. Run `zamba train --help` to list all possible options to pass to `train`.
+The newly trained model will be saved to the specified save directory. The folder will contain a model checkpoint as well as training configuration, model hyperparameters, and validation and test metrics. Run `zamba train --help` to list all possible options to pass to `train`.
 
 See the [Quickstart](https://zamba.drivendata.org/docs/quickstart/) page or the user tutorial on [training a model](https://zamba.drivendata.org/docs/train-tutorial/) for more details.
 
@@ -98,4 +99,4 @@ The command is (from the project root):
 $ make tests
 ```
 
-See the docs page on [contributing to `zamba`](https://zamba.drivendata.org/docs/contribute/index.md) for details.
\ No newline at end of file
+See the docs page on [contributing to `zamba`](https://zamba.drivendata.org/docs/contribute/index.md) for details.
diff --git a/docs/docs/api-reference/densepose_config.md b/docs/docs/api-reference/densepose_config.md
new file mode 100644
index 00000000..382725e0
--- /dev/null
+++ b/docs/docs/api-reference/densepose_config.md
@@ -0,0 +1,3 @@
+# zamba.models.densepose.config
+
+::: zamba.models.densepose.config
diff --git a/docs/docs/api-reference/densepose_manager.md b/docs/docs/api-reference/densepose_manager.md
new file mode 100644
index 00000000..dfff2e55
--- /dev/null
+++ b/docs/docs/api-reference/densepose_manager.md
@@ -0,0 +1,3 @@
+# zamba.models.densepose.densepose_manager
+
+::: zamba.models.densepose.densepose_manager
diff --git a/docs/docs/changelog/index.md b/docs/docs/changelog/index.md
deleted file mode 100644
index e69de29b..00000000
diff --git a/docs/docs/configurations.md b/docs/docs/configurations.md
index 9ca83cce..10caefa1 100644
--- a/docs/docs/configurations.md
+++ b/docs/docs/configurations.md
@@ -1,4 +1,4 @@
-# All Optional Arguments
+# All configuration options
 
 Three main configuration classes are specific in `zamba`:
 
@@ -8,7 +8,7 @@ Three main configuration classes are specific in `zamba`:
 
 Here's a helpful diagram which shows how everything is related.
 
-![](config_diagram.png)
+![](../media/zamba_config_diagram.png)
 
 <a id='video-loading-arguments'></a>
 
@@ -28,17 +28,17 @@ All video loading arguments can be specified either in a [YAML file](yaml-config
     ```
 === "Python"
     ```python
-    from zamba.models.model_manager import predict_model
-    from zamba.models.config import PredictConfig
     from zamba.data.video import VideoLoaderConfig
+    from zamba.models.config import PredictConfig
+    from zamba.models.model_manager import predict_model
 
-    predict_config = PredictConfig(data_directory="example_vids/")
+    predict_config = PredictConfig(data_dir="example_vids/")
     video_loader_config = VideoLoaderConfig(
-        model_input_height=240, 
-        model_input_width=426, 
+        model_input_height=240,
+        model_input_width=426,
         total_frames=16
         # ... other parameters
-    ) 
+    )
     predict_model(
         predict_config=predict_config, video_loader_config=video_loader_config
     )
@@ -51,23 +51,23 @@ Let's look at the class documentation in Python.
 >> help(VideoLoaderConfig)
 
 class VideoLoaderConfig(pydantic.main.BaseModel)
- |  VideoLoaderConfig(*, 
- crop_bottom_pixels: int = None, 
- i_frames: bool = False, 
- scene_threshold: float = None, 
- megadetector_lite_config: zamba.models.megadetector_lite_yolox.MegadetectorLiteYoloXConfig = None, 
- frame_selection_height: int = None, 
- frame_selection_width: int = None, 
- total_frames: int = None, 
- ensure_total_frames: bool = True, 
- fps: float = None, 
- early_bias: bool = False, 
- frame_indices: List[int] = None, 
- evenly_sample_total_frames: bool = False, 
- pix_fmt: str = 'rgb24', 
- model_input_height: int = None, 
- model_input_width: int = None, 
- cache_dir: pathlib.Path = None, 
+ |  VideoLoaderConfig(*,
+ crop_bottom_pixels: int = None,
+ i_frames: bool = False,
+ scene_threshold: float = None,
+ megadetector_lite_config: zamba.models.megadetector_lite_yolox.MegadetectorLiteYoloXConfig = None,
+ frame_selection_height: int = None,
+ frame_selection_width: int = None,
+ total_frames: int = None,
+ ensure_total_frames: bool = True,
+ fps: float = None,
+ early_bias: bool = False,
+ frame_indices: List[int] = None,
+ evenly_sample_total_frames: bool = False,
+ pix_fmt: str = 'rgb24',
+ model_input_height: int = None,
+ model_input_width: int = None,
+ cache_dir: pathlib.Path = None,
  cleanup_cache: bool = False) -> None
 
  ...
@@ -99,7 +99,7 @@ Number of frames that should ultimately be returned. Defaults to `None`
 
 #### `ensure_total_frames (bool)`
 
-Selecting the number of frames by resampling may result in one more or fewer frames due to rounding. If True, ensure the requested number of frames is returned by either clipping or duplicating the final frame. Raises an error if no frames have been selected. Otherwise, return the array unchanged. Defaults to `True`
+Some frame selection methods may yield varying numbers of frames. If `True`, ensure the requested number of frames is returned by either clipping or duplicating the final frame. If no frames are selected, returns an array of the desired shape with all zeros. Otherwise, return the array unchanged. Defaults to `True`
 
 #### `fps (int, optional)`
 
@@ -107,7 +107,7 @@ Resample the video evenly from the entire duration to a specific number of frame
 
 #### `early_bias (bool, optional)`
 
-Resamples to 24 fps and selects 16 frames biased toward the front. This strategy was used by the [Pri-matrix Factorization](https://www.drivendata.org/competitions/49/deep-learning-camera-trap-animals/) machine learning
+Resamples to 24 fps and selects 16 frames biased toward the beginning of the video. This strategy was used by the [Pri-matrix Factorization](https://www.drivendata.org/competitions/49/deep-learning-camera-trap-animals/) machine learning
 competition winner. Defaults to `False`
 
 #### `frame_indices (list(int), optional)`
@@ -128,7 +128,7 @@ After frame selection, resize the video to this height and width in pixels. Defa
 
 #### `cache_dir (Path, optional)`
 
-Cache directory where preprocessed videos will be saved upon first load. Alternatively, can be set with VIDEO_CACHE_DIR environment variable. Provided there is enough space on your machine, it is highly encouraged to cache videos for training as this will speed up all subsequent epochs. If you are predicting on the same videos with the same video loader configuration, this will save time on future runs. Defaults to `None`, which means videos will not be cached.
+Cache directory where preprocessed videos will be saved upon first load. Alternatively, can be set with `VIDEO_CACHE_DIR` environment variable. Provided there is enough space on your machine, it is highly encouraged to cache videos for training as this will speed up all subsequent epochs. If you are predicting on the same videos with the same video loader configuration, this will save time on future runs. Defaults to `None`, which means videos will not be cached.
 
 #### `cleanup_cache (bool, optional)`
 
@@ -145,15 +145,17 @@ All possible model inference parameters are defined by the [`PredictConfig` clas
 >> help(PredictConfig)
 
 class PredictConfig(ZambaBaseModel)
- |  PredictConfig(*, 
- data_directory: DirectoryPath = # your current working directory ,
+ |  PredictConfig(*,
+ data_dir: DirectoryPath = Path.cwd(),
  filepaths: FilePath = None,
  checkpoint: FilePath = None,
  model_name: zamba.models.config.ModelEnum = <ModelEnum.time_distributed: 'time_distributed'>,
- gpus: int = 0, 
+ gpus: int = 0,
  num_workers: int = 3,
  batch_size: int = 2,
- save: Union[bool, pathlib.Path] = True,
+ save: bool = True,
+ save_dir: Optional[Path] = None,
+ overwrite: bool = False,
  dry_run: bool = False,
  proba_threshold: float = None,
  output_class_names: bool = False,
@@ -164,19 +166,19 @@ class PredictConfig(ZambaBaseModel)
  ...
 ```
 
-**Either `data_directory` or `filepaths` must be specified to instantiate `PredictConfig`.** Otherwise the current working directory will be used as the default `data_directory`.
+**Either `data_dir` or `filepaths` must be specified to instantiate `PredictConfig`.** If neither is specified, the current working directory will be used as the default `data_dir`.
 
-#### `data_directory (DirectoryPath, optional)`
+#### `data_dir (DirectoryPath, optional)`
 
-Path to the directory containing training videos. Defaults to the current working directory.
+Path to the directory containing videos for inference. Defaults to the current working directory.
 
 #### `filepaths (FilePath, optional)`
 
-Path to a list of files for classification. Defaults to the files in the current working directory
+Path to a csv containing a `filepath` column with videos for classification.
 
 #### `checkpoint (Path or str, optional)`
 
-Path to a model checkpoint to load and use for inference. The default is `None`, which automatically loads the pretrained checkpoint for the model specified by `model_name`. Since the default `model_name` is `time_distributed` the default `checkpoint` is `zamba_time_distributed.ckpt`
+Path to a model checkpoint to load and use for inference. The default is `None`, which will load the pretrained checkpoint if the model specified by `model_name`.
 
 #### `model_name (time_distributed|slowfast|european, optional)`
 
@@ -194,9 +196,18 @@ The number of CPUs to use during training. The maximum value for `num_workers` i
 
 The batch size to use for inference. Defaults to `2`
 
-#### `save (bool, optional)`
+#### `save (bool)`
 
-Whether to save out the predictions to a CSV file. y default, predictions will be saved at `zamba_predictions.csv` in the current working directory. Defaults to `True`
+Whether to save out predictions. If `False`, predictions are not saved. Defaults to `True`.
+
+#### `save_dir (Path, optional)`
+
+An optional directory in which to save the model predictions and configuration yaml.  If
+no `save_dir` is specified and `save` is True, outputs will be written to the current working directory. Defaults to `None`
+
+#### `overwrite (bool)`
+
+If True, will overwrite `zamba_predictions.csv` and `predict_configuration.yaml` in `save_dir` if they exist. Defaults to False.
 
 #### `dry_run (bool, optional)`
 
@@ -204,7 +215,7 @@ Specifying `True` is useful for trying out model implementations more quickly by
 
 #### `proba_threshold (float between 0 and 1, optional)`
 
-For advanced uses, you may want the algorithm to be more or less sensitive to if a species is present. This parameter is a `FLOAT` number, e.g., `0.64` corresponding to the probability threshold beyond which an animal is considered to be present in the video being analyzed.
+For advanced uses, you may want the algorithm to be more or less sensitive to if a species is present. This parameter is a float, e.g., `0.6` corresponding to the probability threshold beyond which an animal is considered to be present in the video being analyzed.
 
 By default no threshold is passed, `proba_threshold=None`. This will return a probability from 0-1 for each species that could occur in each video. If a threshold is passed, then the final prediction value returned for each class is `probability >= proba_threshold`, so that all class values become `0` (`False`, the species does not appear) or `1` (`True`, the species does appear).
 
@@ -216,7 +227,6 @@ Setting this option to `True` yields the most concise output `zamba` is capable
 
 Because `zamba` needs to download pretrained weights for the neural network architecture, we make these weights available in different regions. `us` is the default, but if you are not in the US you should use either `eu` for the European Union or `asia` for Asia Pacific to make sure that these download as quickly as possible for you.
 
-
 #### `skip_load_validation (bool, optional)`
 
 By default, before kicking off inference `zamba` will iterate through all of the videos in the data and verify that each can be loaded. Setting `skip_load_verification` to `True` skips this step. Validation can be very time intensive depending on the number of videos. It is recommended to run validation once, but not on future inference runs if the videos have not changed. Defaults to `False`
@@ -236,9 +246,9 @@ All possible model training parameters are defined by the [`TrainConfig` class](
 >> help(TrainConfig)
 
 class TrainConfig(ZambaBaseModel)
- |  TrainConfig(*, 
+ |  TrainConfig(*,
  labels: Union[FilePath, pandas.DataFrame],
- data_directory: DirectoryPath = # your current working directory ,
+ data_dir: DirectoryPath = # your current working directory ,
  checkpoint: FilePath = None,
  scheduler_config: Union[str, zamba.models.config.SchedulerConfig, NoneType] = 'default',
  model_name: zamba.models.config.ModelEnum = <ModelEnum.time_distributed: 'time_distributed'>,
@@ -249,7 +259,7 @@ class TrainConfig(ZambaBaseModel)
             BackboneFinetuneConfig(unfreeze_backbone_at_epoch=5,
             backbone_initial_ratio_lr=0.01, multiplier=1,
             pre_train_bn=False, train_bn=False, verbose=True),
- gpus: int = 0, 
+ gpus: int = 0,
  num_workers: int = 3,
  max_epochs: int = None,
  early_stopping_config: zamba.models.config.EarlyStoppingConfig =
@@ -257,8 +267,8 @@ class TrainConfig(ZambaBaseModel)
             verbose=True, mode='max'),
  weight_download_region: zamba.models.utils.RegionEnum = 'us',
  split_proportions: Dict[str, int] = {'train': 3, 'val': 1, 'holdout': 1},
- save_directory: pathlib.Path = # your current working directory , 
- overwrite_save_directory: bool = False, 
+ save_dir: pathlib.Path = # your current working directory ,
+ overwrite: bool = False,
  skip_load_validation: bool = False,
  from_scratch: bool = False,
  predict_all_zamba_species: bool = True,
@@ -267,13 +277,11 @@ class TrainConfig(ZambaBaseModel)
  ...
 ```
 
-**`data_directory` and `labels` must be specified to instantiate `TrainConfig`.** If no `data_directory` is provided, it will default the current working directory.
-
 #### `labels (FilePath or pd.DataFrame, required)`
 
-Either the path to a CSV file with labels for training, or a dataframe of the training labels. There must be columns for `filename` and `label`.
+Either the path to a CSV file with labels for training, or a dataframe of the training labels. There must be columns for `filename` and `label`. **`labels` must be specified to instantiate `TrainConfig`.**
 
-#### `data_directory (DirectoryPath, optional)`
+#### `data_dir (DirectoryPath, optional)`
 
 Path to the directory containing training videos. Defaults to the current working directory.
 
@@ -329,13 +337,13 @@ Because `zamba` needs to download pretrained weights for the neural network arch
 
 The proportion of data to use during training, validation, and as a holdout set. Defaults to `{"train": 3, "val": 1, "holdout": 1}`
 
-#### `save_directory (Path, optional)`
+#### `save_dir (Path, optional)`
 
-Directory in which to save model checkpoint and configuration file. If not specified, will save to a folder called 'zamba_{model_name}' in your working directory.
+Directory in which to save model checkpoint and configuration file. If not specified, will save to a `version_n` folder in your current working directory.
 
-#### `overwrite_save_directory (bool, optional)`
+#### `overwrite (bool, optional)`
 
- If True, will save outputs in `save_directory` and overwrite the directory if it exists. If False, will create an auto-incremented `version_n` folder within `save_directory` with model outputs. Defaults to `False`.
+ If `True`, will save outputs in `save_dir` and overwrite the directory if it exists. If False, will create an auto-incremented `version_n` folder within `save_dir` with model outputs. Defaults to `False`.
 
 #### `skip_load_validation (bool, optional)`
 
@@ -347,8 +355,8 @@ Whether to instantiate the model with base weights. This means starting from the
 
 #### `predict_all_zamba_species (bool, optional)`
 
-Whether the species outputted by the model should be all zamba species. If you want the model classes to only be the species in your labels file, set to `False`. Only used if labels is not `None`. If either `predict_all_zamba_species` is `False` or the labels contain species that are not in the model, the model head will be replaced. Defaults to `True`.
+Whether the species outputted by the model should be all zamba species. If you want the model classes to only be the species in your labels file, set to `False`. Only used if labels is not `None`. If either `predict_all_zamba_species` is `False` or the labels contain species that are not in the model, the model head will be replaced. Defaults to `True`
 
 #### `model_cache_dir (Path, optional)`
 
-Cache directory where downloaded model weights will be saved. If None and the MODEL_CACHE_DIR environment variable is not set, will use your default cache directory, which is often an automatic temp directory at `~/.cache/zamba`. Defaults to `None`.
+Cache directory where downloaded model weights will be saved. If None and the `MODEL_CACHE_DIR` environment variable is not set, will use your default cache directory, which is often an automatic temp directory at `~/.cache/zamba`. Defaults to `None`
diff --git a/docs/docs/contribute/index.md b/docs/docs/contribute/index.md
index a1ec4232..a5ae9b76 100644
--- a/docs/docs/contribute/index.md
+++ b/docs/docs/contribute/index.md
@@ -1,16 +1,13 @@
-# Help Make `zamba` Better
+# Help make `zamba` better
 
 `zamba` is an open source project, which means _you_ can help make it better!
 
-## Develop the Github Repository
+## Develop the github repository
+
 To get involved, check out the Github [code repository](https://github.com/drivendataorg/zamba).
-There you can find [open issues](https://github.com/drivendataorg/zamba/issues)
-, [project goals](https://github.com/drivendataorg/zamba/projects), and plenty
-of comments and links to help you along.
+There you can find [open issues](https://github.com/drivendataorg/zamba/issues) with comments and links to help you along.
 
-`zamba` uses continuous integration and test-driven-development to ensure
-that we always have a working project. So what are you
-waiting for? `git` going!
+`zamba` uses continuous integration and test-driven-development to ensure that we always have a working project. So what are you waiting for? `git` going!
 
 ## Installation for development
 
@@ -20,7 +17,7 @@ To install for development:
 ```console
 $ git clone https://github.com/drivendataorg/zamba.git
 $ cd zamba
-$ pip install --editable .
+$ pip install -r requirements-dev.txt
 ```
 
 ## Running the `zamba` test suite
@@ -36,4 +33,3 @@ $ make tests
 ## Submit additional training videos
 
 If you have additional labeled videos that may be useful for improving the basic models that ship with `zamba`, we'd love to hear from you! You can get in touch at [info@drivendata.org](mailto:info@drivendata.org)
-
diff --git a/docs/docs/debugging.md b/docs/docs/debugging.md
index afd8ab5a..1a09869c 100644
--- a/docs/docs/debugging.md
+++ b/docs/docs/debugging.md
@@ -12,7 +12,7 @@ Before kicking off a full run of inference or model training, we recommend testi
     In Python, add `dry_run=True` to [`PredictConfig`](configurations.md#prediction-arguments) or [`TrainConfig`](configurations.md#training-arguments):
     ```python
     predict_config = PredictConfig(
-        data_directory="example_vids/", dry_run=True
+        data_dir="example_vids/", dry_run=True
     )
     ```
 
@@ -21,7 +21,7 @@ Before kicking off a full run of inference or model training, we recommend testi
 The dry run will also catch any GPU memory errors. If you hit a GPU memory error, there are a couple fixes.
 
 #### Reducing the batch size
-  
+
 === "CLI"
     ```console
     zamba train --data-dir example_vids/ --labels example_labels.csv --batch-size 1
@@ -30,7 +30,7 @@ The dry run will also catch any GPU memory errors. If you hit a GPU memory error
     In Python, add `batch_size` to [`PredictConfig`](configurations.md#prediction-arguments) or [`TrainConfig`](configurations.md#training-arguments):
     ```python
     predict_config = PredictConfig(
-        data_directory="example_vids/", batch_size=1
+        data_dir="example_vids/", batch_size=1
     )
     ```
 
@@ -66,6 +66,10 @@ Reduce the number of workers (subprocesses) used for data loading. By default `n
     In Python, add `num_workers` to [`PredictConfig`](configurations.md#prediction-arguments) or [`TrainConfig`](configurations.md#training-arguments):
     ```python
     predict_config = PredictConfig(
-        data_directory="example_vids/", num_workers=1
+        data_dir="example_vids/", num_workers=1
     )
-    ```
\ No newline at end of file
+    ```
+
+## Logging
+
+To check that videos are getting loaded and cached as expected, set your environment variabe `LOG_LEVEL` to `debug`. The default log level is `info`.
diff --git a/docs/docs/extra-options.md b/docs/docs/extra-options.md
index bb02a8cd..8772844e 100644
--- a/docs/docs/extra-options.md
+++ b/docs/docs/extra-options.md
@@ -1,6 +1,6 @@
-# Guide to Common Optional Parameters
+# Guide to common optional parameters
 
-There are a LOT of ways to customize model training or inference. Here, we take that elephant-sized list of options and condense it to a manageable monkey-sized list of common considerations. To read about all possible customizations, see the [All Optional Arguments](configurations.md) page.
+There are a LOT of ways to customize model training or inference. Here, we take that elephant-sized list of options and condense it to a manageable monkey-sized list of common considerations. To read about all possible customizations, see [All Configuration Options](configurations.md).
 
 Many of the options below cannot be passed directly to the command line. Instead, some must be passed as part of a YAML configuration file. For example:
 
@@ -22,18 +22,18 @@ For using a YAML file with the Python package and other details, see the [YAML C
     In Python this can be specified in [`PredictConfig`](configurations.md#prediction-arguments) or [`TrainConfig`](configurations.md#training-arguments):
     ```python
     predict_config = PredictConfig(
-        data_directory="example_vids/",
+        data_dir="example_vids/",
         weight_download_region='asia',
     )
     ```
 
-The options for `weight_download_region` are `us`, `eu`, and `asia`. Once a model's weights are downloaded, the tool will use the local version and will not need to perform this download again.
+The options for `weight_download_region` are `us`, `eu`, and `asia`. Once a model's weights are downloaded, `zamba` will use the local version and will not need to perform this download again.
 
 ## Video size
 
-When `zamba` loads videos prior to either inference or training, it resizes all of the video frames before feeding them into a model. Higher resolution videos will lead to more detailed accuracy in prediction, but will use more memory and take longer to either predict on or train from. The default video loading configuration for all three pretrained models resizes images to 240x426 pixels. 
+When `zamba` loads videos prior to either inference or training, it resizes all of the video frames before feeding them into a model. Higher resolution videos will lead to superior accuracy in prediction, but will use more memory and take longer to train and/or predict. The default video loading configuration for all three pretrained models resizes images to 240x426 pixels.
 
-Say that you have a large number of videos, and you are more concerned with detecting blank v. non-blank videos than with identifying different species. In this case, you may not need a very high resolution and iterating through all of your videos with a high resolution would take a very long time. To resize all images to 50x50 pixels instead of the default 240x426: 
+Say that you have a large number of videos, and you are more concerned with detecting blank v. non-blank videos than with identifying different species. In this case, you may not need a very high resolution and iterating through all of your videos with a high resolution would take a very long time. To resize all images to 50x50 pixels instead of the default 240x426:
 
 === "YAML file"
     ```yaml
@@ -46,14 +46,16 @@ Say that you have a large number of videos, and you are more concerned with dete
     In Python, video resizing can be specified when `VideoLoaderConfig` is instantiated:
 
     ```python hl_lines="6 7 8"
-    from zamba.models.model_manager import predict_model
-    from zamba.models.config import PredictConfig
     from zamba.data.video import VideoLoaderConfig
+    from zamba.models.config import PredictConfig
+    from zamba.models.model_manager import predict_model
+
+    predict_config = PredictConfig(data_dir="example_vids/")
 
-    predict_config = PredictConfig(data_directory="example_vids/")
     video_loader_config = VideoLoaderConfig(
         model_input_height=50, model_input_width=50, total_frames=16
     ) # total_frames must always be specified
+
     predict_model(
         predict_config=predict_config, video_loader_config=video_loader_config
     )
@@ -63,11 +65,11 @@ Say that you have a large number of videos, and you are more concerned with dete
 
 Each video is simply a series of frames, or images. Most of the videos on which `zamba` was trained had 30 frames per second. That means even just a 15-second video would contain 450 frames.
 
-The model only trains or generates prediction based on a subset of the frames in a video, because using every frame would be far too computationally intensive. There are a number of different ways to select frames. For a full list of options, see the section on [Video loading arguments](configurations.md#video-loading-arguments). A few common approaches are explained below.
+All models only use a subset of the frames in a video, because using every frame would be far too computationally intensive. There are a number of different ways to select frames. For a full list of options, see the section on [Video loading arguments](configurations.md#video-loading-arguments). A few common approaches are explained below.
 
-### Early bias 
+### Early bias
 
-Some camera traps begin recording a video when movement is detected. If this is the case, you may be more likely to see an animal towards when the video starts. Setting `early_bias` to True selects 16 frames towards the beginning of a video. 
+Some camera traps begin recording a video when movement is detected. If this is the case, you may be more likely to see an animal towards when the video starts. Setting `early_bias` to True selects 16 frames towards the beginning of a video.
 
 === "YAML File"
     ```yaml
@@ -100,8 +102,8 @@ A simple option is to sample frames that are evenly distributed throughout a vid
     In Python, these arguments can be specified when `VideoLoaderConfig` is instantiated:
     ```python
     video_loader_config = VideoLoaderConfig(
-        total_frames=32, 
-        evenly_sample_total_frames=True, 
+        total_frames=32,
+        evenly_sample_total_frames=True,
         ensure_total_frames=True,
         ...
     )
@@ -109,7 +111,7 @@ A simple option is to sample frames that are evenly distributed throughout a vid
 
 ### MegadetectorLite
 
-You can use a pretrained object detection model called [MegadetectorLite](models/index.md#megadetectorlite) to select only the frames that are mostly likely to contain an animal. This is the default strategy for all three pretrained models. The parameter `megadetector_lite_config` is used to specify any arguments that should be passed to the MegadetectorLite model. If `megadetector_lite_config` is None, the MegadetectorLite model will not be used. 
+You can use a pretrained object detection model called [MegadetectorLite](models/index.md#megadetectorlite) to select only the frames that are mostly likely to contain an animal. This is the default strategy for all three pretrained models. The parameter `megadetector_lite_config` is used to specify any arguments that should be passed to the MegadetectorLite model. If `megadetector_lite_config` is None, the MegadetectorLite model will not be used.
 
 For example, to take the 16 frames with the highest probability of detection:
 
@@ -137,7 +139,7 @@ For example, to take the 16 frames with the highest probability of detection:
         total_frames=16,
     )
 
-    train_config = TrainConfig(data_directory="example_vids/", labels="example_labels.csv",)
+    train_config = TrainConfig(data_dir="example_vids/", labels="example_labels.csv",)
 
     train_model(video_loader_config=video_loader_config, train_config=train_config)
     ```
@@ -160,7 +162,7 @@ Both can be specified in either [`predict_config`](configurations.md#prediction-
 === "YAML file"
     ```yaml
     predict_config:
-        data_directory: example_vids/
+        data_dir: example_vids/
         num_workers: 5
         batch_size: 4
         # ... other parameters
@@ -168,7 +170,7 @@ Both can be specified in either [`predict_config`](configurations.md#prediction-
 === "Python"
     ```python
     predict_config = PredictConfig(
-        data_directory="example_vids/",
+        data_dir="example_vids/",
         num_workers=5,
         batch_size=4,
         # ... other parameters
@@ -176,4 +178,4 @@ Both can be specified in either [`predict_config`](configurations.md#prediction-
     ```
 
 
-And that's just the tip of the iceberg! See the [All Optional Arguments](configurations.md) page for more possibilities.
+And that's just the tip of the iceberg! See [All Configuration Options](configurations.md) page for more possibilities.
diff --git a/docs/docs/install.md b/docs/docs/install.md
index 93e61e68..7d550862 100644
--- a/docs/docs/install.md
+++ b/docs/docs/install.md
@@ -17,7 +17,6 @@ Prerequisites:
 We recommend [Python installation using Anaconda](https://www.anaconda.com/download/) for all platforms. For more information about how to install Anaconda, here are some useful YouTube videos of installation:
 
  - [Anaconda download link](https://www.anaconda.com/download/)
-
  - [Windows install video](https://www.youtube.com/watch?v=0OXBHvFeH_U)
  - [macOS installation video](https://www.youtube.com/watch?v=nVlrpNf3EdM)
 
@@ -27,7 +26,6 @@ We recommend [Python installation using Anaconda](https://www.anaconda.com/downl
 [FFmpeg](https://ffmpeg.org/ffmpeg.html) is an open source library for loading videos of different codecs. Using FFmpeg means that `zamba` can be flexible in terms of the video formats we support. FFmpeg can be installed on all different platforms, but requires some additional configuration depending on the platform. Here are some videos and instructions walking through FFmpeg installation:
 
  - [FFmpeg download link](https://www.ffmpeg.org/download.html)
-
  - [Install on Ubuntu or Linux](https://www.tecmint.com/install-ffmpeg-in-linux/).
      - In the command line, enter `sudo apt update` and then `sudo apt install ffmpeg`.
  - [MacOS install video](https://www.youtube.com/watch?v=8nbuqYw2OCw&t=5s)
@@ -65,7 +63,7 @@ $ pip install -U zamba
 ```
 
 
-## Operating Systems that have been tested
+## Operating systems that have been tested
 
 ### macOS
 
@@ -75,11 +73,9 @@ $ pip install -U zamba
 
 `zamba` has been tested on [Ubuntu](https://www.ubuntu.com/) versions 16 and 17.
 
-### Windows
-
-`zamba` has been tested on Windows 10.
+**Note: `zamba` has not been tested on Windows 10.**
 
-## Using GPU
+## Using GPU(s)
 
 `zamba` is much faster on a machine with a graphics processing unit (GPU), but has also been developed and tested for machines without GPU(s).
 
diff --git a/docs/docs/config_diagram.png b/docs/docs/media/zamba_config_diagram.png
similarity index 100%
rename from docs/docs/config_diagram.png
rename to docs/docs/media/zamba_config_diagram.png
diff --git a/docs/docs/models/denspose.md b/docs/docs/models/denspose.md
new file mode 100644
index 00000000..9325e787
--- /dev/null
+++ b/docs/docs/models/denspose.md
@@ -0,0 +1,101 @@
+# Densepose
+
+## Background
+
+Facebook AI Research has published a model, DensePose ([Neverova et al, 2021](https://arxiv.org/abs/2011.12438v1)), which can be used to get segmentations for animals that appear in videos. This was trained on the following animals, but often works for other species as well: sheep, zebra, horse, giraffe, elephant, cow, ear, cat, dog. Here's an example of the segmentation output for a frame:
+
+![segmentation of duiker](../media/seg_out.jpg)
+
+Additionally, the model provides mapping of the segmentation output to specific anatomy for chimpanzees. This can be helpful for determining the orientation of chimpanzees in videos and for their behaviors. Here is an example of what that output looks like:
+
+![chimpanzee texture output](../media/texture_out.png)
+
+For more information on the algorithms and outputs of the DensePose model, see the [Facebook DensePose Github Repository](https://github.com/facebookresearch/detectron2/tree/main/projects/DensePose).
+
+## Outputs
+
+The Zamba package supports running Densepose on videos to generate three types of outputs:
+
+ - A `.json` file with details of segmentations per video frame.
+ - A `.mp4` file where the original video has the segmentation rendered on top of animal so that the output can be vsiually inspected.
+ - A `.csv` (when `--output-type chimp_anatomy`) that contains the height and width of the bounding box around each chimpanzee, the frame number and timestamp of the observation, and the percentage of pixels in the bounding box that correspond with each anatomical part.
+
+Generally, running the densepose model is computationally intensive. It is recommended to run the model at a relatively low framerate (e.g., 1 frame per second) to generate outputs for a video. Another caveat is that because the output JSON output contains the full embedding, these files can be quite large. These are not written out by default.
+
+In order to use the densepose model, you must have PyTorch already installed on your system, and then you must install the `densepose` extra:
+
+```bash
+pip install torch  # see https://pytorch.org/get-started/locally/
+pip install "zamba[densepose]"
+```
+
+## Running DensePose
+
+Once that is done, here's how to run the DensePose model:
+
+=== "CLI"
+    ```bash
+    # create a segmentation output video for each input video in PATH_TO_VIDEOS
+    zamba densepose --data-dir PATH_TO_VIDEOS --render-output
+    ```
+=== "Python"
+    ```python
+    from zamba.models.densepose import DensePoseConfig
+    densepose_conf = DensePoseConfig(data_dir="PATH_TO_VIDEOS", render_output=True)
+    densepose_conf.run_model()
+    ```
+
+<video controls>
+  <source src="../../media/densepose_zamba_vid.mp4" type="video/mp4">
+</videp>
+
+
+## Getting help
+
+To see all of the available options, run `zamba densepose --help`.
+
+```console
+$ zamba densepose --help
+Usage: zamba densepose [OPTIONS]
+
+  Run densepose algorithm on videos.
+
+  If an argument is specified in both the command line and in a yaml file, the
+  command line input will take precedence.
+
+Options:
+  --data-dir PATH                 Path to video or image file or folder
+                                  containing images/videos.
+  --filepaths PATH                Path to csv containing `filepath` column
+                                  with videos.
+  --save-dir PATH                 An optional directory for saving the output.
+                                  Defaults to the current working directory.
+  --config PATH                   Specify options using yaml configuration
+                                  file instead of through command line
+                                  options.
+  --fps FLOAT                     Number of frames per second to process.
+                                  Defaults to 1.0 (1 frame per second).
+                                  [default: 1.0]
+  --output-type [segmentation|chimp_anatomy]
+                                  If 'chimp_anatomy' will apply anatomy model
+                                  from densepose to the rendering and create a
+                                  CSV with the anatomy visible in each frame.
+                                  If 'segmentation', will just output the
+                                  segmented area where an animal is
+                                  identified, which works for more species
+                                  than chimpanzees.  [default: chimp_anatomy]
+  --render-output / --no-render-output
+                                  If True, generate an output image or video
+                                  with either the segmentation or anatomy
+                                  rendered depending on the `output_type` that
+                                  is chosen.  [default: no-render-output]
+  --weight-download-region [us|eu|asia]
+                                  Server region for downloading weights.
+  --cache-dir PATH                Path to directory for model weights.
+                                  Alternatively, specify with environment
+                                  variable `ZAMBA_CACHE_DIR`. If not
+                                  specified, user's cache directory is used.
+  -y, --yes                       Skip confirmation of configuration and
+                                  proceed right to prediction.
+  --help                          Show this message and exit.
+```
diff --git a/docs/docs/models/index.md b/docs/docs/models/index.md
deleted file mode 100644
index 4db22f56..00000000
--- a/docs/docs/models/index.md
+++ /dev/null
@@ -1,334 +0,0 @@
-# Available Models
-
-The algorithms in `zamba` are designed to identify species of animals that appear in camera trap videos. There are three models that ship with the `zamba` package: `time_distributed`, `slowfast`, and `european`. For more details of each, read on!
-
-## Basic usage
-
-<table>
-  <tr>
-    <th>Model</th>
-    <th>Use cases</th>
-    <th>Strengths</th>
-    <th>Geography</th>
-  </tr>
-  <tr>
-    <td><code>time_distributed</code></td>
-    <td rowspan=2>Model training or fine tuning</td>
-    <td rowspan=2>Classifying species<br/>Running more quickly</td>
-    <td>Central and west Africa</td>
-  </tr>
-  <tr>
-    <td><code>european</code></td>
-    <td>Western Europe</td>
-  </tr>
-<tr>
-    <td><code>slowfast</code></td>
-    <td>Detailed prediction of blank vs. non-blank</td>
-    <td>Identifying blank vs. non-blank videos</td>
-    <td>Central and west Africa</td>
-  </tr>
-</table>
-
-`time_distributed` and `european` use the same basic algorithm. The main difference is that they predict different species based on their intended geography.
-
-For training or fine tuning, either the `time_distributed` and `european` model is recommended. These run much more quickly thatn the `slowfast` model.
-
-For inference, `slowfast` is recommended if the highest priority is differentiating between blank and non-blank videos. If the priority is species classification, either `time_distributed` or `european` is recommended based on the given geography.
-
-<h2 id="species-classes"></h2>
-
-## What species can `zamba` detect?
-
-`time_distributed` and `slowfast` are both trained to identify 32 common species from central and west Africa. The possible class labels in these models are:
-
-* `aardvark`
-* `antelope_duiker`
-* `badger`
-* `bat`
-* `bird`
-* `blank`
-* `cattle`
-* `cheetah`
-* `chimpanzee_bonobo`
-* `civet_genet`
-* `elephant`
-* `equid`
-* `forest_buffalo`
-* `fox`
-* `giraffe`
-* `gorilla`
-* `hare_rabbit`
-* `hippopotamus`
-* `hog`
-* `human`
-* `hyena`
-* `large_flightless_bird`
-* `leopard`
-* `lion`
-* `mongoose`
-* `monkey_prosimian`
-* `pangolin`
-* `porcupine`
-* `reptile`
-* `rodent`
-* `small_cat`
-* `wild_dog_jackal`
-
-`european` is trained to identify 11 common species in western Europe. The possible class labels are:
-
-* `bird`
-* `blank`
-* `domestic_cat`
-* `european_badger`
-* `european_beaver`
-* `european_hare`
-* `european_roe_deer`
-* `north_american_raccoon`
-* `red_fox`
-* `unidentified`
-* `weasel`
-* `wild_boar`
-
-<a id='time-distributed'></a>
-
-## `time_distributed` model
-
-### Algorithm
-
-The `time_distributed` model was built by re-training a well-known image classification architecture called [EfficientNetV2](https://arxiv.org/abs/1905.11946) to identify the species in our camera trap videos (Tan, M., & Le, Q., 2019). EfficientNetV2 models are convolutional [neural networks](https://www.youtube.com/watch?v=aircAruvnKk&t=995s) designed to jointly optimize model size and training speed. EfficientNetV2 is image native, meaning it classifies each frame separately when generating predictions. It does take into account the relationship between frames in the video.
-
-<a id='time-distributed-training-data'></a>
-
-### Training data
-
-`time_distributed` was trained using data collected and annotated by partners at [The Max Planck Institute for
-Evolutionary Anthropology](https://www.eva.mpg.de/index.html) and [Chimp &
-See](https://www.chimpandsee.org/). The data included camera trap videos from:
-
-* Dzanga-Sangha Protected Area, Central African Republic
-* Gorongosa National Park, Mozambique
-* Grumeti Game Reserve, Tanzania
-* Lopé National Park, Gabon
-* Moyen-Bafing National Park, Guinea
-* Nouabale-Ndoki National Park, Republic of the Congo
-* Salonga National Park, Democratic Republic of the Congo
-* Taï National Park, Côte d'Ivoire
-
-### Default configuration
-
-The full default configuration is available on [Github](https://github.com/drivendataorg/zamba/blob/master/zamba/models/official_models/time_distributed/config.yaml).
-
-By default, an efficient object detection model called [MegadetectorLite](#megadetectorlite) is run on all frames to determine which are the most likely to contain an animal. Then `time_distributed` is run on only the 16 frames with the highest predicted probability of detection. By default, videos are resized to 240x426 pixels.
-
-The full default video loading configuration is:
-```yaml
-video_loader_config:
-  model_input_height: 240
-  model_input_width: 426
-  crop_bottom_pixels: 50
-  ensure_total_frames: True
-  megadetector_lite_config:
-    confidence: 0.25
-    fill_model: "score_sorted"
-    n_frames: 16
-  total_frames: 16
-```
-
-### Requirements
-
-The above is pulled in by default if `time_distributed` is used in the command line. If you are passing in a custom [YAML configuration file](../yaml-config.md) or using `zamba` as a Python package, at a minimum you must specify:
-=== "YAML file"
-    ```yaml
-    video_loader_config:
-      model_input_height: # any integer
-      model_input_width: # any integer
-      total_frames: 16
-    ```
-=== "Python"
-    ```python
-    video_loader_config = VideoLoaderConfig(
-      model_input_height=..., # any integer
-      model_input_width=..., # any integer
-      total_frames=16
-    )
-    ```
-
-<a id='slowfast'></a>
-
-## `slowfast` model
-
-### Algorithm
-
-The `slowfast` model was built by re-training a video classification backbone called [SlowFast](https://arxiv.org/abs/1812.03982) (Feichtenhofer, C., Fan, H., Malik, J., & He, K., 2019). SlowFast refers to the two model pathways involved: one that operates at a low frame rate to capture spatial semantics, and one that operates at a high frame rate to capture motion over time. The basic architectures are deep [neural networks](https://www.youtube.com/watch?v=aircAruvnKk&t=995s) using [pytorch](https://pytorch.org/).
-
-<div style="text-align:center;">
-<img src="https://s3.amazonaws.com/drivendata-public-assets/zamba-slowfast-diagram.png" alt="Architecture showing the two pathways of the slowfast model" style="width:400px;"/>
-<br/><br/>
-<i>Source:</i> Feichtenhofer, C., Fan, H., Malik, J., & He, K. (2019). Slowfast networks for video recognition. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 6202-6211).
-</div>
-
-Unlike `time_distributed`, `slowfast` is video native. This means it takes into account the relationship between frames in a video, rather than running independently on each frame.
-
-### Training data
-
-The `slowfast` model was trained using the same data as the [`time_distributed` model](#time-distributed-training-data).
-
-### Default configuration
-
-The full default configuration is available on [Github](https://github.com/drivendataorg/zamba/blob/master/zamba/models/official_models/slowfast/config.yaml).
-
-By default, an efficient object detection model called [MegadetectorLite](#megadetectorlite) is run on all frames to determine which are the most likely to contain an animal. Then `slowfast` is run on only the 32 frames with the highest predicted probability of detection. By default, videos are resized to 240x426 pixels.
-
-The full default video loading configuration is:
-
-```yaml
-video_loader_config:
-  model_input_height: 240
-  model_input_width: 426
-  crop_bottom_pixels: 50
-  ensure_total_frames: True
-  megadetector_lite_config:
-    confidence: 0.25
-    fill_model: "score_sorted"
-    n_frames: 32
-  total_frames: 32
-```
-
-### Requirements
-
-The above is pulled in by default if `slowfast` is used in the command line. If you are passing in a custom [YAML configuration file](../yaml-config.md) or using `zamba` as a Python package, at a minimum you must specify:
-=== "YAML file"
-    ```yaml
-    video_loader_config:
-      model_input_height: # any integer >= 200
-      model_input_width: # any integer >= 200
-      total_frames: 32
-    ```
-=== "Python"
-    ```python
-    video_loader_config = VideoLoaderConfig(
-      model_input_height=..., # any integer >= 200
-      model_input_width=..., # any integer >= 200
-      total_frames=32
-    )
-    ```
-
-<a id='european'></a>
-
-## `european` model
-
-### Algorithm
-
-The `european` model has the same backbone as the `time_distributed` model, but is trained on data from camera traps in western Europe instead of central and west Africa.
-
-The `european` model was built by re-training a well-known image classification architecture called [EfficientNetV2](https://arxiv.org/abs/1905.11946) to identify the species in our camera trap videos (Tan, M., & Le, Q., 2019). EfficientNetV2 models are convolutional [neural networks](https://www.youtube.com/watch?v=aircAruvnKk&t=995s) designed to jointly optimize model size and training speed. EfficientNetV2 is image native, meaning it classifies each frame separately when generating predictions. It does take into account the relationship between frames in the video.
-
-`european` combines the EfficientNetV2 architecture with an open-source image object detection model to implement frame selection. The [YOLOX](https://arxiv.org/abs/2107.08430) detection model is run on all frames in a video. Only the frames with the highest probability of detection are then passed to the more computationally intensive EfficientNetV2 for detailed detection and classification.
-
-### Training data
-
-The `european` model is built by starting with the fully trained `time_distributed` model. The network is then finetuned with data collected and annotated by partners at [The Max Planck Institute for
-Evolutionary Anthropology](https://www.eva.mpg.de/index.html). The finetuning data included camera trap videos from Hintenteiche bei Biesenbrow, Germany.
-
-### Default configuration
-
-The full default configuration is available on [Github](https://github.com/drivendataorg/zamba/blob/master/zamba/models/official_models/european/config.yaml).
-
-By default, an efficient object detection model called [MegadetectorLite](#megadetectorlite) is run on all frames to determine which are the most likely to contain an animal. Then `european` is run on only the 16 frames with the highest predicted probability of detection. By default, videos are resized to 240x426 pixels.
-
-The full default video loading configuration is:
-```yaml
-video_loader_config:
-  model_input_height: 240
-  model_input_width: 426
-  crop_bottom_pixels: 50
-  ensure_total_frames: True
-  megadetector_lite_config:
-    confidence: 0.25
-    fill_model: "score_sorted"
-    n_frames: 16
-  total_frames: 16
-```
-
-### Requirements
-
-The above is pulled in by default if `european` is used in the command line. If you are passing in a custom [YAML configuration file](../yaml-config.md) or using `zamba` as a Python package, at a minimum you must specify:
-
-=== "YAML file"
-    ```yaml
-    video_loader_config:
-      model_input_height: # any integer
-      model_input_width: # any integer
-      total_frames: 16
-    ```
-=== "Python"
-    ```python
-    video_loader_config = VideoLoaderConfig(
-      model_input_height=..., # any integer
-      model_input_width=..., # any integer
-      total_frames=16
-    )
-    ```
-
-<a id='megadetectorlite'></a>
-
-## MegadetectorLite
-
-Running any of the three models that ship with `zamba` on all frames of a video would be incredibly time consuming and computationally intensive. Instead, `zamba` uses a more efficient object detection model called MegadetectorLite to determine the likelihood that each frame contains an animal. Then, only the frames with the highest probability of detection can be passed to the model.
-
-MegadetectorLite combines two open-source models:
-
-* [Megadetector](https://github.com/microsoft/CameraTraps/blob/master/megadetector.md) is a pretrained image model designed to detect animals, people, and vehicles in camera trap videos.
-* [YOLOX](https://github.com/Megvii-BaseDetection/YOLOX) is a high-performance, lightweight object detection model that is much less computationally intensive than Megadetector.
-
-While highly accurate, Megadetector is too computationally intensive to run on every frame. MegadetectorLite was created by training a YOLOX model using the predictions of the Megadetector as ground truth - this method is called [student-teacher training](https://towardsdatascience.com/knowledge-distillation-simplified-dd4973dbc764).
-
-
-## Densepose
-
-Facebook AI Research has published a model, DensePose ([Neverova et al, 2021](https://arxiv.org/abs/2011.12438v1)), which can be used to get segmentations for animals that appear in videos. This was trained on the following animals, but often works for other species as well: sheep, zebra, horse, giraffe, elephant, cow, ear, cat, dog. Here's an example of the segmentation output for a frame:
-
-![segmentation of duiker](../media/seg_out.jpg)
-
-Additionally, the model provides mapping of the segmentation output to specific anatomy for chimpanzees. This can be helpful for determining the orientation of chimpanzees in videos and for their behaviors. Here is an example of what that output looks like:
-
-![chimpanzee texture output](../media/texture_out.png)
-
-For more information on the algorithms and outputs of the DensePose model, see the [Facebook DensePose Github Repository](https://github.com/facebookresearch/detectron2/tree/main/projects/DensePose).
-
-The Zamba package supports running Densepose on videos to generate three types of outputs:
-
- - A `.json` file with details of segmentations per video frame.
- - A `.mp4` file where the original video has the segmentation rendered on top of animal so that the output can be vsiually inspected.
- - A `.csv` (when `--output-type chimp_anatomy`) that contains the height and width of the bounding box around each chimpanzee, the frame number and timestamp of the observation, and the percentage of pixels in the bounding box that correspond with each anatomical part.
-
-Generally, running the densepose model is computationally intensive. It is recommended to run the model at a relatively low framerate (e.g., 1 frame per second) to generate outputs for a video. Another caveat is that because the output JSON output contains the full embedding, these files can be quite large. These are not written out by default.
-
-In order to use the densepose model, you must have PyTorch already installed on your system, and then you must install the `densepose` extra:
-
-```bash
-pip install torch  # see https://pytorch.org/get-started/locally/
-pip install "zamba[densepose]"
-```
-
-Once that is done, here's how to run the DensePose model:
-
-=== "CLI"
-    ```bash
-    # create a segmentation output video for each input video in PATH_TO_VIDEOS
-    zamba densepose --data-dir PATH_TO_VIDEOS --render-output
-    ```
-=== "Python"
-    ```python
-    from zamba.models.densepose import DensePoseConfig
-    densepose_conf = DensePoseConfig(data_directory="PATH_TO_VIDEOS", render_output=True)
-    densepose_conf.run_model()
-    ```
-
-
-<video controls>
-  <source src="../media/densepose_zamba_vid.mp4" type="video/mp4">
-</videp>
-
-To see all of the available options, run `zamba densepose --help`.
-
diff --git a/docs/docs/models/species-detection.md b/docs/docs/models/species-detection.md
new file mode 100644
index 00000000..43b14719
--- /dev/null
+++ b/docs/docs/models/species-detection.md
@@ -0,0 +1,318 @@
+# Available models
+
+The algorithms in `zamba` are designed to identify species of animals that appear in camera trap videos. There are three models that ship with the `zamba` package: `time_distributed`, `slowfast`, and `european`. For more details of each, read on!
+
+## Model summary
+
+<table>
+  <tr>
+    <th>Model</th>
+    <th>Geography</th>
+    <th>Relative strengths</th>
+    <th>Architecture</th>
+  </tr>
+  <tr>
+    <td><code>time_distributed</code></td>
+    <td>Central and West Africa</td>
+    <td>Better than <code>slowfast</code> at duikers, chimps, and gorillas and other larger species</td>
+    <td>Image-based <code>TimeDistributedEfficientNet</code></td>
+  </tr>
+  <tr>
+      <td><code>slowfast</code></td>
+      <td>Central and West Africa</td>
+      <td>Better than <code>time_distributed</code> at blank detection and small species detection</td>
+      <td>Video-native <code>SlowFast</code></td>
+    </tr>
+  <tr>
+    <td><code>european</code></td>
+    <td>Western Europe</td>
+    <td>Trained on non-jungle ecologies</td>
+    <td>Finetuned <code>time_distributed</code>model</td>
+  </tr>
+</table>
+
+All models support training, fine-tuning, and inference. For fine-tuning, we recommend using the `time_distributed` model as the starting point.
+
+<h2 id="species-classes"></h2>
+
+## What species can `zamba` detect?
+
+`time_distributed` and `slowfast` are both trained to identify 32 common species from Central and West Africa. The output labels in these models are:
+
+* `aardvark`
+* `antelope_duiker`
+* `badger`
+* `bat`
+* `bird`
+* `blank`
+* `cattle`
+* `cheetah`
+* `chimpanzee_bonobo`
+* `civet_genet`
+* `elephant`
+* `equid`
+* `forest_buffalo`
+* `fox`
+* `giraffe`
+* `gorilla`
+* `hare_rabbit`
+* `hippopotamus`
+* `hog`
+* `human`
+* `hyena`
+* `large_flightless_bird`
+* `leopard`
+* `lion`
+* `mongoose`
+* `monkey_prosimian`
+* `pangolin`
+* `porcupine`
+* `reptile`
+* `rodent`
+* `small_cat`
+* `wild_dog_jackal`
+
+`european` is trained to identify 11 common species in western Europe. The possible class labels are:
+
+* `bird`
+* `blank`
+* `domestic_cat`
+* `european_badger`
+* `european_beaver`
+* `european_hare`
+* `european_roe_deer`
+* `north_american_raccoon`
+* `red_fox`
+* `unidentified`
+* `weasel`
+* `wild_boar`
+
+<a id='time-distributed'></a>
+
+## `time_distributed` model
+
+### Architecture
+
+The `time_distributed` model was built by re-training a well-known image classification architecture called [EfficientNetV2](https://arxiv.org/abs/1905.11946) (Tan, M., & Le, Q., 2019) to identify the species in our camera trap videos. EfficientNetV2 models are convolutional [neural networks](https://www.youtube.com/watch?v=aircAruvnKk&t=995s) designed to jointly optimize model size and training speed. EfficientNetV2 is image native, meaning it classifies each frame separately when generating predictions. The model is wrapped in a [`TimeDistributed` layer](https://docs.fast.ai/layers.html#TimeDistributed) which enables a single prediction per video.
+
+<a id='time-distributed-training-data'></a>
+
+### Training data
+
+`time_distributed` was trained using data collected and annotated by partners at [The Max Planck Institute for
+Evolutionary Anthropology](https://www.eva.mpg.de/index.html) and [Chimp &
+See](https://www.chimpandsee.org/).
+
+The data included camera trap videos from:
+
+<table>
+  <tr>
+    <th>Country</th>
+    <th>Location</th>
+  </tr>
+  <tr>
+    <td rowspan=2>Cameroon</td>
+    <td>Campo Ma'an National Park</td>
+  </tr>
+  <tr>
+    <td>Korup National Park</td>
+  </tr>
+  <tr>
+    <td>Central African Republic</td>
+    <td>Dzanga-Sangha Protected Area</td>
+  </tr>
+  <tr>
+    <td rowspan=3>Côte d'Ivoire</td>
+    <td>Comoé National Park</td>
+  </tr>
+  <tr>
+    <td>Guiroutou</td>
+  </tr>
+  <tr>
+    <td>Taï National Park</td>
+  </tr>
+  <tr>
+    <td rowspan=2>Democratic Republic of the Congo</td>
+    <td>Bili-Uele Protect Area</td>
+  </tr>
+  <tr>
+    <td>Salonga National Park</td>
+  </tr>
+  <tr>
+    <td rowspan=2>Gabon</td>
+    <td>Loango National Park</td>
+  </tr>
+  <tr>
+    <td>Lopé National Park</td>
+  </tr>
+  <tr>
+    <td rowspan=2>Guinea</td>
+    <td>Bakoun Classified Forest</td>
+  </tr>
+  <tr>
+    <td>Moyen-Bafing National Park</td>
+  </tr>
+  <tr>
+    <td rowspan=3>Liberia</td>
+    <td>East Nimba Nature Reserve</td>
+  </tr>
+  <tr>
+    <td>Grebo-Krahn National Park</td>
+  </tr>
+  <tr>
+    <td>Sapo National Park</td>
+  </tr>
+  <tr>
+    <td>Mozambique</td>
+    <td>Gorongosa National Park</td>
+  </tr>
+  <tr>
+    <td>Nigeria</td>
+    <td>Gashaka-Gumti National Park</td>
+  </tr>
+  <tr>
+    <td rowspan=2>Republic of the Congo</td>
+    <td>Conkouati-Douli National Park</td>
+  </tr>
+  <tr>
+    <td>Nouabale-Ndoki National Park</td>
+  </tr>
+  <tr>
+    <td>Senegal</td>
+    <td>Kayan</td>
+  </tr>
+  <tr>
+    <td rowspan=2>Tanzania</td>
+    <td>Grumeti Game Reserve</td>
+  </tr>
+  <tr>
+    <td>Ugalla River National Park</td>
+  </tr>
+  <tr>
+    <td rowspan=3>Uganda</td>
+    <td>Budongo Forest Reserve</td>
+  </tr>
+  <tr>
+    <td>Bwindi Forest National Park</td>
+  </tr>
+  <tr>
+    <td>Ngogo and Kibale National Park</td>
+  </tr>
+</table>
+
+### Default configuration
+
+The full default configuration is available on [Github](https://github.com/drivendataorg/zamba/blob/master/zamba/models/official_models/time_distributed/config.yaml).
+
+By default, an efficient object detection model called [MegadetectorLite](#megadetectorlite) is run on all frames to determine which are the most likely to contain an animal. Then `time_distributed` is run on only the 16 frames with the highest predicted probability of detection. By default, videos are resized to 240x426 pixels following frame selection.
+
+The default video loading configuration for `time_distributed` is:
+```yaml
+video_loader_config:
+  model_input_height: 240
+  model_input_width: 426
+  crop_bottom_pixels: 50
+  fps: 4
+  total_frames: 16
+  ensure_total_frames: true
+  megadetector_lite_config:
+    confidence: 0.25
+    fill_mode: score_sorted
+    n_frames: 16
+```
+
+You can choose different frame selection methods and vary the size of the images that are used by passing in a custom [YAML configuration file](../yaml-config.md). The only requirement for the `time_distributed` model is that the video loader must return 16 frames.
+
+<a id='slowfast'></a>
+
+## `slowfast` model
+
+### Architecture
+
+The `slowfast` model was built by re-training a video classification backbone called [SlowFast](https://arxiv.org/abs/1812.03982) (Feichtenhofer, C., Fan, H., Malik, J., & He, K., 2019). SlowFast refers to the two model pathways involved: one that operates at a low frame rate to capture spatial semantics, and one that operates at a high frame rate to capture motion over time.
+
+<div style="text-align:center;">
+<img src="https://s3.amazonaws.com/drivendata-public-assets/zamba-slowfast-diagram.png" alt="Architecture showing the two pathways of the slowfast model" style="width:400px;"/>
+<br/><br/>
+<i>Source:</i> Feichtenhofer, C., Fan, H., Malik, J., & He, K. (2019). Slowfast networks for video recognition. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 6202-6211).
+</div>
+
+Unlike `time_distributed`, `slowfast` is video native. This means it takes into account the relationship between frames in a video, rather than running independently on each frame.
+
+### Training data
+
+The `slowfast` model was trained using the same data as the [`time_distributed` model](#time-distributed-training-data).
+
+### Default configuration
+
+The full default configuration is available on [Github](https://github.com/drivendataorg/zamba/blob/master/zamba/models/official_models/slowfast/config.yaml).
+
+By default, an efficient object detection model called [MegadetectorLite](#megadetectorlite) is run on all frames to determine which are the most likely to contain an animal. Then `slowfast` is run on only the 32 frames with the highest predicted probability of detection. By default, videos are resized to 240x426 pixels.
+
+The full default video loading configuration is:
+
+```yaml
+video_loader_config:
+  model_input_height: 240
+  model_input_width: 426
+  crop_bottom_pixels: 50
+  fps: 8
+  total_frames: 32
+  ensure_total_frames: true
+  megadetector_lite_config:
+    confidence: 0.25
+    fill_mode: score_sorted
+    n_frames: 32
+```
+
+You can choose different frame selection methods and vary the size of the images that are used by passing in a custom [YAML configuration file](../yaml-config.md). The two requirements for the `slowfast` model are that:
+- the video loader must return 32 frames
+- videos inputted into the model must be at least 200 x 200 pixels
+
+<a id='european'></a>
+
+## `european` model
+
+### Architecture
+
+The `european` model starts from the trained `time_distributed` model, and then replaces and trains the final output layer to predict European species.
+
+### Training data
+
+The `european` model is finetuned with data collected and annotated by partners at [The Max Planck Institute for Evolutionary Anthropology](https://www.eva.mpg.de/index.html). The finetuning data included camera trap videos from Hintenteiche bei Biesenbrow, Germany.
+
+### Default configuration
+
+The full default configuration is available on [Github](https://github.com/drivendataorg/zamba/blob/master/zamba/models/official_models/european/config.yaml).
+
+The `european` model uses the same frame selection as the `time_distributed` model. By default, an efficient object detection model called [MegadetectorLite](#megadetectorlite) is run on all frames to determine which are the most likely to contain an animal. Then `european` is run on only the 16 frames with the highest predicted probability of detection. By default, videos are resized to 240x426 pixels following frame selection.
+
+The full default video loading configuration is:
+```yaml
+video_loader_config:
+  model_input_height: 240
+  model_input_width: 426
+  crop_bottom_pixels: 50
+  fps: 4
+  total_frames: 16
+  ensure_total_frames: true
+  megadetector_lite_config:
+    confidence: 0.25
+    fill_mode: score_sorted
+    n_frames: 16
+```
+
+As with all models, you can choose different frame selection methods and vary the size of the images that are used by passing in a custom [YAML configuration file](../yaml-config.md). The only requirement for the `european` model is that the video loader must return 16 frames.
+
+<a id='megadetectorlite'></a>
+
+## MegadetectorLite
+
+Frame selection for video models is critical as it would be infeasible to train neural networks on all the frames in a video. For all the species detection models that ship with `zamba`, the default frame selection method is an efficient object detection model called MegadetectorLite that determines the likelihood that each frame contains an animal. Then, only the frames with the highest probability of detection are passed to the model.
+
+MegadetectorLite combines two open-source models:
+
+* [Megadetector](https://github.com/microsoft/CameraTraps/blob/master/megadetector.md) is a pretrained image model designed to detect animals, people, and vehicles in camera trap videos.
+* [YOLOX](https://github.com/Megvii-BaseDetection/YOLOX) is a high-performance, lightweight object detection model that is much less computationally intensive than Megadetector.
+
+While highly accurate, Megadetector is too computationally intensive to run on every frame. MegadetectorLite was created by training a YOLOX model using the predictions of the Megadetector as ground truth - this method is called [student-teacher training](https://towardsdatascience.com/knowledge-distillation-simplified-dd4973dbc764).
diff --git a/docs/docs/predict-tutorial.md b/docs/docs/predict-tutorial.md
index 4bea5638..f788c862 100644
--- a/docs/docs/predict-tutorial.md
+++ b/docs/docs/predict-tutorial.md
@@ -1,4 +1,4 @@
-# User Tutorial: Classifying Unlabeled Videos
+# User tutorial: Classifying unlabeled videos
 
 This section walks through how to classify videos using `zamba`. If you are new to `zamba` and just want to classify some videos as soon as possible, see the [Quickstart](quickstart.md) guide.
 
@@ -20,10 +20,10 @@ $ zamba predict --data-dir example_vids/
 
 ### Required arguments
 
-To run `zamba predict` in the command line, you must specify either `--data-dir` or `--filepaths`. 
+To run `zamba predict` in the command line, you must specify `--data-dir` and/or `--filepaths`.
 
 * **`--data-dir PATH`:** Path to the folder containing your videos.
-* **`--filepaths PATH`:** Path to a CSV file with a column for the filepath to each video you want to classify. The CSV must have a column for `filepath`.
+* **`--filepaths PATH`:** Path to a CSV file with a column for the filepath to each video you want to classify. The CSV must have a column for `filepath`. Filepaths can be absolute or relative to the data directory.
 
 All other flags are optional. To choose a model, either `--model` or `--checkpoint` must be specified. Use `--model` to specify one of the three [pretrained models](models/index.md) that ship with `zamba`. Use `--checkpoint` to run inference with a locally saved model. `--model` defaults to `time_distributed`.
 
@@ -37,27 +37,27 @@ Minimum example for prediction using the Python package:
 from zamba.models.model_manager import predict_model
 from zamba.models.config import PredictConfig
 
-predict_config = PredictConfig(data_directory="example_vids/")
+predict_config = PredictConfig(data_dir="example_vids/")
 predict_model(predict_config=predict_config)
 ```
 
-The only two arguments that can be passed to `predict_model` are `predict_config` and (optionally) `video_loader_config`. The first step is to instantiate [`PredictConfig`](configurations.md#prediction-arguments). Optionally, you can also specify video loading arguments by instantiating and passing in [`VideoLoaderConfig`](configurations.md#video-loading-arguments). 
+The only two arguments that can be passed to `predict_model` are `predict_config` and (optionally) `video_loader_config`. The first step is to instantiate [`PredictConfig`](configurations.md#prediction-arguments). Optionally, you can also specify video loading arguments by instantiating and passing in [`VideoLoaderConfig`](configurations.md#video-loading-arguments).
 
 ### Required arguments
 
-To run `predict_model` in Python, you must specify either `data_directory` or `filepaths` when `PredictConfig` is instantiated.
+To run `predict_model` in Python, you must specify either `data_dir` or `filepaths` when `PredictConfig` is instantiated.
 
-* **`data_directory (DirectoryPath)`:** Path to the folder containing your videos.
+* **`data_dir (DirectoryPath)`:** Path to the folder containing your videos.
 
-* **`filepaths (FilePath)`:** Path to a CSV file with a column for the filepath to each video you want to classify. The CSV must have a column for `filepath`.
+* **`filepaths (FilePath)`:** Path to a CSV file with a column for the filepath to each video you want to classify. The CSV must have a column for `filepath`. Filepaths can be absolute or relative to the data directory.
 
 For detailed explanations of all possible configuration arguments, see [All Optional Arguments](configurations.md).
 
 ## Default behavior
 
-By default, the [`time_distributed`](models/index.md#time-distributed) model will be used. `zamba` will output a `.csv` file with rows labeled by each video filename and columns for each class (ie. species). The default prediction will store all class probabilities, so that cell (i,j) can be interpreted as *the probability that animal j is present in video i.* 
+By default, the [`time_distributed`](models/index.md#time-distributed) model will be used. `zamba` will output a `.csv` file with rows labeled by each video filename and columns for each class (ie. species). The default prediction will store all class probabilities, so that cell (i,j) can be interpreted as *the probability that animal j is present in video i.*
 
-By default, predictions will be saved to `zamba_predictions.csv`. You can save predictions to a custom path using the `--save-path` argument.
+By default, predictions will be saved to `zamba_predictions.csv` in your working directory. You can save predictions to a custom directory using the `--save-dir` argument.
 
 ```console
 $ cat zamba_predictions.csv
@@ -80,7 +80,7 @@ $ zamba predict --config predict_configuration.yaml
 Save all of your videos within one folder.
 
 * They can be in nested subdirectories within the folder.
-* Your videos should all be saved in formats that are suppored by FFmpeg, [which are listed here](https://www.ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features). Any videos that fail a set of FFmpeg checks will be skipped during inference or training.
+* Your videos should be in be saved in formats that are suppored by FFmpeg, [which are listed here](https://www.ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features). Any videos that fail a set of FFmpeg checks will be skipped during inference or training. By default, `zamba` will look for files with the following suffixes: `.avi`, `.mp4`, `.asf`. To use other video suffixes that are supported by FFmpeg, set your `VIDEO_SUFFIXES` environment variable.
 
 Add the path to your video folder. For example, if your videos are in a folder called `example_vids`:
 
@@ -90,16 +90,13 @@ Add the path to your video folder. For example, if your videos are in a folder c
     ```
 === "Python"
     ```python
-    predict_config = PredictConfig(data_directory='example_vids/')
+    predict_config = PredictConfig(data_dir='example_vids/')
     predict_model(predict_config=predict_config)
     ```
 
 ### 2. Choose a model for prediction
 
-If your camera videos contain species common to central or west Africa:
-
-* Use the [`time_distributed` model](models/index.md#time-distributed) if your priority is species classification
-* Use the [`slowfast` model](models/index.md#slowfast) if your priority is blank vs. non-blank video detection
+If your camera videos contain species common to Central or West Africa, use either the [`time_distributed` model](models/index.md#time-distributed) or [`slowfast` model](models/index.md#slowfast) model. `slowfast` is better for blank and small species detection. `time_distributed` performs better if you have many different species of interest, or are focused on duikers, chimpanzees, and/or gorillas.
 
 If your videos contain species common to Europe, use the [`european` model](models/index.md#european).
 
@@ -112,7 +109,7 @@ Add the model name to your command. The `time_distributed` model will be used if
 === "Python"
     ```python
     predict_config = PredictConfig(
-        data_directory='example_vids/', model_name='slowfast'
+        data_dir='example_vids/', model_name='slowfast'
     )
     predict_model(predict_config=predict_config)
     ```
@@ -121,8 +118,8 @@ Add the model name to your command. The `time_distributed` model will be used if
 
 There are three options for how to format predictions, listed from most information to least:
 
-1. **Store all probabilities (default):** Return predictions with a row for each filename and a column for each class label, with probabilities between 0 and 1. Cell (i,j) is the probability that animal j is present in video i.
-2. **Presence/absence:** Return predictions with a row for each filename and a column for each class label, with cells indicating either presence or absense based on a user-specified probability threshold. Cell (i, j) indicates whether animal j is present (`1`) or not present (`0`) in video i. The probability threshold cutoff is specified with `--proba-threshold` in the CLI. 
+1. **Store all probabilities (default):** Return predictions with a row for each filename and a column for each class label, with probabilities between 0 and 1. Cell `(i,j)` is the probability that animal `j` is present in video `i`.
+2. **Presence/absence:** Return predictions with a row for each filename and a column for each class label, with cells indicating either presence or absense based on a user-specified probability threshold. Cell `(i, j)` indicates whether animal `j` is present (`1`) or not present (`0`) in video `i`. The probability threshold cutoff is specified with `--proba-threshold` in the CLI.
 3. **Most likely class:** Return predictions with a row for each filename and one column for the most likely class in each video. The most likely class can also be blank. To get the most likely class, add `--output-class-names` to your command. In Python, it can be specified by adding `output_class_names=True` when `PredictConfig` is instantiated. This is not recommended if you'd like to detect more than one species in each video.
 
 Say we want to generate predictions for the videos in `example_vids` indicating which animals are present in each video based on a probability threshold of 50%:
@@ -140,7 +137,7 @@ Say we want to generate predictions for the videos in `example_vids` indicating
 === "Python"
     ```python
     predict_config = PredictConfig(
-        data_directory="example_vids/", proba_threshold=0.5
+        data_dir="example_vids/", proba_threshold=0.5
     )
     predict_model(predict_config=predict_config)
     predictions = pd.read_csv("zamba_predictions.csv")
@@ -156,8 +153,8 @@ Say we want to generate predictions for the videos in `example_vids` indicating
 
 ### 4. Specify any additional parameters
 
-And there's so much more! You can also do things like specify your region for faster model download (`--weight-download-region`), use a saved model checkpoint (`--checkpoint`), or specify a different path where your predictions should be saved (`--save`). To read about a few common considerations, see the [Guide to Common Optional Parameters](extra-options.md) page.
+And there's so much more! You can also do things like specify your region for faster model download (`--weight-download-region`), use a saved model checkpoint (`--checkpoint`), or specify a different folder where your predictions should be saved (`--save-dir`). To read about a few common considerations, see the [Guide to Common Optional Parameters](extra-options.md) page.
 
 ### 5. Test your configuration with a dry run
 
-Before kicking off a full run of inference, we recommend testing your code with a "dry run". This will run one batch of inference to quickly detect any bugs. See the [Debugging](debugging.md) page for details.
\ No newline at end of file
+Before kicking off a full run of inference, we recommend testing your code with a "dry run". This will run one batch of inference to quickly detect any bugs. See the [Debugging](debugging.md) page for details.
diff --git a/docs/docs/python-tutorial.ipynb b/docs/docs/python-tutorial.ipynb
deleted file mode 100644
index 09021bea..00000000
--- a/docs/docs/python-tutorial.ipynb
+++ /dev/null
@@ -1,423 +0,0 @@
-{
- "cells": [
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# User Tutorial: Using `zamba` as a Python Package\n",
-    "\n",
-    "***\n",
-    "\n",
-    "Before getting started, you need to have `zamba` and it's prerequisites installed. See the [Installation](https://zamba.drivendata.org/docs/install.md) page for details."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 1,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# %load_ext nb_black\n"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 2,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# import packages\n",
-    "from pathlib import Path"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Classifying unlabeled videos\n",
-    "\n",
-    "Say we want to generate predictions for the videos in a folder called `example_vids` in our home directory."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 8,
-   "metadata": {},
-   "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "[PosixPath('/Users/katewetstone/example_vids/chimp.MP4'),\n",
-       " PosixPath('/Users/katewetstone/example_vids/eleph.MP4'),\n",
-       " PosixPath('/Users/katewetstone/example_vids/blank.MP4'),\n",
-       " PosixPath('/Users/katewetstone/example_vids/leopard.MP4')]"
-      ]
-     },
-     "execution_count": 8,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "video_dir = Path('/Users/katewetstone/example_vids')\n",
-    "list(video_dir.iterdir())"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "There are four videos in the folder: `chimp.MP4`, `eleph.MP4`, `blank.MP4`, and `leopard.MP4`. For clarity, here we've named the videos based on what we know each contains. In the real world, your videos will probably have much less helpful names."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 9,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# import from zamba\n",
-    "from zamba.models.model_manager import predict_model\n",
-    "from zamba.models.config import PredictConfig"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 10,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Help on class PredictConfig in module zamba.models.config:\n",
-      "\n",
-      "class PredictConfig(ZambaBaseModel)\n",
-      " |  PredictConfig(*, data_directory: pydantic.types.DirectoryPath = PosixPath('/Users/katewetstone/Repos/zamba/docs/docs'), filepaths: pydantic.types.FilePath = None, checkpoint: pydantic.types.FilePath = None, model_name: zamba.models.config.ModelEnum = <ModelEnum.time_distributed: 'time_distributed'>, gpus: int = 0, num_workers: int = 3, batch_size: int = 8, save: Union[bool, pathlib.Path] = True, dry_run: bool = False, proba_threshold: float = None, output_class_names: bool = False, weight_download_region: zamba.models.utils.RegionEnum = 'us', cache_dir: pathlib.Path = None, skip_load_validation: bool = False) -> None\n",
-      " |  \n",
-      " |  Configuration for using a model for inference.\n",
-      " |  \n",
-      " |  Args:\n",
-      " |      filepaths (FilePath): Path to a CSV containing videos for inference, with\n",
-      " |          one row per video in the data_directory. There must be a column called\n",
-      " |          'filepath' (absolute or relative to the data_directory). If None, uses\n",
-      " |          all files in data_directory. Defaults to None.\n",
-      " |      data_directory (DirectoryPath): Path to a directory containing videos for\n",
-      " |          inference. Defaults to the working directory.\n",
-      " |      model_name (str, optional): Name of the model to use for inference. Options\n",
-      " |          are: time_distributed, slowfast, european. Defaults to time_distributed.\n",
-      " |      checkpoint (FilePath, optional): Path to a custom checkpoint file (.ckpt)\n",
-      " |          generated by zamba that can be used to generate predictions. If None,\n",
-      " |          defaults to a pretrained model. Defaults to None.\n",
-      " |      gpus (int): Number of GPUs to use for inference.\n",
-      " |          Defaults to all of the available GPUs found on the machine.\n",
-      " |      num_workers (int): Number of subprocesses to use for data loading. 0 means\n",
-      " |          that the data will be loaded in the main process. Defauts to one less\n",
-      " |          than the number of CPUs in the system, or 1 if there is one CPU in the\n",
-      " |          system.\n",
-      " |      batch_size (int): Batch size to use for inference. Defaults to 8.\n",
-      " |      save (bool or Path): Path to a CSV to save predictions. If True is passed,\n",
-      " |          \"zamba_predictions.csv\" is written to the current working directory.\n",
-      " |          If False is passed, predictions are not saved. Defaults to True.\n",
-      " |      dry_run (bool): Perform inference on a single batch for testing. Predictions\n",
-      " |          will not be saved. Defaults to False.\n",
-      " |      proba_threshold (float, optional): Probability threshold for classification.\n",
-      " |          If specified, binary predictions are returned with 1 being greater than the\n",
-      " |          threshold and 0 being less than or equal to the threshold. If None, return\n",
-      " |          probability scores for each species. Defaults to None.\n",
-      " |      output_class_names (bool): Output the species with the highest probability\n",
-      " |          score as a single prediction for each video. If False, return probabilty\n",
-      " |          scores for each species. Defaults to False.\n",
-      " |      weight_download_region (str): s3 region to download pretrained weights from.\n",
-      " |          Options are \"us\" (United States), \"eu\" (European Union), or \"asia\"\n",
-      " |          (Asia Pacific). Defaults to \"us\".\n",
-      " |      cache_dir (Path, optional): Cache directory where downloaded model weights\n",
-      " |          will be saved. If None and no environment variable is set, will use your\n",
-      " |          default cache directory. Defaults to None.\n",
-      " |      skip_load_validation (bool). By default, zamba runs a check to verify that\n",
-      " |          all videos can be loaded and skips files that cannot be loaded. This can\n",
-      " |          be time intensive, depending on how many videos there are. If you are very\n",
-      " |          confident all your videos can be loaded, you can set this to True and skip\n",
-      " |          this check. Defaults to False.\n",
-      " |  \n",
-      " |  Method resolution order:\n",
-      " |      PredictConfig\n",
-      " |      ZambaBaseModel\n",
-      " |      pydantic.main.BaseModel\n",
-      " |      pydantic.utils.Representation\n",
-      " |      builtins.object\n",
-      " |  \n",
-      " |  Class methods defined here:\n",
-      " |  \n",
-      " |  get_filepaths(values) from pydantic.main.ModelMetaclass\n",
-      " |      If no file list is passed, get all files in data directory. Warn if there\n",
-      " |      are unsupported suffixes. Filepaths is set to a dataframe, where column `filepath`\n",
-      " |      contains files with valid suffixes.\n",
-      " |  \n",
-      " |  validate_dry_run_and_save(values) from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  validate_files(values) from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  validate_proba_threshold(values) from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  validate_save(values) from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  ----------------------------------------------------------------------\n",
-      " |  Static methods defined here:\n",
-      " |  \n",
-      " |  __json_encoder__ = pydantic_encoder(obj: Any) -> Any\n",
-      " |  \n",
-      " |  ----------------------------------------------------------------------\n",
-      " |  Data and other attributes defined here:\n",
-      " |  \n",
-      " |  __abstractmethods__ = frozenset()\n",
-      " |  \n",
-      " |  __annotations__ = {'batch_size': <class 'int'>, 'cache_dir': typing.Un...\n",
-      " |  \n",
-      " |  __class_vars__ = set()\n",
-      " |  \n",
-      " |  __config__ = <class 'zamba.models.config.Config'>\n",
-      " |  \n",
-      " |  \n",
-      " |  __custom_root_type__ = False\n",
-      " |  \n",
-      " |  __fields__ = {'batch_size': ModelField(name='batch_size', type=int, re...\n",
-      " |  \n",
-      " |  __hash__ = None\n",
-      " |  \n",
-      " |  __post_root_validators__ = [(True, <function PredictConfig.validate_dr...\n",
-      " |  \n",
-      " |  __pre_root_validators__ = []\n",
-      " |  \n",
-      " |  __private_attributes__ = {}\n",
-      " |  \n",
-      " |  __schema_cache__ = {}\n",
-      " |  \n",
-      " |  __signature__ = <Signature (*, data_directory: pydantic.types.Di...one...\n",
-      " |  \n",
-      " |  __validators__ = {'cache_dir': [<pydantic.class_validators.Validator o...\n",
-      " |  \n",
-      " |  ----------------------------------------------------------------------\n",
-      " |  Data and other attributes inherited from ZambaBaseModel:\n",
-      " |  \n",
-      " |  Config = <class 'zamba.models.config.ZambaBaseModel.Config'>\n",
-      " |  \n",
-      " |  \n",
-      " |  ----------------------------------------------------------------------\n",
-      " |  Methods inherited from pydantic.main.BaseModel:\n",
-      " |  \n",
-      " |  __eq__(self, other: Any) -> bool\n",
-      " |  \n",
-      " |  __getstate__(self) -> 'DictAny'\n",
-      " |  \n",
-      " |  __init__(__pydantic_self__, **data: Any) -> None\n",
-      " |      Create a new model by parsing and validating input data from keyword arguments.\n",
-      " |      \n",
-      " |      Raises ValidationError if the input data cannot be parsed to form a valid model.\n",
-      " |  \n",
-      " |  __iter__(self) -> 'TupleGenerator'\n",
-      " |      so `dict(model)` works\n",
-      " |  \n",
-      " |  __repr_args__(self) -> 'ReprArgs'\n",
-      " |  \n",
-      " |  __setattr__(self, name, value)\n",
-      " |  \n",
-      " |  __setstate__(self, state: 'DictAny') -> None\n",
-      " |  \n",
-      " |  copy(self: 'Model', *, include: Union[ForwardRef('AbstractSetIntStr'), ForwardRef('MappingIntStrAny')] = None, exclude: Union[ForwardRef('AbstractSetIntStr'), ForwardRef('MappingIntStrAny')] = None, update: 'DictStrAny' = None, deep: bool = False) -> 'Model'\n",
-      " |      Duplicate a model, optionally choose which fields to include, exclude and change.\n",
-      " |      \n",
-      " |      :param include: fields to include in new model\n",
-      " |      :param exclude: fields to exclude from new model, as with values this takes precedence over include\n",
-      " |      :param update: values to change/add in the new model. Note: the data is not validated before creating\n",
-      " |          the new model: you should trust this data\n",
-      " |      :param deep: set to `True` to make a deep copy of the model\n",
-      " |      :return: new model instance\n",
-      " |  \n",
-      " |  dict(self, *, include: Union[ForwardRef('AbstractSetIntStr'), ForwardRef('MappingIntStrAny')] = None, exclude: Union[ForwardRef('AbstractSetIntStr'), ForwardRef('MappingIntStrAny')] = None, by_alias: bool = False, skip_defaults: bool = None, exclude_unset: bool = False, exclude_defaults: bool = False, exclude_none: bool = False) -> 'DictStrAny'\n",
-      " |      Generate a dictionary representation of the model, optionally specifying which fields to include or exclude.\n",
-      " |  \n",
-      " |  json(self, *, include: Union[ForwardRef('AbstractSetIntStr'), ForwardRef('MappingIntStrAny')] = None, exclude: Union[ForwardRef('AbstractSetIntStr'), ForwardRef('MappingIntStrAny')] = None, by_alias: bool = False, skip_defaults: bool = None, exclude_unset: bool = False, exclude_defaults: bool = False, exclude_none: bool = False, encoder: Union[Callable[[Any], Any], NoneType] = None, **dumps_kwargs: Any) -> 'unicode'\n",
-      " |      Generate a JSON representation of the model, `include` and `exclude` arguments as per `dict()`.\n",
-      " |      \n",
-      " |      `encoder` is an optional function to supply as `default` to json.dumps(), other arguments as per `json.dumps()`.\n",
-      " |  \n",
-      " |  ----------------------------------------------------------------------\n",
-      " |  Class methods inherited from pydantic.main.BaseModel:\n",
-      " |  \n",
-      " |  __get_validators__() -> 'CallableGenerator' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  construct(_fields_set: Union[ForwardRef('SetStr'), NoneType] = None, **values: Any) -> 'Model' from pydantic.main.ModelMetaclass\n",
-      " |      Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data.\n",
-      " |      Default values are respected, but no other validation is performed.\n",
-      " |      Behaves as if `Config.extra = 'allow'` was set since it adds all passed values\n",
-      " |  \n",
-      " |  from_orm(obj: Any) -> 'Model' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  parse_file(path: Union[str, pathlib.Path], *, content_type: 'unicode' = None, encoding: 'unicode' = 'utf8', proto: pydantic.parse.Protocol = None, allow_pickle: bool = False) -> 'Model' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  parse_obj(obj: Any) -> 'Model' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  parse_raw(b: Union[str, bytes], *, content_type: 'unicode' = None, encoding: 'unicode' = 'utf8', proto: pydantic.parse.Protocol = None, allow_pickle: bool = False) -> 'Model' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  schema(by_alias: bool = True, ref_template: 'unicode' = '#/definitions/{model}') -> 'DictStrAny' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  schema_json(*, by_alias: bool = True, ref_template: 'unicode' = '#/definitions/{model}', **dumps_kwargs: Any) -> 'unicode' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  update_forward_refs(**localns: Any) -> None from pydantic.main.ModelMetaclass\n",
-      " |      Try to update ForwardRefs on fields based on this Model, globalns and localns.\n",
-      " |  \n",
-      " |  validate(value: Any) -> 'Model' from pydantic.main.ModelMetaclass\n",
-      " |  \n",
-      " |  ----------------------------------------------------------------------\n",
-      " |  Data descriptors inherited from pydantic.main.BaseModel:\n",
-      " |  \n",
-      " |  __dict__\n",
-      " |      dictionary for instance variables (if defined)\n",
-      " |  \n",
-      " |  __fields_set__\n",
-      " |  \n",
-      " |  ----------------------------------------------------------------------\n",
-      " |  Methods inherited from pydantic.utils.Representation:\n",
-      " |  \n",
-      " |  __pretty__(self, fmt: Callable[[Any], Any], **kwargs: Any) -> Generator[Any, NoneType, NoneType]\n",
-      " |      Used by devtools (https://python-devtools.helpmanual.io/) to provide a human readable representations of objects\n",
-      " |  \n",
-      " |  __repr__(self) -> 'unicode'\n",
-      " |  \n",
-      " |  __repr_name__(self) -> 'unicode'\n",
-      " |      Name of the instance's class, used in __repr__.\n",
-      " |  \n",
-      " |  __repr_str__(self, join_str: 'unicode') -> 'unicode'\n",
-      " |  \n",
-      " |  __str__(self) -> 'unicode'\n",
-      "\n"
-     ]
-    }
-   ],
-   "source": [
-    "help(PredictConfig)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "#### Simple prediction\n",
-    "\n",
-    "Predict the most likely class for each video, and save the results to `example_predictions.csv`"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 11,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "2021-10-18 13:34:57.135 | INFO     | zamba.models.config:get_filepaths:664 - Getting files in example_vids.\n",
-      "2021-10-18 13:34:57.137 | INFO     | zamba.models.config:get_filepaths:685 - Found 4 videos in example_vids.\n",
-      "2021-10-18 13:34:57.142 | INFO     | zamba.models.config:check_files_exist_and_load:118 - Checking all 4 filepaths exist. Can take up to a minute for every couple thousand files.\n",
-      "2021-10-18 13:34:57.144 | INFO     | zamba.models.config:check_files_exist_and_load:143 - Checking that all videos can be loaded. If you're very confident all your videos can be loaded, you can skip this with `skip_load_validation`, but it's not recommended.\n",
-      "100%|██████████| 4/4 [00:00<00:00, 12.59it/s]\n"
-     ]
-    }
-   ],
-   "source": [
-    "# set up our prediction configuration\n",
-    "predict_config = PredictConfig(data_directory=\"example_vids/\", \n",
-    "save=\"example_predictions.csv\", \n",
-    "output_class_names=True)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 12,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "2021-10-18 13:35:00.031 | INFO     | zamba.models.config:get_default_video_loader_config:754 - No video loader config specified. Using default for time_distributed.\n",
-      "2021-10-18 13:35:00.040 | INFO     | zamba.models.model_manager:instantiate_model:74 - Downloading weights for model.\n"
-     ]
-    },
-    {
-     "ename": "ClientError",
-     "evalue": "An error occurred (AccessDenied) when calling the ListObjects operation: Access Denied",
-     "output_type": "error",
-     "traceback": [
-      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
-      "\u001b[0;31mClientError\u001b[0m                               Traceback (most recent call last)",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/cloudpathlib/s3/s3client.py\u001b[0m in \u001b[0;36m_s3_file_query\u001b[0;34m(self, cloud_path)\u001b[0m\n\u001b[1;32m    138\u001b[0m             \u001b[0mobj\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0ms3\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mObject\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcloud_path\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbucket\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcloud_path\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 139\u001b[0;31m             \u001b[0mobj\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mload\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    140\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mobj\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/boto3/resources/factory.py\u001b[0m in \u001b[0;36mdo_action\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m    504\u001b[0m             \u001b[0;32mdef\u001b[0m \u001b[0mdo_action\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 505\u001b[0;31m                 \u001b[0mresponse\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0maction\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    506\u001b[0m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmeta\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mresponse\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/boto3/resources/action.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, parent, *args, **kwargs)\u001b[0m\n\u001b[1;32m     82\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 83\u001b[0;31m         \u001b[0mresponse\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparent\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmeta\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moperation_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mparams\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     84\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/botocore/client.py\u001b[0m in \u001b[0;36m_api_call\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m    387\u001b[0m             \u001b[0;31m# The \"self\" in this scope is referring to the BaseClient.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 388\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_api_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moperation_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    389\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/botocore/client.py\u001b[0m in \u001b[0;36m_make_api_call\u001b[0;34m(self, operation_name, api_params)\u001b[0m\n\u001b[1;32m    707\u001b[0m             \u001b[0merror_class\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexceptions\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfrom_code\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0merror_code\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 708\u001b[0;31m             \u001b[0;32mraise\u001b[0m \u001b[0merror_class\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparsed_response\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moperation_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    709\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;31mClientError\u001b[0m: An error occurred (403) when calling the HeadObject operation: Forbidden",
-      "\nDuring handling of the above exception, another exception occurred:\n",
-      "\u001b[0;31mClientError\u001b[0m                               Traceback (most recent call last)",
-      "\u001b[0;32m/var/folders/hh/_nbd8pkn08x5ty_mx954bjvm0000gn/T/ipykernel_11531/1464135739.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mpredict_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpredict_config\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mpredict_config\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
-      "\u001b[0;32m~/Repos/zamba/zamba/models/model_manager.py\u001b[0m in \u001b[0;36mpredict_model\u001b[0;34m(predict_config, video_loader_config)\u001b[0m\n\u001b[1;32m    329\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    330\u001b[0m     \u001b[0;31m# set up model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 331\u001b[0;31m     model = instantiate_model(\n\u001b[0m\u001b[1;32m    332\u001b[0m         \u001b[0mcheckpoint\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mpredict_config\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheckpoint\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    333\u001b[0m         \u001b[0mweight_download_region\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mpredict_config\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mweight_download_region\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m~/Repos/zamba/zamba/models/model_manager.py\u001b[0m in \u001b[0;36minstantiate_model\u001b[0;34m(checkpoint, weight_download_region, scheduler_config, cache_dir, labels, from_scratch, predict_all_zamba_species)\u001b[0m\n\u001b[1;32m     73\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mPath\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcheckpoint\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexists\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     74\u001b[0m         \u001b[0mlogger\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minfo\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Downloading weights for model.\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 75\u001b[0;31m         checkpoint = download_weights(\n\u001b[0m\u001b[1;32m     76\u001b[0m             \u001b[0mfilename\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mstr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcheckpoint\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     77\u001b[0m             \u001b[0mweight_region\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mweight_download_region\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m~/Repos/zamba/zamba/models/utils.py\u001b[0m in \u001b[0;36mdownload_weights\u001b[0;34m(filename, destination_dir, weight_region)\u001b[0m\n\u001b[1;32m     31\u001b[0m         \u001b[0mclient\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mS3Client\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlocal_cache_dir\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdestination_dir\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mno_sign_request\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     32\u001b[0m     )\n\u001b[0;32m---> 33\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0ms3p\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfspath\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/cloudpathlib/cloudpath.py\u001b[0m in \u001b[0;36mfspath\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    304\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0mproperty\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    305\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mfspath\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 306\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__fspath__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    307\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    308\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mglob\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpattern\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mIterable\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"CloudPath\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/cloudpathlib/cloudpath.py\u001b[0m in \u001b[0;36m__fspath__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    208\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    209\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__fspath__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 210\u001b[0;31m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_file\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    211\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_refresh_cache\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mforce_overwrite_from_cloud\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    212\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_local\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/cloudpathlib/s3/s3path.py\u001b[0m in \u001b[0;36mis_file\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m     37\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     38\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mis_file\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mbool\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 39\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclient\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_is_file_or_dir\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"file\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     40\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     41\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mmkdir\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparents\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mexist_ok\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/cloudpathlib/s3/s3client.py\u001b[0m in \u001b[0;36m_is_file_or_dir\u001b[0;34m(self, cloud_path)\u001b[0m\n\u001b[1;32m    119\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    120\u001b[0m         \u001b[0;31m# get first item by listing at least one key\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 121\u001b[0;31m         \u001b[0ms3_obj\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_s3_file_query\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcloud_path\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    122\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    123\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0ms3_obj\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/cloudpathlib/s3/s3client.py\u001b[0m in \u001b[0;36m_s3_file_query\u001b[0;34m(self, cloud_path)\u001b[0m\n\u001b[1;32m    142\u001b[0m         \u001b[0;31m# else, confirm it is a dir by filtering to the first item under the prefix\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    143\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mClientError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 144\u001b[0;31m             return next(\n\u001b[0m\u001b[1;32m    145\u001b[0m                 (\n\u001b[1;32m    146\u001b[0m                     \u001b[0mobj\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/cloudpathlib/s3/s3client.py\u001b[0m in \u001b[0;36m<genexpr>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m    143\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mClientError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    144\u001b[0m             return next(\n\u001b[0;32m--> 145\u001b[0;31m                 (\n\u001b[0m\u001b[1;32m    146\u001b[0m                     \u001b[0mobj\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    147\u001b[0m                     for obj in (\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/boto3/resources/collection.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m     81\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     82\u001b[0m         \u001b[0mcount\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 83\u001b[0;31m         \u001b[0;32mfor\u001b[0m \u001b[0mpage\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpages\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     84\u001b[0m             \u001b[0;32mfor\u001b[0m \u001b[0mitem\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mpage\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     85\u001b[0m                 \u001b[0;32myield\u001b[0m \u001b[0mitem\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/boto3/resources/collection.py\u001b[0m in \u001b[0;36mpages\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    164\u001b[0m         \u001b[0;31m# we start processing and yielding individual items.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    165\u001b[0m         \u001b[0mcount\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 166\u001b[0;31m         \u001b[0;32mfor\u001b[0m \u001b[0mpage\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mpages\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    167\u001b[0m             \u001b[0mpage_items\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    168\u001b[0m             \u001b[0;32mfor\u001b[0m \u001b[0mitem\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_handler\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_parent\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mparams\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/botocore/paginate.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    253\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_inject_starting_params\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcurrent_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    254\u001b[0m         \u001b[0;32mwhile\u001b[0m \u001b[0;32mTrue\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 255\u001b[0;31m             \u001b[0mresponse\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_request\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcurrent_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    256\u001b[0m             \u001b[0mparsed\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_extract_parsed_response\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mresponse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    257\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mfirst_request\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/botocore/paginate.py\u001b[0m in \u001b[0;36m_make_request\u001b[0;34m(self, current_kwargs)\u001b[0m\n\u001b[1;32m    332\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    333\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_make_request\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcurrent_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 334\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_method\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mcurrent_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    335\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    336\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_extract_parsed_response\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresponse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/botocore/client.py\u001b[0m in \u001b[0;36m_api_call\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m    386\u001b[0m                     \"%s() only accepts keyword arguments.\" % py_operation_name)\n\u001b[1;32m    387\u001b[0m             \u001b[0;31m# The \"self\" in this scope is referring to the BaseClient.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 388\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_api_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moperation_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    389\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    390\u001b[0m         \u001b[0m_api_call\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__name__\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpy_operation_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;32m/opt/anaconda3/envs/zamba/lib/python3.8/site-packages/botocore/client.py\u001b[0m in \u001b[0;36m_make_api_call\u001b[0;34m(self, operation_name, api_params)\u001b[0m\n\u001b[1;32m    706\u001b[0m             \u001b[0merror_code\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mparsed_response\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Error\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Code\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    707\u001b[0m             \u001b[0merror_class\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexceptions\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfrom_code\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0merror_code\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 708\u001b[0;31m             \u001b[0;32mraise\u001b[0m \u001b[0merror_class\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mparsed_response\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moperation_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    709\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    710\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mparsed_response\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
-      "\u001b[0;31mClientError\u001b[0m: An error occurred (AccessDenied) when calling the ListObjects operation: Access Denied"
-     ]
-    }
-   ],
-   "source": [
-    "predict_model(predict_config=predict_config)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "outline\n",
-    "\n",
-    "- requirements\n",
-    "- generating predictions for unlabeled videos\n",
-    "- training a model on labeled videos"
-   ]
-  }
- ],
- "metadata": {
-  "interpreter": {
-   "hash": "7c5f4cd3bd54a8e2ae4ec218923447b01b0f54366b28f342e656c452f99042eb"
-  },
-  "kernelspec": {
-   "display_name": "Python 3.8.12 64-bit ('zamba': conda)",
-   "name": "python3"
-  },
-  "language_info": {
-   "codemirror_mode": {
-    "name": "ipython",
-    "version": 3
-   },
-   "file_extension": ".py",
-   "mimetype": "text/x-python",
-   "name": "python",
-   "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython3",
-   "version": "3.8.12"
-  },
-  "orig_nbformat": 4
- },
- "nbformat": 4,
- "nbformat_minor": 2
-}
diff --git a/docs/docs/quickstart.md b/docs/docs/quickstart.md
index c861a049..5fce19d0 100644
--- a/docs/docs/quickstart.md
+++ b/docs/docs/quickstart.md
@@ -18,7 +18,7 @@ macOS, this can be done in the terminal (⌘+space, "Terminal"). On Windows, thi
 
 ## How do I organize my videos for `zamba`?
 
-You can input the path to a directory of videos or specify a list of file paths. `zamba` supports the same video formats as FFmpeg, [which are listed here](https://www.ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features). Any videos that fail a set of FFmpeg checks will be skipped during inference or training.
+You can specify the path to a directory of videos or specify a list of filepaths in a `.csv` file. `zamba` supports the same video formats as FFmpeg, [which are listed here](https://www.ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features). Any videos that fail a set of FFmpeg checks will be skipped during inference or training.
 
 For example, say we have a directory of videos called `example_vids` that we want to generate predictions for using `zamba`. Let's list the videos:
 
@@ -64,8 +64,9 @@ To generate and save predictions for your videos using the default settings, run
 $ zamba predict --data-dir example_vids/
 ```
 
-`zamba` will output a `.csv` file with rows labeled by each video filename and columns for each class (ie. species). The default prediction will store all class probabilities, so that cell (i,j) is *the probability that animal j is present in video i.*  Comprehensive predictions are helpful when a single video contains multiple species.
-Predictions will be saved to `zamba_predictions.csv` in the current working directory by default. You can save out predictions under a different name or in a different folder using the `--save-path` argument.
+`zamba` will output a `.csv` file with rows labeled by each video filename and columns for each class (ie. species). The default prediction will store all class probabilities, so that cell `(i,j)` is *the probability that animal `j` is present in video `i`.*  Comprehensive predictions are helpful when a single video contains multiple species.
+
+Predictions will be saved to `zamba_predictions.csv` in the current working directory by default. You can save out predictions to a different folder using the `--save-dir` argument.
 
 Adding the argument `--output-class-names` will simplify the predictions to return only the *most likely* animal in each video:
 
@@ -88,8 +89,8 @@ $ zamba predict --data-dir example_vids/ --model slowfast
 
 You can continue training one of the [models](models/index.md) that ships with `zamba` by either:
 
-* Fine-tuning with additional labeled videos where the species are included in the list of [`zamba` class labels](models/index.md#species-classes)
-* Fine-tuning with labeled videos that include new species
+* Finetuning with additional labeled videos where the species are included in the list of [`zamba` class labels](models/index.md#species-classes)
+* Finetuning with labeled videos that include new species
 
 In either case, the commands for training are the same. Say that we have labels for the videos in the `example_vids` folder saved in `example_labels.csv`. To train a model, run:
 
@@ -108,31 +109,32 @@ eleph.MP4,elephant
 leopard.MP4,leopard
 ```
 
-By default, the trained model and additional training output will be saved to a folder in the current working directory called `zamba_{model_name}`. For example, a model finetuned from the provided `time_distributed` model will be saved in `zamba_time_distributed`. 
+By default, the trained model and additional training output will be saved to a `version_n` folder in the current working directory. For example,
 
 ```console
 $ zamba train --data-dir example_vids/ --labels example_labels.csv
-$ ls zamba_time_distributed
-time_distributed.ckpt 
+$ ls version_0/
+hparams.yaml
+time_distributed.ckpt
+train_configuration.yamml
+val_metrics.json
 ...
 ```
 
 ## Downloading model weights
 
-**`zamba` needs to download the "weights" files for the neural networks that it uses to make predictions. On first run it will download ~200-500 MB of files with these weights depending which model you choose.** 
-Once a model's weights are downloaded, the tool will use the local version and will not need to perform this download again. If you are not in the US, we recommend running the above command with the additional flag either `--weight_download_region eu` or `--weight_download_region asia` depending on your location. The closer you are to the server the faster the downloads will be.
+**`zamba` needs to download the "weights" files for the models it uses to make predictions. On first run, it will download ~200-500 MB of files with these weights depending which model you choose.**
+Once a model's weights are downloaded, `zamba` will use the local version and will not need to perform this download again. If you are not in the United States, we recommend running the above command with the additional flag either `--weight_download_region eu` or `--weight_download_region asia` depending on your location. The closer you are to the server, the faster the downloads will be.
 
 <a id='getting-help'></a>
 
 ## Getting help
 
-Once zamba is installed, you can see more details of each function with `--help`. 
+Once zamba is installed, you can see more details of each function with `--help`.
 
 To get help with `zamba predict`:
 
 ```console
-$ zamba predict --help
-
 Usage: zamba predict [OPTIONS]
 
   Identify species in a video.
@@ -159,11 +161,13 @@ Options:
                                   specifiied, will use all GPUs found on
                                   machine.
   --batch-size INTEGER            Batch size to use for training.
-  --save / --no-save              Whether to save out predictions to a csv
-                                  file. If you want to specify the location of
-                                  the csv, use save_path instead.
-  --save-path PATH                Full path for prediction CSV file. Any
-                                  needed parent directories will be created.
+  --save / --no-save              Whether to save out predictions. If you want
+                                  to specify the output directory, use
+                                  save_dir instead.
+  --save-dir PATH                 An optional directory in which to save the
+                                  model predictions and configuration yaml.
+                                  Defaults to the current working directory if
+                                  save is True.
   --dry-run / --no-dry-run        Runs one batch of inference to check for
                                   bugs.
   --config PATH                   Specify options using yaml configuration
@@ -190,6 +194,8 @@ Options:
                                   loaded prior to inference. Only use if
                                   you're very confident all your videos can be
                                   loaded.
+  -o, --overwrite                 Overwrite outputs in the save directory if
+                                  they exist.
   -y, --yes                       Skip confirmation of configuration and
                                   proceed right to prediction.
   --help                          Show this message and exit.
@@ -225,11 +231,10 @@ Options:
                                   machine.
   --dry-run / --no-dry-run        Runs one batch of train and validation to
                                   check for bugs.
-  --save-dir PATH                 Directory in which to save model checkpoint
-                                  and configuration file. If not specified,
-                                  will save to a folder called
-                                  'zamba_{model_name}' in your working
-                                  directory.
+  --save-dir PATH                 An optional directory in which to save the
+                                  model checkpoint and configuration file. If
+                                  not specified, will save to a `version_n`
+                                  folder in your working directory.
   --num-workers INTEGER           Number of subprocesses to use for data
                                   loading.
   --weight-download-region [us|eu|asia]
diff --git a/docs/docs/train-tutorial.md b/docs/docs/train-tutorial.md
index d13977e2..dfdd0dbd 100644
--- a/docs/docs/train-tutorial.md
+++ b/docs/docs/train-tutorial.md
@@ -1,4 +1,4 @@
-# User Tutorial: Training a Model on Labaled Videos
+# User tutorial: Training a model on labaled videos
 
 This section walks through how to train a model using `zamba`. If you are new to `zamba` and just want to classify some videos as soon as possible, see the [Quickstart](quickstart.md) guide.
 
@@ -9,14 +9,14 @@ This tutorial goes over the steps for using `zamba` if:
 
 `zamba` can run two types of model training:
 
-* Fine-tuning a model with labels that are a subset of the possible [zamba labels](models/index.md#species-classes)
-* Fine-tuning a model to predict an entirely new set of labels
+* Finetuning a model with labels that are a subset of the possible [zamba labels](models/index.md#species-classes)
+* Finetuning a model to predict an entirely new set of labels
 
 The process is the same for both cases.
 
 ## Basic usage: command line interface
 
-Say that we want to finetune the `time_distributed` model based on the videos in `example_vids` and the labels in `example_labels.csv`. 
+Say that we want to finetune the `time_distributed` model based on the videos in `example_vids` and the labels in `example_labels.csv`.
 
 Minimum example for training in the command line:
 
@@ -43,7 +43,7 @@ leopard.MP4,leopard
 
 ## Basic usage: Python package
 
-Say that we want to finetune the `time_distributed` model based on the videos in `example_vids` and the labels in `example_labels.csv`. 
+Say that we want to finetune the `time_distributed` model based on the videos in `example_vids` and the labels in `example_labels.csv`.
 
 Minimum example for training using the Python package:
 
@@ -52,66 +52,51 @@ from zamba.models.model_manager import train_model
 from zamba.models.config import TrainConfig
 
 train_config = TrainConfig(
-    data_directory="example_vids/", labels="example_labels.csv"
+    data_dir="example_vids/", labels="example_labels.csv"
 )
 train_model(train_config=train_config)
 ```
 
-The only two arguments that can be passed to `train_model` are `train_config` and (optionally) `video_loader_config`. The first step is to instantiate [`TrainConfig`](configurations.md#training-arguments). Optionally, you can also specify video loading arguments by instantiating and passing in [`VideoLoaderConfig`](configurations.md#video-loading-arguments). 
+The only two arguments that can be passed to `train_model` are `train_config` and (optionally) `video_loader_config`. The first step is to instantiate [`TrainConfig`](configurations.md#training-arguments). Optionally, you can also specify video loading arguments by instantiating and passing in [`VideoLoaderConfig`](configurations.md#video-loading-arguments).
 
 ### Required arguments
 
-To run `train_model` in Python, you must specify both `data_directory` and `labels` when `TrainConfig` is instantiated.
+To run `train_model` in Python, you must specify both `data_dir` and `labels` when `TrainConfig` is instantiated.
 
-* **`data_directory (DirectoryPath)`:** Path to the folder containing your videos.
+* **`data_dir (DirectoryPath)`:** Path to the folder containing your videos.
 
 * **`labels (FilePath or pd.DataFrame)`:** Either the path to a CSV file with labels for training, or a dataframe of the training labels. There must be columns for `filename` and `label`.
 
-For detailed explanations of all possible configuration arguments, see [All Optional Arguments](configurations.md).
+For detailed explanations of all possible configuration arguments, see [All Configuration Options](configurations.md).
 
 ## Default behavior
 
-By default, the [`time_distributed`](models/index.md#time-distributed) model will be used as a starting point. The newly trained model will be saved to a folder in the current working directory called `zamba_{model_name}`. For example, a model finetuned from the provided `time_distributed` model (the default) will be saved in `zamba_time_distributed`. 
-
-`zamba_time_distributed` contains:
-
-* `train_configuration.yaml`: The full model configuration used to generate the given model, including `video_loader_config` and `train_config`. To continue training using the same configuration, or to train another model using the same configuration, you can pass in `train_configurations.yaml` (see [Specifying Model Configurations with a YAML File](yaml-config.md)).
-* `hparams.yaml`: Model hyperparameters. For example, the YAML file below tells us that the model was trained with a learning rate (`lr`) of 0.001:
-    ```yaml
-    $ cat zamba_time_distributed/hparams.yaml
-
-    lr: 0.001
-    model_class: TimeDistributedEfficientNetMultiLayerHead
-    num_frames: 16
-    scheduler: MultiStepLR
-    scheduler_params:
-    gamma: 0.5
-    milestones:
-    - 3
-    verbose: true
-    species:
-    - species_blank
-    - species_chimpanzee_bonobo
-    - species_elephant
-    - species_leopard
+By default, the [`time_distributed`](models/index.md#time-distributed) model will be used as a starting point. You can specify where the outputs should be saved with `--save-dir`. If no save directory is specified, `zamba` will write out incremental `version_n` folders to your current working directory. For example, a model finetuned from the provided `time_distributed` model (the default) will be saved in `version_0`.
+
+`version_0` contains:
+
+* `train_configuration.yaml`: The full model configuration used to generate the given model, including `video_loader_config` and `train_config`. To continue training using the same configuration, or to train another model using the same configuration, you can pass in `train_configurations.yaml` (see [Specifying Model Configurations with a YAML File](yaml-config.md)) along with the `labels` filepath.
+* `hparams.yaml`: Model hyperparameters. These are included in the checkpoint file as well.
+* `time_distributed.ckpt`: Model checkpoint. You can continue training from this checkpoint by passing it to `zamba train` with the `--checkpoint` flag:
+    ```console
+    $ zamba train --checkpoint version_0/time_distributed.ckpt --data-dir example_vids/ --labels example_labels.csv
     ```
-* `time_distributed.ckpt`: Model checkpoint. The model checkpoint also includes both the model configuration in `train_configuration.yaml` and the model hyperparameters in `hparams.yaml`. You can continue training from this checkpoint by passing it to `zamba train` with the `--checkpoint` flag:
+* `events.out.tfevents.1632250686.ip-172-31-15-179.14229.0`: [TensorBoard](https://www.tensorflow.org/tensorboard/get_started) logs. You can view these with tensorboard:
     ```console
-    $ zamba train --checkpoint time_distributed.ckpt --data-dir example_vids/ --labels example_labels.csv
+    $ tensorboard --logdir version_0/
     ```
-* `events.out.tfevents.1632250686.ip-172-31-15-179.14229.0`: [TensorBoard](https://www.tensorflow.org/tensorboard/get_started) logs
-* `test_metrics.json`: The model's performance on the test subset
 * `val_metrics.json`: The model's performance on the validation subset
+* `test_metrics.json`: The model's performance on the test (holdout) subset
 * `splits.csv`: Which files were used for training, validation, and as a holdout set. If split is specified in the labels file passed to training, `splits.csv` will not be saved out.
 
 ## Step-by-step tutorial
 
-### 1. Specify the path to your videos 
+### 1. Specify the path to your videos
 
-Save all of your videos within one folder.
+Save all of your videos in a folder.
 
 * They can be in nested directories within the folder.
-* Your videos should all be saved in formats that are suppored by FFmpeg, [which are listed here](https://www.ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features). Any videos that fail a set of FFmpeg checks will be skipped during inference or training.
+* Your videos should all be saved in formats that are suppored by FFmpeg, [which are listed here](https://www.ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features). Any videos that fail a set of FFmpeg checks will be skipped during inference or training. By default, `zamba` will look for files with the following suffixes: `.avi`, `.mp4`, `.asf`. To use other video suffixes that are supported by FFmpeg, set your `VIDEO_SUFFIXES` environment variable.
 
 Add the path to your video folder with `--data-dir`. For example, if your videos are in a folder called `example_vids`, add `--data-dir example_vids/` to your command.
 
@@ -122,10 +107,10 @@ Add the path to your video folder with `--data-dir`. For example, if your videos
 
 === "Python"
     ```python
-    from zamba.models.model_manager import train_model
     from zamba.models.config import TrainConfig
+    from zamba.models.model_manager import train_model
 
-    train_config = TrainConfig(data_directory='example_vids/')
+    train_config = TrainConfig(data_dir='example_vids/')
     train_model(train_config=train_config)
     ```
 Note that the above will not run yet because labels are not specified.
@@ -156,7 +141,7 @@ Add the path to your labels with `--labels`.  For example, if your videos are in
     ```python
     labels_dataframe = pd.read_csv('example_labels.csv', index_col='filepath')
     train_config = TrainConfig(
-        data_directory='example_vids/', labels=labels_dataframe
+        data_dir='example_vids/', labels=labels_dataframe
     )
     train_model(train_config=train_config)
     ```
@@ -167,11 +152,13 @@ Your labels may be included in the list of [`zamba` class labels](models/index.m
 
 #### Completely new labels
 
-You can also train a model to predict completely new labels - the world is your oyster! (We'd love to see a model trained to predict oysters.) If this is the case, the model architecture will replace the final [neural network](https://www.youtube.com/watch?v=aircAruvnKk&t=995s) layer with a new head that predicts *your* labels instead of those that ship with `zamba`. [Backpropogation](https://www.youtube.com/watch?v=Ilg3gGewQ5U) will continue from that point with the new head. This process is called [transfer learning](https://keras.io/guides/transfer_learning/).
+You can also train a model to predict completely new labels - the world is your oyster! (We'd love to see a model trained to predict oysters.) If this is the case, the model architecture will replace the final [neural network](https://www.youtube.com/watch?v=aircAruvnKk&t=995s) layer with a new head that predicts *your* labels instead of those that ship with `zamba`.
 
 ### 3. Choose a model for training
 
-If your videos contain species common to central or west Africa, use the [`time_distributed` model](models/index.md#time-distributed). If they contain species common to western Europe, use the [`european` model](models/index.md#european). We do not recommend using the [`slowfast` model](models/index.md#slowfast) for training because it is much more computationally intensive and slower to run.
+Any of the models that ship with `zamba` can be trained. If you're training on entirely new species or new ecologies, we recommend starting with the [`time_distributed` model](models/index.md#time-distributed) as this model is less computationally intensive than the [`slowfast` model](models/index.md#slowfast).
+
+However, if you're tuning a model to a subset of species (e.g. a `european_beaver` or `blank` model), use the model that was trained on data that is most similar to your new data.
 
 Add the model name to your command with `--model`. The `time_distributed` model will be used if no model is specified. For example, if you want to continue training the `european` model based on the videos in `example_euro_vids` and the labels in `example_euro_labels.csv`:
 
@@ -182,7 +169,7 @@ Add the model name to your command with `--model`. The `time_distributed` model
 === "Python"
     ```python
     train_config = TrainConfig(
-        data_directory="example_euro_vids/",
+        data_dir="example_euro_vids/",
         labels="example_euro_labels.csv",
         model_name="european",
     )
@@ -191,8 +178,8 @@ Add the model name to your command with `--model`. The `time_distributed` model
 
 ### 4. Specify any additional parameters
 
-And there's so much more! You can also do things like specify your region for faster model download (`--weight-download-region`), start training from a saved model checkpoint (`--checkpoint`), or specify a different path where your model should be saved (`--save-directory`). To read about a few common considerations, see the [Guide to Common Optional Parameters](extra-options.md) page.
+And there's so much more! You can also do things like specify your region for faster model download (`--weight-download-region`), start training from a saved model checkpoint (`--checkpoint`), or specify a different path where your model should be saved (`--save-dir`). To read about a few common considerations, see the [Guide to Common Optional Parameters](extra-options.md) page.
 
 ### 5. Test your configuration with a dry run
 
-Before kicking off the full model training, we recommend testing your code with a "dry run". This will run one training and validation batch for one epoch to quickly detect any bugs. See the [Debugging](debugging.md) page for details.
\ No newline at end of file
+Before kicking off the full model training, we recommend testing your code with a "dry run". This will run one training and validation batch for one epoch to quickly detect any bugs. See the [Debugging](debugging.md) page for details.
diff --git a/docs/docs/yaml-config.md b/docs/docs/yaml-config.md
index f8273daa..f965073d 100644
--- a/docs/docs/yaml-config.md
+++ b/docs/docs/yaml-config.md
@@ -1,4 +1,4 @@
-# Using YAML Configuration Files
+# Using YAML configuration files
 
 In both the command line and the Python module, options for video loading, training, and prediction can be set by passing a YAML file instead of passing arguments directly. YAML files (`.yml` or `.yaml`) are commonly used to serialize data in an easily readable way.
 
@@ -12,15 +12,15 @@ video_loader_config:
   total_frames: 16
   # other video loading parameters
 
-predict_config:
+train_config:
   model_name: time_distributed
-  data_directoty: example_vids/
+  data_dir: example_vids/
+  labels: example_labels.csv
   # other training parameters, eg. batch_size
 
-train_config:
+predict_config:
   model_name: time_distributed
-  data_directory: example_vids/
-  labels: example_labels.csv
+  data_directoty: example_vids/
   # other training parameters, eg. batch_size
 ```
 
@@ -39,15 +39,13 @@ predict_config:
 
 ## Required arguments
 
-Either `predict_config` or `train_config` is required, based on whether you will be running inference or training a model. See [All Optional Arguments](configurations.md) for a full list of what can be specified under each class. To run inference, either `data_directory` or `filepaths` must be specified. To train a model, both `data_directory` and `labels` must be specified.
+Either `predict_config` or `train_config` is required, based on whether you will be running inference or training a model. See [All Configuration Options](configurations.md) for a full list of what can be specified under each class. To run inference, `data_dir`and/or `filepaths` must be specified. To train a model, `labels` must be specified.
 
-In `video_loader_config`, you must specify at least `model_input_height`, `model_input_width`, and `total_frames`. 
+In `video_loader_config`, you must specify at least `model_input_height`, `model_input_width`, and `total_frames`. While this is the minimum required, we strongly recommend being intentional in your choice of frame selection method. `total_frames` by itself will just take the first `n` frames. For a full list of frame selection methods, see the section on [Video loading arguments](configurations.md#video-loading-arguments).
 
 * For `time_distributed` or `european`, `total_frames` must be 16
 * For `slowfast`, `total_frames` must be 32
 
-See the [Available Models](models/index.md) page for more details on each model's requirements.
-
 ## Command line interface
 
 A YAML configuration file can be passed to the command line interface with the `--config` argument. For example, say the example configuration above is saved as `example_config.yaml`. To run prediction:
@@ -66,36 +64,11 @@ The main API for zamba is the [`ModelManager` class](api-reference/models-model_
 from zamba.models.manager import ModelManager
 ```
 
-The `ModelManager` class is used by `zamba`’s command line interface to handle preprocessing the filenames, loading the videos, serving them to the model, and saving predictions. Therefore any functionality available to the command line interface is accessible via the `ModelManager` class.
+The `ModelManager` class is used by `zamba`’s command line interface to handle preprocessing the filenames, loading the videos, training the model, performing inference, and saving predictions. Therefore any functionality available to the command line interface is accessible via the `ModelManager` class.
 
 To instantiate the `ModelManager` based on a configuration file saved at `test_config.yaml`:
 ```python
 >>> manager = ModelManager.from_yaml('test_config.yaml')
->>> manager.config
-
-ModelConfig(
-  video_loader_config=VideoLoaderConfig(crop_bottom_pixels=None, i_frames=False, 
-                                        scene_threshold=None, megadetector_lite_config=None, 
-                                        model_input_height=240, model_input_width=426, 
-                                        total_frames=16, ensure_total_frames=True, 
-                                        fps=None, early_bias=False, frame_indices=None,
-                                        evenly_sample_total_frames=False, pix_fmt='rgb24'
-                                      ), 
-  train_config=None, 
-  predict_config=PredictConfig(data_directory=PosixPath('vids'), 
-                               filepaths=                         filepath
-                                          0    /home/ubuntu/vids/eleph.MP4
-                                          1  /home/ubuntu/vids/leopard.MP4
-                                          2    /home/ubuntu/vids/blank.MP4
-                                          3    /home/ubuntu/vids/chimp.MP4, 
-                                checkpoint='zamba_time_distributed.ckpt', 
-                                model_params=ModelParams(scheduler=None, scheduler_params=None),
-                                model_name='time_distributed', species=None, 
-                                gpus=1, num_workers=3, batch_size=8, 
-                                save=True, dry_run=False, proba_threshold=None,
-                                output_class_names=False, weight_download_region='us', 
-                                cache_dir=None, skip_load_validation=False)
-                              )
 ```
 
 We can now run inference or model training without specifying any additional parameters, because they are already associated with our instance of the `ModelManager` class. To run inference or training:
@@ -116,16 +89,6 @@ For example, the default configuration for the [`time_distributed` model](models
 
 ```yaml
 train_config:
-  model_name: time_distributed
-  backbone_finetune_config:
-    backbone_initial_ratio_lr: 0.01
-    multiplier: 1
-    pre_train_bn: True
-    train_bn: False
-    unfreeze_backbone_at_epoch: 3
-    verbose: True
-  early_stopping_config:
-    patience: 5
   scheduler_config:
     scheduler: MultiStepLR
     scheduler_params:
@@ -133,65 +96,32 @@ train_config:
       milestones:
       - 3
       verbose: true
-
+  model_name: time_distributed
+  backbone_finetune_config:
+    backbone_initial_ratio_lr: 0.01
+    multiplier: 1
+    pre_train_bn: true
+    train_bn: false
+    unfreeze_backbone_at_epoch: 3
+    verbose: true
+  early_stopping_config:
+    patience: 5
 video_loader_config:
   model_input_height: 240
   model_input_width: 426
   crop_bottom_pixels: 50
   fps: 4
   total_frames: 16
-  ensure_total_frames: True
+  ensure_total_frames: true
   megadetector_lite_config:
     confidence: 0.25
     fill_mode: score_sorted
     n_frames: 16
-
 predict_config:
   model_name: time_distributed
+public_checkpoint: time_distributed_9e710aa8c92d25190a64b3b04b9122bdcb456982.ckpt
 ```
 
-For reference, the below shows how to specify the same video loading and training parameters using only the Python package:
+## Templates
 
-```python
-from zamba.data.video import VideoLoaderConfig
-from zamba.models.config import TrainConfig
-from zamba.models.model_manager import train_model
-
-video_loader_config = VideoLoaderConfig(
-    model_input_height=240,
-    model_input_width=426,
-    crop_bottom_pixels=50,
-    fps=4,
-    total_frames=16,
-    ensure_total_frames=True,
-    megadetector_lite_config={
-        "confidence": 0.25, 
-        "fill_mode": "score_sorted", 
-        "n_frames": 16,
-    },
-)
-
-train_config = TrainConfig(
-    # data_directory=YOUR_DATA_DIRECTORY_HERE,
-    # labels=YOUR_LABELS_CSV_HERE,
-    model_name="time_distributed",
-    backbone_finetune_config={
-        "backbone_initial_ratio_lr": 0.01,
-        "unfreeze_backbone_at_epoch": 3,
-        "verbose": True,
-        "pre_train_bn": True,
-        "train_bn": False,
-        "multiplier": 1,
-    },
-    early_stopping_config={"patience": 5},
-    scheduler_config={
-        "scheduler": "MultiStepLR",
-        "scheduler_params": {"gamma": 0.5, "milestones": 3, "verbose": True,},
-    },
-)
-
-train_model(
-    train_config=train_config,
-    video_loader_config=video_loader_config,
-)
-```
\ No newline at end of file
+To make modifying existing mod easier, we've set up the official models as templates in the [`templates` folder](https://github.com/drivendataorg/zamba/tree/master/templates). Just fill in your data directory and labels, make any desired tweaks to the model config, and then kick off some [training](train_tutorial.md). Happy modeling!
diff --git a/docs/mkdocs.yml b/docs/mkdocs.yml
index 319c8a9f..61fcf87b 100644
--- a/docs/mkdocs.yml
+++ b/docs/mkdocs.yml
@@ -14,35 +14,39 @@ extra_css:
 
 nav:
   - Home: "index.md"
-  - "Getting Started":
-      - Installing Zamba: "install.md"
+  - "Getting started":
+      - Installing zamba: "install.md"
       - Quickstart: "quickstart.md"
   - "User Tutorials":
-      - Classifying Unlabeled Videos: "predict-tutorial.md"
-      - Training a Model on Labeled Videos: "train-tutorial.md"
+      - Classifying unlabeled videos: "predict-tutorial.md"
+      - Training a model on labeled videos: "train-tutorial.md"
       - Debugging: "debugging.md"
+      - Guide to common optional parameters: "extra-options.md"
   - "Available Models":
-      - "models/index.md"
+      - Species detection: "models/species-detection.md"
+      - DensePose: "models/denspose.md"
   - "Advanced Options":
-      - All Optional Arguments: "configurations.md"
+      - All configuration options: "configurations.md"
       - Using YAML configuration files: "yaml-config.md"
-      - Guide to Common Optional Parameters: "extra-options.md"
-  - "Contribute to zamba": 
+  - "Contribute to zamba":
       - "contribute/index.md"
   - "Changelog":
-      - "changelog/index.md"
+      - "changelog.md"
   - API Reference:
       - zamba.data:
           - zamba.data.metadata: "api-reference/data-metadata.md"
           - zamba.data.video: "api-reference/data-video.md"
       - zamba.models:
-          - zamba.models.yolox_models: "api-reference/models-yolox_models.md"
           - zamba.models.config: "api-reference/models-config.md"
+          - zamba.models.densepose:
+            - zamba.models.densepose.config: "api-reference/densepose_config.md"
+            - zamba.models.densepose.densepose_manager: "api-reference/densepose_manager.md"
           - zamba.models.efficientnet_models: "api-reference/models-efficientnet_models.md"
           - zamba.models.megadetector_lite_yolox: "api-reference/models-megadetector_lite_yolox.md"
           - zamba.models.model_manager: "api-reference/models-model_manager.md"
           - zamba.models.slowfast_models: "api-reference/models-slowfast_models.md"
           - zamba.models.utils: "api-reference/models-utils.md"
+          - zamba.models.yolox_models: "api-reference/models-yolox_models.md"
       - zamba.pytorch:
           - zamba.pytorch.dataloaders: "api-reference/pytorch-dataloaders.md"
           - zamba.pytorch.finetuning: "api-reference/pytorch-finetuning.md"
diff --git a/templates/european.yaml b/templates/european.yaml
index fff5681b..9448ef16 100644
--- a/templates/european.yaml
+++ b/templates/european.yaml
@@ -1,5 +1,5 @@
 train_config:
-  # data_directory: YOUR_DATA_DIRECTORY_HERE
+  # data_dir: YOUR_DATA_DIR HERE
   # labels: YOUR_LABELS_CSV_HERE
   model_name: european
   backbone_finetune_config:
@@ -9,6 +9,8 @@ train_config:
     train_bn: false
     unfreeze_backbone_at_epoch: 15
     verbose: true
+  early_stopping_config:
+    patience: 3
 
 video_loader_config:
   model_input_height: 240
@@ -23,7 +25,7 @@ video_loader_config:
     n_frames: 16
 
 predict_config:
-  # data_directory: YOUR_DATA_DIRECTORY_HERE
+  # data_dir: YOUR_DATA_DIR HERE
   # or
   # filepaths: YOUR_FILEPATH_CSV_HERE
   model_name: european
diff --git a/templates/slowfast.yaml b/templates/slowfast.yaml
index 046b6a3d..518c7e6b 100644
--- a/templates/slowfast.yaml
+++ b/templates/slowfast.yaml
@@ -1,5 +1,5 @@
 train_config:
-  # data_directory: YOUR_DATA_DIRECTORY_HERE
+  # data_dir: YOUR_DATA_DIR HERE
   # labels: YOUR_LABELS_CSV_HERE
   model_name: slowfast
   backbone_finetune_config:
@@ -32,7 +32,7 @@ video_loader_config:
     n_frames: 32
 
 predict_config:
-  # data_directory: YOUR_DATA_DIRECTORY_HERE
+  # data_dir: YOUR_DATA_DIR HERE
   # or
   # filepaths: YOUR_FILEPATH_CSV_HERE
   model_name: slowfast
diff --git a/templates/time_distributed.yaml b/templates/time_distributed.yaml
index cc036f66..f89662fb 100644
--- a/templates/time_distributed.yaml
+++ b/templates/time_distributed.yaml
@@ -1,14 +1,14 @@
 train_config:
-  # data_directory: YOUR_DATA_DIRECTORY_HERE
+  # data_dir: YOUR_DATA_DIR HERE
   # labels: YOUR_LABELS_CSV_HERE
   model_name: time_distributed
   backbone_finetune_config:
     backbone_initial_ratio_lr: 0.01
     multiplier: 1
-    pre_train_bn: True
-    train_bn: False
+    pre_train_bn: true
+    train_bn: false
     unfreeze_backbone_at_epoch: 3
-    verbose: True
+    verbose: true
   early_stopping_config:
     patience: 5
   scheduler_config:
@@ -25,14 +25,14 @@ video_loader_config:
   crop_bottom_pixels: 50
   fps: 4
   total_frames: 16
-  ensure_total_frames: True
+  ensure_total_frames: true
   megadetector_lite_config:
     confidence: 0.25
     fill_mode: score_sorted
     n_frames: 16
 
 predict_config:
-  # data_directory: YOUR_DATA_DIRECTORY_HERE
+  # data_dir: YOUR_DATA_DIR HERE
   # or
   # filepaths: YOUR_FILEPATH_CSV_HERE
   model_name: time_distributed
diff --git a/tests/assets/sample_predict_config.yaml b/tests/assets/sample_predict_config.yaml
index bfa3603e..af081300 100644
--- a/tests/assets/sample_predict_config.yaml
+++ b/tests/assets/sample_predict_config.yaml
@@ -7,5 +7,5 @@ video_loader_config:
   total_frames: 16
 
 predict_config:
-  data_directory: tests/assets/videos
+  data_dir: tests/assets/videos
   model_name: time_distributed
diff --git a/tests/assets/sample_train_config.yaml b/tests/assets/sample_train_config.yaml
index 1ec30b1b..b3b16215 100644
--- a/tests/assets/sample_train_config.yaml
+++ b/tests/assets/sample_train_config.yaml
@@ -7,7 +7,7 @@ video_loader_config:
   total_frames: 16
 
 train_config:
-  data_directory: tests/assets/videos
+  data_dir: tests/assets/videos
   labels: tests/assets/labels.csv
   model_name: time_distributed
   predict_all_zamba_species: True
diff --git a/tests/conftest.py b/tests/conftest.py
index ca814de7..3e21dd5b 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -161,14 +161,14 @@ def dummy_train_config(labels_absolute_path, dummy_checkpoint, tmp_path_factory)
     tmp_path = tmp_path_factory.mktemp("dummy-model-dir")
     return DummyTrainConfig(
         labels=labels_absolute_path,
-        data_directory=TEST_VIDEOS_DIR,
+        data_dir=TEST_VIDEOS_DIR,
         model_name="dummy",
         checkpoint=dummy_checkpoint,
         max_epochs=1,
         batch_size=1,
         auto_lr_find=False,
         num_workers=2,
-        save_directory=tmp_path / "my_model",
+        save_dir=tmp_path / "my_model",
         skip_load_validation=True,
     )
 
diff --git a/tests/test_cli.py b/tests/test_cli.py
index 78f3c7c8..fcd45302 100644
--- a/tests/test_cli.py
+++ b/tests/test_cli.py
@@ -145,13 +145,21 @@ def test_predict_specific_options(mocker, minimum_valid_predict, tmp_path):  # n
     )
     assert result.exit_code == 0
 
+    # test save overwrite
+    (tmp_path / "zamba_predictions.csv").touch()
+    result = runner.invoke(
+        app,
+        minimum_valid_predict + ["--output-class-names", "--save-dir", str(tmp_path), "-o"],
+    )
+    assert result.exit_code == 0
+
 
 def test_actual_prediction_on_single_video(tmp_path):  # noqa: F811
     data_dir = tmp_path / "videos"
     data_dir.mkdir()
     shutil.copy(TEST_VIDEOS_DIR / "data" / "raw" / "benjamin" / "04250002.MP4", data_dir)
 
-    save_path = tmp_path / "zamba" / "my_preds.csv"
+    save_dir = tmp_path / "zamba"
 
     result = runner.invoke(
         app,
@@ -162,17 +170,20 @@ def test_actual_prediction_on_single_video(tmp_path):  # noqa: F811
             "--config",
             str(ASSETS_DIR / "sample_predict_config.yaml"),
             "--yes",
-            "--save-path",
-            str(save_path),
+            "--save-dir",
+            str(save_dir),
         ],
     )
     assert result.exit_code == 0
     # check preds file got saved out
-    assert save_path.exists()
+    assert save_dir.exists()
     # check config got saved out too
-    assert (save_path.parent / "predict_configuration.yaml").exists()
+    assert (save_dir / "predict_configuration.yaml").exists()
     assert (
-        pd.read_csv(save_path, index_col="filepath").idxmax(axis=1).values[0] == "monkey_prosimian"
+        pd.read_csv(save_dir / "zamba_predictions.csv", index_col="filepath")
+        .idxmax(axis=1)
+        .values[0]
+        == "monkey_prosimian"
     )
 
 
diff --git a/tests/test_config.py b/tests/test_config.py
index 3b3abfc4..5d4cb0d6 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -20,7 +20,7 @@
 
 def test_train_data_dir_only():
     with pytest.raises(ValidationError) as error:
-        TrainConfig(data_directory=TEST_VIDEOS_DIR)
+        TrainConfig(data_dir=TEST_VIDEOS_DIR)
     # labels is missing
     assert error.value.errors() == [
         {"loc": ("labels",), "msg": "field required", "type": "value_error.missing"}
@@ -29,19 +29,19 @@ def test_train_data_dir_only():
 
 def test_train_data_dir_and_labels(tmp_path, labels_relative_path, labels_absolute_path):
     # correct data dir
-    config = TrainConfig(data_directory=TEST_VIDEOS_DIR, labels=labels_relative_path)
-    assert config.data_directory is not None
+    config = TrainConfig(data_dir=TEST_VIDEOS_DIR, labels=labels_relative_path)
+    assert config.data_dir is not None
     assert config.labels is not None
 
     # data dir ignored if absolute path provided in filepath
-    config = TrainConfig(data_directory=tmp_path, labels=labels_absolute_path)
-    assert config.data_directory is not None
+    config = TrainConfig(data_dir=tmp_path, labels=labels_absolute_path)
+    assert config.data_dir is not None
     assert config.labels is not None
     assert not config.labels.filepath.str.startswith(str(tmp_path)).any()
 
     # incorrect data dir with relative filepaths
     with pytest.raises(ValidationError) as error:
-        TrainConfig(data_directory=ASSETS_DIR, labels=labels_relative_path)
+        TrainConfig(data_dir=ASSETS_DIR, labels=labels_relative_path)
     assert "None of the video filepaths exist" in error.value.errors()[0]["msg"]
 
 
@@ -51,8 +51,8 @@ def test_train_labels_only(labels_absolute_path):
 
 
 def test_predict_data_dir_only():
-    config = PredictConfig(data_directory=TEST_VIDEOS_DIR)
-    assert config.data_directory == TEST_VIDEOS_DIR
+    config = PredictConfig(data_dir=TEST_VIDEOS_DIR)
+    assert config.data_dir == TEST_VIDEOS_DIR
     assert isinstance(config.filepaths, pd.DataFrame)
     assert sorted(config.filepaths.filepath.values) == sorted(
         [str(f) for f in TEST_VIDEOS_DIR.rglob("*") if f.is_file()]
@@ -62,14 +62,14 @@ def test_predict_data_dir_only():
 
 def test_predict_data_dir_and_filepaths(labels_absolute_path, labels_relative_path):
     # correct data dir
-    config = PredictConfig(data_directory=TEST_VIDEOS_DIR, filepaths=labels_relative_path)
-    assert config.data_directory is not None
+    config = PredictConfig(data_dir=TEST_VIDEOS_DIR, filepaths=labels_relative_path)
+    assert config.data_dir is not None
     assert config.filepaths is not None
     assert config.filepaths.filepath.str.startswith(str(TEST_VIDEOS_DIR)).all()
 
     # incorrect data dir
     with pytest.raises(ValidationError) as error:
-        PredictConfig(data_directory=ASSETS_DIR, filepaths=labels_relative_path)
+        PredictConfig(data_dir=ASSETS_DIR, filepaths=labels_relative_path)
     assert "None of the video filepaths exist" in error.value.errors()[0]["msg"]
 
 
@@ -203,18 +203,16 @@ def test_labels_with_invalid_split(labels_absolute_path):
 
 
 def test_labels_no_splits(labels_no_splits, tmp_path):
-    config = TrainConfig(
-        data_directory=TEST_VIDEOS_DIR, labels=labels_no_splits, save_directory=tmp_path
-    )
+    config = TrainConfig(data_dir=TEST_VIDEOS_DIR, labels=labels_no_splits, save_dir=tmp_path)
     assert set(config.labels.split.unique()) == set(("holdout", "train", "val"))
 
 
 def test_labels_split_proportions(labels_no_splits, tmp_path):
     config = TrainConfig(
-        data_directory=TEST_VIDEOS_DIR,
+        data_dir=TEST_VIDEOS_DIR,
         labels=labels_no_splits,
         split_proportions={"a": 3, "b": 1},
-        save_directory=tmp_path,
+        save_dir=tmp_path,
     )
     assert config.labels.split.value_counts().to_dict() == {"a": 14, "b": 5}
 
@@ -230,9 +228,18 @@ def test_from_scratch(labels_absolute_path):
     assert "If from_scratch=True, model_name cannot be None." == error.value.errors()[0]["msg"]
 
 
-def test_predict_dry_run_and_save(labels_absolute_path, caplog):
-    PredictConfig(filepaths=labels_absolute_path, dry_run=True, save=True)
-    assert "Cannot save when predicting with dry_run=True. Setting save=False." in caplog.text
+def test_predict_dry_run_and_save(labels_absolute_path, caplog, tmp_path):
+    config = PredictConfig(filepaths=labels_absolute_path, dry_run=True, save=True)
+    assert (
+        "Cannot save when predicting with dry_run=True. Setting save=False and save_dir=None."
+        in caplog.text
+    )
+    assert not config.save
+    assert config.save_dir is None
+
+    config = PredictConfig(filepaths=labels_absolute_path, dry_run=True, save_dir=tmp_path)
+    assert not config.save
+    assert config.save_dir is None
 
 
 def test_predict_filepaths_with_duplicates(labels_absolute_path, tmp_path, caplog):
@@ -257,40 +264,60 @@ def test_model_cache_dir(labels_absolute_path, tmp_path):
 
 
 def test_predict_save(labels_absolute_path, tmp_path, dummy_trained_model_checkpoint):
-    # if save is True, use default save path
+    # if save is True, save in current working directory
     config = PredictConfig(filepaths=labels_absolute_path, skip_load_validation=True)
-    assert config.save == Path.cwd() / "zamba_predictions.csv"
+    assert config.save_dir == Path.cwd()
 
     config = PredictConfig(filepaths=labels_absolute_path, save=False, skip_load_validation=True)
     assert config.save is False
+    assert config.save_dir is None
 
-    # use checkpoint directory if checkpoint exists
+    # if save_dir is specified, set save to True
     config = PredictConfig(
         filepaths=labels_absolute_path,
+        save=False,
+        save_dir=tmp_path / "my_dir",
         skip_load_validation=True,
-        checkpoint=dummy_trained_model_checkpoint,
     )
-    assert config.save == Path(dummy_trained_model_checkpoint).parent / "zamba_predictions.csv"
+    assert config.save is True
+    # save dir gets created
+    assert (tmp_path / "my_dir").exists()
+
+    # empty save dir does not error
+    save_dir = tmp_path / "save_dir"
+    save_dir.mkdir()
 
-    # case does not matter as long as it's a csv
     config = PredictConfig(
-        filepaths=labels_absolute_path, save=tmp_path / "zamba/my_model/my_predictions.CSV"
+        filepaths=labels_absolute_path, save_dir=save_dir, skip_load_validation=True
     )
-    assert config.save == tmp_path / "zamba/my_model/my_predictions.CSV"
-
-    # cannot pass in directories, must specify full path
-    with pytest.raises(ValueError) as error:
-        PredictConfig(
-            filepaths=labels_absolute_path, save="zamba/my_model/", skip_load_validation=True
+    assert config.save_dir == save_dir
+
+    # save dir with prediction csv or yaml will error
+    for pred_file in [
+        (save_dir / "zamba_predictions.csv"),
+        (save_dir / "predict_configuration.yaml"),
+    ]:
+        # just takes one of the two files to raise error
+        pred_file.touch()
+        with pytest.raises(ValueError) as error:
+            PredictConfig(
+                filepaths=labels_absolute_path, save_dir=save_dir, skip_load_validation=True
+            )
+        assert (
+            f"zamba_predictions.csv and/or predict_configuration.yaml already exist in {save_dir}. If you would like to overwrite, set overwrite=True"
+            == error.value.errors()[0]["msg"]
         )
-    assert "Save path must end with .csv" in error.value.errors()[0]["msg"]
+        pred_file.unlink()
 
-    # cannot use path that already exists
-    save_path = tmp_path / "pred.csv"
-    save_path.touch()
-    with pytest.raises(ValueError) as error:
-        PredictConfig(filepaths=labels_absolute_path, save=save_path, skip_load_validation=True)
-    assert "already exists" in error.value.errors()[0]["msg"]
+    # can overwrite
+    pred_file.touch()
+    config = PredictConfig(
+        filepaths=labels_absolute_path,
+        save_dir=save_dir,
+        skip_load_validation=True,
+        overwrite=True,
+    )
+    assert config.save_dir == save_dir
 
 
 def test_validate_scheduler(labels_absolute_path):
diff --git a/tests/test_densepose.py b/tests/test_densepose.py
index 48e84bb0..ca26a219 100644
--- a/tests/test_densepose.py
+++ b/tests/test_densepose.py
@@ -122,8 +122,8 @@ def test_denseposeconfig(model, tmp_path):
             output_type="bananas",
             render_output=True,
             embeddings_in_json=False,
-            data_directory=ASSETS_DIR / "densepose_tests",
-            save_path=tmp_path,
+            data_dir=ASSETS_DIR / "densepose_tests",
+            save_dir=tmp_path,
         )
 
     dpc = DensePoseConfig(
@@ -131,13 +131,13 @@ def test_denseposeconfig(model, tmp_path):
         output_type="segmentation" if model == "animals" else "chimp_anatomy",
         render_output=True,
         embeddings_in_json=False,
-        data_directory=ASSETS_DIR / "densepose_tests",
-        save_path=tmp_path,
+        data_dir=ASSETS_DIR / "densepose_tests",
+        save_dir=tmp_path,
     )
 
     dpc.run_model()
 
-    # ensure all outputs are saved in save_path
+    # ensure all outputs are saved in save_dir
     assert (tmp_path / "chimp_denspose_video.mp4").exists()
     assert (tmp_path / "chimp_denspose_labels.json").exists()
 
diff --git a/tests/test_model_manager.py b/tests/test_model_manager.py
index 21948250..682a032d 100644
--- a/tests/test_model_manager.py
+++ b/tests/test_model_manager.py
@@ -75,14 +75,14 @@ def test_save_metrics_less_than_two_classes(
     trainer = train_model(
         train_config=DummyTrainConfig(
             labels=labels,
-            data_directory=TEST_VIDEOS_DIR,
+            data_dir=TEST_VIDEOS_DIR,
             model_name="dummy",
             checkpoint=dummy_checkpoint,
             max_epochs=1,
             batch_size=1,
             auto_lr_find=False,
             num_workers=2,
-            save_directory=tmp_path / "my_model",
+            save_dir=tmp_path / "my_model",
             skip_load_validation=True,
         ),
         video_loader_config=dummy_video_loader_config,
@@ -125,22 +125,22 @@ def test_save_configuration(dummy_trainer):
     }
 
 
-def test_train_save_directory(dummy_trainer):
+def test_train_save_dir(dummy_trainer):
     assert Path(dummy_trainer.logger.root_dir).name == "my_model"
     assert Path(dummy_trainer.logger.log_dir).name == "version_0"
 
 
-def test_train_save_directory_overwrite(
+def test_train_save_dir_overwrite(
     labels_absolute_path, dummy_checkpoint, tmp_path, dummy_video_loader_config
 ):
     config = DummyTrainConfig(
         labels=labels_absolute_path,
-        data_directory=TEST_VIDEOS_DIR,
+        data_dir=TEST_VIDEOS_DIR,
         model_name="dummy",
         checkpoint=dummy_checkpoint,
-        save_directory=tmp_path / "my_model",
+        save_dir=tmp_path / "my_model",
         skip_load_validation=True,
-        overwrite_save_directory=True,
+        overwrite=True,
         max_epochs=1,
         batch_size=1,
         auto_lr_find=False,
@@ -151,9 +151,9 @@ def test_train_save_directory_overwrite(
         train_config=config, video_loader_config=dummy_video_loader_config
     )
 
-    assert Path(overwrite_trainer.logger.log_dir).resolve() == config.save_directory.resolve()
+    assert Path(overwrite_trainer.logger.log_dir).resolve() == config.save_dir.resolve()
 
-    assert not any([f.name.startswith("version_") for f in config.save_directory.iterdir()])
+    assert not any([f.name.startswith("version_") for f in config.save_dir.iterdir()])
 
     # when training from checkpoint, model_name is None so get PTL default ckpt name
     for f in [
@@ -162,7 +162,7 @@ def test_train_save_directory_overwrite(
         "val_metrics.json",
         "epoch=0-step=10.ckpt",
     ]:
-        assert (config.save_directory / f).exists()
+        assert (config.save_dir / f).exists()
 
 
 @pytest.mark.parametrize("model_name", ["time_distributed", "slowfast", "european"])
diff --git a/zamba/cli.py b/zamba/cli.py
index 6f1b4f6d..aa212425 100644
--- a/zamba/cli.py
+++ b/zamba/cli.py
@@ -53,7 +53,7 @@ def train(
     ),
     save_dir: Path = typer.Option(
         None,
-        help="Directory in which to save model checkpoint and configuration file. If not specified, will save to a folder called 'zamba_{model_name}' in your working directory.",
+        help="An optional directory in which to save the model checkpoint and configuration file. If not specified, will save to a `version_n` folder in your working directory.",
     ),
     num_workers: int = typer.Option(
         None,
@@ -95,7 +95,7 @@ def train(
 
     # override if any command line arguments are passed
     if data_dir is not None:
-        train_dict["data_directory"] = data_dir
+        train_dict["data_dir"] = data_dir
 
     if labels is not None:
         train_dict["labels"] = labels
@@ -116,7 +116,7 @@ def train(
         train_dict["dry_run"] = dry_run
 
     if save_dir is not None:
-        train_dict["save_directory"] = save_dir
+        train_dict["save_dir"] = save_dir
 
     if num_workers is not None:
         train_dict["num_workers"] = num_workers
@@ -154,17 +154,17 @@ def train(
     msg = f"""The following configuration will be used for training:
 
     Config file: {config_file}
-    Data directory: {data_dir if data_dir is not None else config_dict["train_config"].get("data_directory")}
+    Data directory: {data_dir if data_dir is not None else config_dict["train_config"].get("data_dir")}
     Labels csv: {labels if labels is not None else config_dict["train_config"].get("labels")}
     Species: {species}
     Model name: {config.train_config.model_name}
     Checkpoint: {checkpoint if checkpoint is not None else config_dict["train_config"].get("checkpoint")}
-    Weight download region: {config.train_config.weight_download_region}
     Batch size: {config.train_config.batch_size}
     Number of workers: {config.train_config.num_workers}
     GPUs: {config.train_config.gpus}
     Dry run: {config.train_config.dry_run}
-    Save directory: {config.train_config.save_directory}
+    Save directory: {config.train_config.save_dir}
+    Weight download region: {config.train_config.weight_download_region}
     """
 
     if yes:
@@ -203,11 +203,12 @@ def predict(
     batch_size: int = typer.Option(None, help="Batch size to use for training."),
     save: bool = typer.Option(
         None,
-        help="Whether to save out predictions to a csv file. If you want to specify the location of the csv, use save_path instead.",
+        help="Whether to save out predictions. If you want to specify the output directory, use save_dir instead.",
     ),
-    save_path: Path = typer.Option(
+    save_dir: Path = typer.Option(
         None,
-        help="Full path for prediction CSV file. Any needed parent directories will be created.",
+        help="An optional directory in which to save the model predictions and configuration yaml. "
+        "Defaults to the current working directory if save is True.",
     ),
     dry_run: bool = typer.Option(None, help="Runs one batch of inference to check for bugs."),
     config: Path = typer.Option(
@@ -238,6 +239,9 @@ def predict(
         None,
         help="Skip check that verifies all videos can be loaded prior to inference. Only use if you're very confident all your videos can be loaded.",
     ),
+    overwrite: bool = typer.Option(
+        None, "--overwrite", "-o", help="Overwrite outputs in the save directory if they exist."
+    ),
     yes: bool = typer.Option(
         False,
         "--yes",
@@ -271,7 +275,7 @@ def predict(
 
     # override if any command line arguments are passed
     if data_dir is not None:
-        predict_dict["data_directory"] = data_dir
+        predict_dict["data_dir"] = data_dir
 
     if filepaths is not None:
         predict_dict["filepaths"] = filepaths
@@ -294,9 +298,9 @@ def predict(
     if save is not None:
         predict_dict["save"] = save
 
-    # save path takes precedence over save
-    if save_path is not None:
-        predict_dict["save"] = save_path
+    # save_dir takes precedence over save
+    if save_dir is not None:
+        predict_dict["save_dir"] = save_dir
 
     if proba_threshold is not None:
         predict_dict["proba_threshold"] = proba_threshold
@@ -313,6 +317,9 @@ def predict(
     if skip_load_validation is not None:
         predict_dict["skip_load_validation"] = skip_load_validation
 
+    if overwrite is not None:
+        predict_dict["overwrite"] = overwrite
+
     try:
         manager = ModelManager(
             ModelConfig(
@@ -329,7 +336,7 @@ def predict(
     msg = f"""The following configuration will be used for inference:
 
     Config file: {config_file}
-    Data directory: {data_dir if data_dir is not None else config_dict["predict_config"].get("data_directory")}
+    Data directory: {data_dir if data_dir is not None else config_dict["predict_config"].get("data_dir")}
     Filepath csv: {filepaths if filepaths is not None else config_dict["predict_config"].get("filepaths")}
     Model: {config.predict_config.model_name}
     Checkpoint: {checkpoint if checkpoint is not None else config_dict["predict_config"].get("checkpoint")}
@@ -337,7 +344,7 @@ def predict(
     Number of workers: {config.predict_config.num_workers}
     GPUs: {config.predict_config.gpus}
     Dry run: {config.predict_config.dry_run}
-    Save: {config.predict_config.save}
+    Save directory: {config.predict_config.save_dir}
     Proba threshold: {config.predict_config.proba_threshold}
     Output class names: {config.predict_config.output_class_names}
     Weight download region: {config.predict_config.weight_download_region}
@@ -387,7 +394,7 @@ def densepose(
     filepaths: Path = typer.Option(
         None, exists=True, help="Path to csv containing `filepath` column with videos."
     ),
-    save_path: Path = typer.Option(
+    save_dir: Path = typer.Option(
         None,
         help="An optional directory for saving the output. Defaults to the current working directory.",
     ),
@@ -447,13 +454,13 @@ def densepose(
 
     # override if any command line arguments are passed
     if data_dir is not None:
-        predict_dict["data_directory"] = data_dir
+        predict_dict["data_dir"] = data_dir
 
     if filepaths is not None:
         predict_dict["filepaths"] = filepaths
 
-    if save_path is not None:
-        predict_dict["save_path"] = save_path
+    if save_dir is not None:
+        predict_dict["save_dir"] = save_dir
 
     if weight_download_region is not None:
         predict_dict["weight_download_region"] = weight_download_region
@@ -478,7 +485,7 @@ def densepose(
     Config file: {config_file}
     Output type: {densepose_config.output_type}
     Render output: {densepose_config.render_output}
-    Data directory: {data_dir if data_dir is not None else config_dict.get("data_directory")}
+    Data directory: {data_dir if data_dir is not None else config_dict.get("data_dir")}
     Filepath csv: {filepaths if filepaths is not None else config_dict.get("filepaths")}
     Weight download region: {densepose_config.weight_download_region}
     Cache directory: {densepose_config.cache_dir}
diff --git a/zamba/models/config.py b/zamba/models/config.py
index faba49dc..d12f13d0 100644
--- a/zamba/models/config.py
+++ b/zamba/models/config.py
@@ -78,14 +78,14 @@ def validate_model_cache_dir(model_cache_dir: Optional[Path]):
 
 
 def check_files_exist_and_load(
-    df: pd.DataFrame, data_directory: DirectoryPath, skip_load_validation: bool
+    df: pd.DataFrame, data_dir: DirectoryPath, skip_load_validation: bool
 ):
     """Check whether files in file list exist and can be loaded with ffmpeg.
     Warn and skip files that don't exist or can't be loaded.
 
     Args:
         df (pd.DataFrame): DataFrame with a "filepath" column
-        data_directory (Path): Data folder to prepend if filepath is not an
+        data_dir (Path): Data folder to prepend if filepath is not an
             absolute path.
         skip_load_validation (bool): Skip ffprobe check that verifies all videos
             can be loaded.
@@ -94,7 +94,7 @@ def check_files_exist_and_load(
         pd.DataFrame: DataFrame with valid and loadable videos.
     """
     # update filepath column to prepend data_dir if filepath column is not an absolute path
-    data_dir = Path(data_directory).resolve()
+    data_dir = Path(data_dir).resolve()
     df["filepath"] = str(data_dir) / df.filepath.path
 
     # we can have multiple rows per file with labels so limit just to one row per file for these checks
@@ -109,7 +109,7 @@ def check_files_exist_and_load(
     # if no files exist
     if len(invalid_files) == len(files_df):
         raise ValueError(
-            f"None of the video filepaths exist. Are you sure they're specified correctly? Here's an example invalid path: {invalid_files.filepath.values[0]}. Either specify absolute filepaths in the csv or provide filepaths relative to `data_directory`."
+            f"None of the video filepaths exist. Are you sure they're specified correctly? Here's an example invalid path: {invalid_files.filepath.values[0]}. Either specify absolute filepaths in the csv or provide filepaths relative to `data_dir`."
         )
 
     # if at least some files exist
@@ -280,10 +280,10 @@ class TrainConfig(ZambaBaseModel):
     Args:
         labels (FilePath or pandas DataFrame): Path to a CSV or pandas DataFrame
             containing labels for training, with one row per label. There must be
-            columns called 'filepath' (absolute or relative to the data_directory) and
+            columns called 'filepath' (absolute or relative to the data_dir) and
             'label', and optionally columns called 'split' ("train", "val", or "holdout")
             and 'site'. Labels must be specified to train a model.
-        data_directory (DirectoryPath): Path to a directory containing training
+        data_dir (DirectoryPath): Path to a directory containing training
             videos. Defaults to the working directory.
         model_name (str, optional): Name of the model to use for training. Options
             are: time_distributed, slowfast, european. Defaults to time_distributed.
@@ -326,16 +326,15 @@ class TrainConfig(ZambaBaseModel):
         split_proportions (dict): Proportions used to divide data into training,
             validation, and holdout sets if a if a "split" column is not included in
             labels. Defaults to "train": 3, "val": 1, "holdout": 1.
-        save_directory (Path, optional): Path to a directory where training files
+        save_dir (Path, optional): Path to a directory where training files
             will be saved. Files include the best model checkpoint (``model_name``.ckpt),
             training configuration (configuration.yaml), Tensorboard logs
             (events.out.tfevents...), test metrics (test_metrics.json), validation
             metrics (val_metrics.json), and model hyperparameters (hparams.yml).
-            If None, a folder is created in the working directory called
-            "zamba_``model_name``". Defaults to None.
-        overwrite_save_directory (bool): If True, will save outputs in `save_directory`
-            overwriting if those exist. If False, will create auto-incremented `version_n` folder
-            in `save_directory` with model outputs. Defaults to False.
+            If None, a folder is created in the working directory. Defaults to None.
+        overwrite (bool): If True, will save outputs in `save_dir` overwriting if those
+            exist. If False, will create auto-incremented `version_n` folder in `save_dir`
+            with model outputs. Defaults to False.
         skip_load_validation (bool). Skip ffprobe check, which verifies that all
             videos can be loaded and skips files that cannot be loaded. Defaults
             to False.
@@ -351,7 +350,7 @@ class TrainConfig(ZambaBaseModel):
     """
 
     labels: Union[FilePath, pd.DataFrame]
-    data_directory: DirectoryPath = Path.cwd()
+    data_dir: DirectoryPath = Path.cwd()
     checkpoint: Optional[FilePath] = None
     scheduler_config: Optional[Union[str, SchedulerConfig]] = "default"
     model_name: Optional[ModelEnum] = ModelEnum.time_distributed
@@ -365,8 +364,8 @@ class TrainConfig(ZambaBaseModel):
     early_stopping_config: Optional[EarlyStoppingConfig] = EarlyStoppingConfig()
     weight_download_region: RegionEnum = "us"
     split_proportions: Optional[Dict[str, int]] = {"train": 3, "val": 1, "holdout": 1}
-    save_directory: Path = Path.cwd()
-    overwrite_save_directory: bool = False
+    save_dir: Path = Path.cwd()
+    overwrite: bool = False
     skip_load_validation: bool = False
     from_scratch: bool = False
     predict_all_zamba_species: bool = True
@@ -466,7 +465,7 @@ def validate_filepaths_and_labels(cls, values):
         # check that all videos exist and can be loaded
         values["labels"] = check_files_exist_and_load(
             df=labels,
-            data_directory=values["data_directory"],
+            data_dir=values["data_dir"],
             skip_load_validation=values["skip_load_validation"],
         )
         return values
@@ -508,49 +507,20 @@ def preprocess_labels(cls, values):
                 )
 
                 logger.info(
-                    f"Writing out split information to {values['save_directory'] / 'splits.csv'}."
+                    f"Writing out split information to {values['save_dir'] / 'splits.csv'}."
                 )
 
                 # create the directory to save if we need to.
-                values["save_directory"].mkdir(parents=True, exist_ok=True)
+                values["save_dir"].mkdir(parents=True, exist_ok=True)
 
                 labels.reset_index()[["filepath", "split"]].drop_duplicates().to_csv(
-                    values["save_directory"] / "splits.csv", index=False
+                    values["save_dir"] / "splits.csv", index=False
                 )
 
         # filepath becomes column instead of index
         values["labels"] = labels.reset_index()
         return values
 
-    def get_model_only_params(self):
-        """Return only params that are not data or machine specific.
-        Used for generating official configs.
-        """
-        train_config = self.dict()
-
-        # remove data and machine specific params
-        for key in [
-            "labels",
-            "data_directory",
-            "dry_run",
-            "batch_size",
-            "auto_lr_find",
-            "gpus",
-            "num_workers",
-            "max_epochs",
-            "weight_download_region",
-            "split_proportions",
-            "save_directory",
-            "overwrite_save_directory",
-            "skip_load_validation",
-            "from_scratch",
-            "model_cache_dir",
-            "predict_all_zamba_species",
-        ]:
-            train_config.pop(key)
-
-        return train_config
-
 
 class PredictConfig(ZambaBaseModel):
     """
@@ -558,10 +528,10 @@ class PredictConfig(ZambaBaseModel):
 
     Args:
         filepaths (FilePath): Path to a CSV containing videos for inference, with
-            one row per video in the data_directory. There must be a column called
-            'filepath' (absolute or relative to the data_directory). If None, uses
-            all files in data_directory. Defaults to None.
-        data_directory (DirectoryPath): Path to a directory containing videos for
+            one row per video in the data_dir. There must be a column called
+            'filepath' (absolute or relative to the data_dir). If None, uses
+            all files in data_dir. Defaults to None.
+        data_dir (DirectoryPath): Path to a directory containing videos for
             inference. Defaults to the working directory.
         model_name (str, optional): Name of the model to use for inference. Options
             are: time_distributed, slowfast, european. Defaults to time_distributed.
@@ -574,9 +544,13 @@ class PredictConfig(ZambaBaseModel):
             that the data will be loaded in the main process. The maximum value is
             the number of CPUs in the system. Defaults to 3.
         batch_size (int): Batch size to use for inference. Defaults to 2.
-        save (bool or Path): Path to a CSV to save predictions. If True is passed,
-            "zamba_predictions.csv" is written to the current working directory.
-            If False is passed, predictions are not saved. Defaults to True.
+        save (bool): Whether to save out predictions. If False, predictions are
+            not saved. Defaults to True.
+        save_dir (Path, optional): An optional directory in which to save the model
+             predictions and configuration yaml. If no save_dir is specified and save=True,
+             outputs will be written to the current working directory. Defaults to None.
+        overwrite (bool): If True, overwrite outputs in save_dir if they exist.
+            Defaults to False.
         dry_run (bool): Perform inference on a single batch for testing. Predictions
             will not be saved. Defaults to False.
         proba_threshold (float, optional): Probability threshold for classification.
@@ -599,14 +573,16 @@ class PredictConfig(ZambaBaseModel):
             default cache directory. Defaults to None.
     """
 
-    data_directory: DirectoryPath = Path.cwd()
+    data_dir: DirectoryPath = Path.cwd()
     filepaths: Optional[FilePath] = None
     checkpoint: Optional[FilePath] = None
     model_name: Optional[ModelEnum] = ModelEnum.time_distributed
     gpus: int = GPUS_AVAILABLE
     num_workers: int = 3
     batch_size: int = 2
-    save: Union[bool, Path] = True
+    save: bool = True
+    save_dir: Optional[Path] = None
+    overwrite: bool = False
     dry_run: bool = False
     proba_threshold: Optional[float] = None
     output_class_names: bool = False
@@ -622,42 +598,45 @@ class PredictConfig(ZambaBaseModel):
 
     @root_validator(skip_on_failure=True)
     def validate_dry_run_and_save(cls, values):
-        if values["dry_run"] and (values["save"] is not False):
-            logger.warning("Cannot save when predicting with dry_run=True. Setting save=False.")
+        if values["dry_run"] and (
+            (values["save"] is not False) or (values["save_dir"] is not None)
+        ):
+            logger.warning(
+                "Cannot save when predicting with dry_run=True. Setting save=False and save_dir=None."
+            )
             values["save"] = False
+            values["save_dir"] = None
 
         return values
 
     @root_validator(skip_on_failure=True)
-    def validate_save(cls, values):
-        # do this check before we look up checkpoints based on model name so we can see if checkpoint is None
+    def validate_save_dir(cls, values):
+        save_dir = values["save_dir"]
         save = values["save"]
-        checkpoint = values["checkpoint"]
 
-        # if False, no predictions will be written out
-        if save is False:
-            return values
+        # if no save_dir but save is True, use current working directory
+        if save_dir is None and save:
+            save_dir = Path.cwd()
 
-        else:
-            # if save=True and we have a local checkpoint, save in checkpoint directory
-            if save is True and checkpoint is not None:
-                save = checkpoint.parent / "zamba_predictions.csv"
-
-            # else, save to current working directory
-            elif save is True and checkpoint is None:
-                save = Path.cwd() / "zamba_predictions.csv"
-
-        # validate save path
-        if isinstance(save, Path):
-            if save.suffix.lower() != ".csv":
-                raise ValueError("Save path must end with .csv")
-            elif save.exists():
-                raise ValueError(f"Save path {save} already exists.")
-            else:
-                values["save"] = save
+        if save_dir is not None:
+            # check if files exist
+            if (
+                (save_dir / "zamba_predictions.csv").exists()
+                or (save_dir / "predict_configuration.yaml").exists()
+            ) and not values["overwrite"]:
+                raise ValueError(
+                    f"zamba_predictions.csv and/or predict_configuration.yaml already exist in {save_dir}. If you would like to overwrite, set overwrite=True"
+                )
+
+            # make a directory if needed
+            save_dir.mkdir(parents=True, exist_ok=True)
+
+            # set save to True if save_dir is set
+            if not save:
+                save = True
 
-            # create any needed parent directories
-            save.parent.mkdir(parents=True, exist_ok=True)
+        values["save_dir"] = save_dir
+        values["save"] = save
 
         return values
 
@@ -686,12 +665,12 @@ def get_filepaths(cls, values):
         contains files with valid suffixes.
         """
         if values["filepaths"] is None:
-            logger.info(f"Getting files in {values['data_directory']}.")
+            logger.info(f"Getting files in {values['data_dir']}.")
             files = []
             new_suffixes = []
 
             # iterate over all files in data directory
-            for f in values["data_directory"].rglob("*"):
+            for f in values["data_dir"].rglob("*"):
                 if f.is_file():
                     # keep just files with supported suffixes
                     if f.suffix.lower() in VIDEO_SUFFIXES:
@@ -705,9 +684,9 @@ def get_filepaths(cls, values):
                 )
 
             if len(files) == 0:
-                raise ValueError(f"No video files found in {values['data_directory']}.")
+                raise ValueError(f"No video files found in {values['data_dir']}.")
 
-            logger.info(f"Found {len(files)} videos in {values['data_directory']}.")
+            logger.info(f"Found {len(files)} videos in {values['data_dir']}.")
             values["filepaths"] = pd.DataFrame(files, columns=["filepath"])
         return values
 
@@ -733,7 +712,7 @@ def validate_files(cls, values):
 
         values["filepaths"] = check_files_exist_and_load(
             df=files_df,
-            data_directory=values["data_directory"],
+            data_dir=values["data_dir"],
             skip_load_validation=values["skip_load_validation"],
         )
         return values
diff --git a/zamba/models/densepose/config.py b/zamba/models/densepose/config.py
index ab12ca6f..65f2f480 100644
--- a/zamba/models/densepose/config.py
+++ b/zamba/models/densepose/config.py
@@ -34,10 +34,10 @@ class DensePoseConfig(ZambaBaseModel):
             Defaults to False.
         embeddings_in_json (bool): Whether to save the embeddings matrices in the json of the
             DensePose result. Setting to True can result in large json files. Defaults to False.
-        data_directory (Path): Where to find the files listed in filepaths (or where to look if
+        data_dir (Path): Where to find the files listed in filepaths (or where to look if
             filepaths is not provided).
         filepaths (Path, optional): Path to a CSV file with a list of filepaths to process.
-        save_path (Path, optional): Directory for where to save the output files;
+        save_dir (Path, optional): Directory for where to save the output files;
             defaults to os.getcwd().
         cache_dir (Path, optional): Path for downloading and saving model weights. Defaults
             to env var `MODEL_CACHE_DIR` or the OS app cache dir.
@@ -49,9 +49,9 @@ class DensePoseConfig(ZambaBaseModel):
     output_type: DensePoseOutputEnum
     render_output: bool = False
     embeddings_in_json: bool = False
-    data_directory: Path
+    data_dir: Path
     filepaths: Optional[Path] = None
-    save_path: Optional[Path] = None
+    save_dir: Optional[Path] = None
     cache_dir: Optional[Path] = None
     weight_download_region: RegionEnum = RegionEnum("us")
 
@@ -71,7 +71,7 @@ def run_model(self):
         else:
             raise Exception(f"invalid {self.output_type}")
 
-        output_dir = Path(os.getcwd()) if self.save_path is None else self.save_path
+        output_dir = Path(os.getcwd()) if self.save_dir is None else self.save_dir
 
         dpm = DensePoseManager(
             model, model_cache_dir=self.cache_dir, download_region=self.weight_download_region
@@ -112,12 +112,12 @@ def get_filepaths(cls, values):
         contains files with valid suffixes.
         """
         if values["filepaths"] is None:
-            logger.info(f"Getting files in {values['data_directory']}.")
+            logger.info(f"Getting files in {values['data_dir']}.")
             files = []
             new_suffixes = []
 
             # iterate over all files in data directory
-            for f in values["data_directory"].rglob("*"):
+            for f in values["data_dir"].rglob("*"):
                 if f.is_file():
                     # keep just files with supported suffixes
                     if f.suffix.lower() in VIDEO_SUFFIXES:
@@ -131,9 +131,9 @@ def get_filepaths(cls, values):
                 )
 
             if len(files) == 0:
-                raise ValueError(f"No video files found in {values['data_directory']}.")
+                raise ValueError(f"No video files found in {values['data_dir']}.")
 
-            logger.info(f"Found {len(files)} videos in {values['data_directory']}.")
+            logger.info(f"Found {len(files)} videos in {values['data_dir']}.")
             values["filepaths"] = pd.DataFrame(files, columns=["filepath"])
         return values
 
@@ -159,7 +159,7 @@ def validate_files(cls, values):
 
         values["filepaths"] = check_files_exist_and_load(
             df=files_df,
-            data_directory=values["data_directory"],
+            data_dir=values["data_dir"],
             skip_load_validation=True,
         )
         return values
diff --git a/zamba/models/model_manager.py b/zamba/models/model_manager.py
index 35eb5b81..e367b3c5 100644
--- a/zamba/models/model_manager.py
+++ b/zamba/models/model_manager.py
@@ -81,7 +81,8 @@ def instantiate_model(
             hparams = yaml.safe_load(f)
 
     else:
-        if not (model_cache_dir / checkpoint).exists():
+        # download if neither local checkpoint nor cached checkpoint exist
+        if not checkpoint.exists() and not (model_cache_dir / checkpoint).exists():
             logger.info("Downloading weights for model.")
             checkpoint = download_weights(
                 filename=str(checkpoint),
@@ -225,16 +226,12 @@ def train_model(
 
     validate_species(model, data_module)
 
-    train_config.save_directory.mkdir(parents=True, exist_ok=True)
+    train_config.save_dir.mkdir(parents=True, exist_ok=True)
 
     # add folder version_n that auto increments if we are not overwriting
-    tensorboard_version = (
-        train_config.save_directory.name if train_config.overwrite_save_directory else None
-    )
+    tensorboard_version = train_config.save_dir.name if train_config.overwrite else None
     tensorboard_save_dir = (
-        train_config.save_directory.parent
-        if train_config.overwrite_save_directory
-        else train_config.save_directory
+        train_config.save_dir.parent if train_config.overwrite else train_config.save_dir
     )
 
     tensorboard_logger = TensorBoardLogger(
@@ -245,9 +242,7 @@ def train_model(
     )
 
     logging_and_save_dir = (
-        tensorboard_logger.log_dir
-        if not train_config.overwrite_save_directory
-        else train_config.save_directory
+        tensorboard_logger.log_dir if not train_config.overwrite else train_config.save_dir
     )
 
     model_checkpoint = ModelCheckpoint(
@@ -381,6 +376,13 @@ def predict_model(
         "video_loader_config": json.loads(video_loader_config.json()),
     }
 
+    if predict_config.save is not False:
+
+        config_path = predict_config.save_dir / "predict_configuration.yaml"
+        logger.info(f"Writing out full configuration to {config_path}.")
+        with config_path.open("w") as fp:
+            yaml.dump(configuration, fp)
+
     dataloader = data_module.predict_dataloader()
     logger.info("Starting prediction...")
     probas = trainer.predict(model=model, dataloaders=dataloader)
@@ -401,14 +403,9 @@ def predict_model(
 
     if predict_config.save is not False:
 
-        config_path = predict_config.save.parent / "predict_configuration.yaml"
-        config_path.parent.mkdir(exist_ok=True, parents=True)
-        logger.info(f"Writing out full configuration to {config_path}.")
-        with config_path.open("w") as fp:
-            yaml.dump(configuration, fp)
-
-        logger.info(f"Saving out predictions to {predict_config.save}.")
-        with predict_config.save.open("w") as fp:
+        preds_path = predict_config.save_dir / "zamba_predictions.csv"
+        logger.info(f"Saving out predictions to {preds_path}.")
+        with preds_path.open("w") as fp:
             df.to_csv(fp, index=True)
 
     return df
diff --git a/zamba/models/official_models/european/config.yaml b/zamba/models/official_models/european/config.yaml
index 0b4177e0..66c46f99 100644
--- a/zamba/models/official_models/european/config.yaml
+++ b/zamba/models/official_models/european/config.yaml
@@ -1,6 +1,4 @@
 train_config:
-  scheduler_config: default
-  model_name: european
   backbone_finetune_config:
     backbone_initial_ratio_lr: 0.01
     multiplier: 1
@@ -9,21 +7,36 @@ train_config:
     unfreeze_backbone_at_epoch: 15
     verbose: true
   early_stopping_config:
+    mode: max
     monitor: val_macro_f1
     patience: 3
     verbose: true
-    mode: max
+  model_name: european
+  scheduler_config: default
 video_loader_config:
-  model_input_height: 240
-  model_input_width: 426
   crop_bottom_pixels: 50
-  fps: 4
-  total_frames: 16
+  early_bias: false
   ensure_total_frames: true
+  evenly_sample_total_frames: false
+  fps: 4.0
+  frame_indices: null
+  frame_selection_height: null
+  frame_selection_width: null
+  i_frames: false
   megadetector_lite_config:
     confidence: 0.25
     fill_mode: score_sorted
+    image_height: 416
+    image_width: 416
     n_frames: 16
+    nms_threshold: 0.45
+    seed: 55
+    sort_by_time: true
+  model_input_height: 240
+  model_input_width: 426
+  pix_fmt: rgb24
+  scene_threshold: null
+  total_frames: 16
 predict_config:
   model_name: european
-public_checkpoint: european_0c69da8a888c499411deaa040a91d76546ddf78a.ckpt
+public_checkpoint: european_0a80dc77bf.ckpt
diff --git a/zamba/models/official_models/european/predict_configuration.yaml b/zamba/models/official_models/european/predict_configuration.yaml
index 3e4eb57b..be82b672 100644
--- a/zamba/models/official_models/european/predict_configuration.yaml
+++ b/zamba/models/official_models/european/predict_configuration.yaml
@@ -2,16 +2,18 @@ inference_start_time: '2021-10-15T03:46:08.256520'
 model_class: TimeDistributedEfficientNet
 predict_config:
   batch_size: 2
-  cache_dir: /home/ubuntu/.cache/zamba
   checkpoint: experiments/european_td_dev_base/version_0/time_distributed.ckpt
-  data_directory: /home/ubuntu/zamba-algorithms
+  data_dir: /home/ubuntu/zamba-algorithms
   dry_run: false
   gpus: 1
-  model_name: time_distributed
+  model_cache_dir: /home/ubuntu/.cache/zamba
+  model_name: null
   num_workers: 3
   output_class_names: false
+  overwrite: false
   proba_threshold: null
-  save: experiments/european_td_dev_base/version_0/zamba_predictions.csv
+  save: true
+  save_dir: experiments/european_td_dev_base/version_0
   skip_load_validation: true
   weight_download_region: us
 species:
@@ -27,6 +29,8 @@ species:
 - weasel
 - wild_boar
 video_loader_config:
+  cache_dir: /tmp/zamba_cache
+  cleanup_cache: false
   crop_bottom_pixels: 50
   early_bias: false
   ensure_total_frames: true
diff --git a/zamba/models/official_models/european/train_configuration.yaml b/zamba/models/official_models/european/train_configuration.yaml
index 329aabe7..38eb9622 100644
--- a/zamba/models/official_models/european/train_configuration.yaml
+++ b/zamba/models/official_models/european/train_configuration.yaml
@@ -23,9 +23,8 @@ train_config:
     unfreeze_backbone_at_epoch: 15
     verbose: true
   batch_size: 2
-  cache_dir: /home/ubuntu/.cache/zamba
   checkpoint: data/results/experiments/td_dev_set_full_size_mdlite/results/version_1/time_distributed_zamba.ckpt
-  data_directory: /home/ubuntu/zamba-algorithms
+  data_dir: /home/ubuntu/zamba-algorithms
   dry_run: false
   early_stopping_config:
     mode: max
@@ -35,17 +34,20 @@ train_config:
   from_scratch: false
   gpus: 1
   max_epochs: null
+  model_cache_dir: /home/ubuntu/.cache/zamba
   model_name: null
   num_workers: 3
-  overwrite_save_directory: false
+  overwrite: false
   predict_all_zamba_species: true
-  save_directory: experiments/european_td_dev_base
+  save_dir: experiments/european_td_dev_base
   scheduler_config: default
   skip_load_validation: true
   split_proportions: null
   weight_download_region: us
 training_start_time: '2021-10-13T16:46:39.593515'
 video_loader_config:
+  cache_dir: /tmp/zamba_cache
+  cleanup_cache: false
   crop_bottom_pixels: 50
   early_bias: false
   ensure_total_frames: true
diff --git a/zamba/models/official_models/slowfast/config.yaml b/zamba/models/official_models/slowfast/config.yaml
index 48baa85e..f77a4595 100644
--- a/zamba/models/official_models/slowfast/config.yaml
+++ b/zamba/models/official_models/slowfast/config.yaml
@@ -1,12 +1,4 @@
 train_config:
-  scheduler_config:
-    scheduler: MultiStepLR
-    scheduler_params:
-      gamma: 0.5
-      milestones:
-      - 1
-      verbose: true
-  model_name: slowfast
   backbone_finetune_config:
     backbone_initial_ratio_lr: 0.01
     multiplier: 10
@@ -15,18 +7,42 @@ train_config:
     unfreeze_backbone_at_epoch: 3
     verbose: true
   early_stopping_config:
+    mode: max
+    monitor: val_macro_f1
     patience: 5
+    verbose: true
+  model_name: slowfast
+  scheduler_config:
+    scheduler: MultiStepLR
+    scheduler_params:
+      gamma: 0.5
+      milestones:
+      - 1
+      verbose: true
 video_loader_config:
-  model_input_height: 240
-  model_input_width: 426
   crop_bottom_pixels: 50
-  fps: 8
-  total_frames: 32
+  early_bias: false
   ensure_total_frames: true
+  evenly_sample_total_frames: false
+  fps: 8.0
+  frame_indices: null
+  frame_selection_height: null
+  frame_selection_width: null
+  i_frames: false
   megadetector_lite_config:
     confidence: 0.25
     fill_mode: score_sorted
+    image_height: 416
+    image_width: 416
     n_frames: 32
+    nms_threshold: 0.45
+    seed: 55
+    sort_by_time: true
+  model_input_height: 240
+  model_input_width: 426
+  pix_fmt: rgb24
+  scene_threshold: null
+  total_frames: 32
 predict_config:
   model_name: slowfast
-public_checkpoint: slowfast_501182d969aabf49805829a2b09ed8078b4255a3.ckpt
+public_checkpoint: slowfast_3c9d5d0c72.ckpt
diff --git a/zamba/models/official_models/slowfast/predict_configuration.yaml b/zamba/models/official_models/slowfast/predict_configuration.yaml
index b46a417b..c8b6ef36 100644
--- a/zamba/models/official_models/slowfast/predict_configuration.yaml
+++ b/zamba/models/official_models/slowfast/predict_configuration.yaml
@@ -3,15 +3,17 @@ model_class: SlowFast
 predict_config:
   batch_size: 2
   checkpoint: experiments/slowfast_small_set_full_size_mdlite/version_2/slowfast.ckpt
-  data_directory: /home/ubuntu/zamba-algorithms
+  data_dir: /home/ubuntu/zamba-algorithms
   dry_run: false
   gpus: 1
   model_cache_dir: /home/ubuntu/.cache/zamba
   model_name: null
   num_workers: 3
   output_class_names: false
+  overwrite: false
   proba_threshold: null
-  save: experiments/slowfast_small_set_full_size_mdlite/version_2/zamba_predictions.csv
+  save: true
+  save_dir: experiments/slowfast_small_set_full_size_mdlite/version_2
   skip_load_validation: true
   weight_download_region: us
 species:
diff --git a/zamba/models/official_models/slowfast/train_configuration.yaml b/zamba/models/official_models/slowfast/train_configuration.yaml
index 70cc2434..653014ce 100644
--- a/zamba/models/official_models/slowfast/train_configuration.yaml
+++ b/zamba/models/official_models/slowfast/train_configuration.yaml
@@ -45,7 +45,7 @@ train_config:
     verbose: true
   batch_size: 2
   checkpoint: null
-  data_directory: /home/ubuntu/zamba-algorithms
+  data_dir: /home/ubuntu/zamba-algorithms
   dry_run: false
   early_stopping_config:
     mode: max
@@ -58,9 +58,9 @@ train_config:
   model_cache_dir: /home/ubuntu/.cache/zamba
   model_name: slowfast
   num_workers: 3
-  overwrite_save_directory: false
+  overwrite: false
   predict_all_zamba_species: true
-  save_directory: experiments/slowfast_small_set_full_size_mdlite
+  save_dir: experiments/slowfast_small_set_full_size_mdlite
   scheduler_config:
     scheduler: MultiStepLR
     scheduler_params:
diff --git a/zamba/models/official_models/time_distributed/config.yaml b/zamba/models/official_models/time_distributed/config.yaml
index aa161965..0adcfe0c 100644
--- a/zamba/models/official_models/time_distributed/config.yaml
+++ b/zamba/models/official_models/time_distributed/config.yaml
@@ -1,12 +1,4 @@
 train_config:
-  scheduler_config:
-    scheduler: MultiStepLR
-    scheduler_params:
-      gamma: 0.5
-      milestones:
-      - 3
-      verbose: true
-  model_name: time_distributed
   backbone_finetune_config:
     backbone_initial_ratio_lr: 0.01
     multiplier: 1
@@ -15,18 +7,42 @@ train_config:
     unfreeze_backbone_at_epoch: 3
     verbose: true
   early_stopping_config:
+    mode: max
+    monitor: val_macro_f1
     patience: 5
+    verbose: true
+  model_name: time_distributed
+  scheduler_config:
+    scheduler: MultiStepLR
+    scheduler_params:
+      gamma: 0.5
+      milestones:
+      - 3
+      verbose: true
 video_loader_config:
-  model_input_height: 240
-  model_input_width: 426
   crop_bottom_pixels: 50
-  fps: 4
-  total_frames: 16
+  early_bias: false
   ensure_total_frames: true
+  evenly_sample_total_frames: false
+  fps: 4.0
+  frame_indices: null
+  frame_selection_height: null
+  frame_selection_width: null
+  i_frames: false
   megadetector_lite_config:
     confidence: 0.25
     fill_mode: score_sorted
+    image_height: 416
+    image_width: 416
     n_frames: 16
+    nms_threshold: 0.45
+    seed: 55
+    sort_by_time: true
+  model_input_height: 240
+  model_input_width: 426
+  pix_fmt: rgb24
+  scene_threshold: null
+  total_frames: 16
 predict_config:
   model_name: time_distributed
-public_checkpoint: time_distributed_9e710aa8c92d25190a64b3b04b9122bdcb456982.ckpt
+public_checkpoint: time_distributed_7f74686b7b.ckpt
diff --git a/zamba/models/official_models/time_distributed/predict_configuration.yaml b/zamba/models/official_models/time_distributed/predict_configuration.yaml
index ae39bbdf..9c08eab1 100644
--- a/zamba/models/official_models/time_distributed/predict_configuration.yaml
+++ b/zamba/models/official_models/time_distributed/predict_configuration.yaml
@@ -2,16 +2,18 @@ inference_start_time: '2021-09-30T10:55:29.100719'
 model_class: TimeDistributedEfficientNet
 predict_config:
   batch_size: 1
-  cache_dir: /home/ubuntu/.cache/zamba
   checkpoint: experiments/td_small_set_full_size_mdlite/version_1/time_distributed.ckpt
-  data_directory: /home/ubuntu/zamba-algorithms
+  data_dir: /home/ubuntu/zamba-algorithms
   dry_run: false
   gpus: 1
-  model_name: time_distributed
+  model_cache_dir: /home/ubuntu/.cache/zamba
+  model_name: null
   num_workers: 5
   output_class_names: false
+  overwrite: false
   proba_threshold: null
-  save: experiments/td_small_set_full_size_mdlite/version_1/zamba_predictions.csv
+  save: true
+  save_dir: experiments/td_small_set_full_size_mdlite/version_1
   skip_load_validation: true
   weight_download_region: us
 species:
@@ -48,6 +50,8 @@ species:
 - small_cat
 - wild_dog_jackal
 video_loader_config:
+  cache_dir: /tmp/zamba_cache
+  cleanup_cache: false
   crop_bottom_pixels: 50
   early_bias: false
   ensure_total_frames: true
diff --git a/zamba/models/official_models/time_distributed/train_configuration.yaml b/zamba/models/official_models/time_distributed/train_configuration.yaml
index 427260c5..d4627cc6 100644
--- a/zamba/models/official_models/time_distributed/train_configuration.yaml
+++ b/zamba/models/official_models/time_distributed/train_configuration.yaml
@@ -44,9 +44,8 @@ train_config:
     unfreeze_backbone_at_epoch: 3
     verbose: true
   batch_size: 1
-  cache_dir: /home/ubuntu/.cache/zamba
   checkpoint: null
-  data_directory: /home/ubuntu/zamba-algorithms
+  data_dir: /home/ubuntu/zamba-algorithms
   dry_run: false
   early_stopping_config:
     mode: max
@@ -56,11 +55,12 @@ train_config:
   from_scratch: true
   gpus: 1
   max_epochs: null
+  model_cache_dir: /home/ubuntu/.cache/zamba
   model_name: time_distributed
   num_workers: 5
-  overwrite_save_directory: false
+  overwrite: false
   predict_all_zamba_species: true
-  save_directory: experiments/td_small_set_full_size_mdlite
+  save_dir: experiments/td_small_set_full_size_mdlite
   scheduler_config:
     scheduler: MultiStepLR
     scheduler_params:
@@ -73,6 +73,8 @@ train_config:
   weight_download_region: us
 training_start_time: '2021-09-29T23:50:09.687298'
 video_loader_config:
+  cache_dir: /tmp/zamba_cache
+  cleanup_cache: false
   crop_bottom_pixels: 50
   early_bias: false
   ensure_total_frames: true
diff --git a/zamba/models/publish_models.py b/zamba/models/publish_models.py
index a32ff4e1..de731f7e 100644
--- a/zamba/models/publish_models.py
+++ b/zamba/models/publish_models.py
@@ -8,10 +8,47 @@
 import yaml
 
 from zamba import MODELS_DIRECTORY
-from zamba.models.config import WEIGHT_LOOKUP, ModelEnum, TrainConfig
+from zamba.models.config import WEIGHT_LOOKUP, ModelEnum
 from zamba.models.densepose import MODELS as DENSEPOSE_MODELS
 
 
+def get_model_only_params(full_configuration, subset="train_config"):
+    """Return only params that are not data or machine specific.
+    Used for generating official configs.
+    """
+    if subset == "train_config":
+        config = full_configuration[subset]
+        for key in [
+            "data_dir",
+            "dry_run",
+            "batch_size",
+            "auto_lr_find",
+            "gpus",
+            "num_workers",
+            "max_epochs",
+            "weight_download_region",
+            "split_proportions",
+            "save_dir",
+            "overwrite",
+            "skip_load_validation",
+            "from_scratch",
+            "model_cache_dir",
+            "predict_all_zamba_species",
+        ]:
+            config.pop(key)
+
+    elif subset == "video_loader_config":
+        config = full_configuration[subset]
+
+        if "megadetector_lite_config" in config.keys():
+            config["megadetector_lite_config"].pop("device")
+
+        for key in ["cache_dir", "cleanup_cache"]:
+            config.pop(key)
+
+    return config
+
+
 def publish_model(model_name, trained_model_dir):
     """
     Creates the files for the model folder in `official_models` and uploads the model to the three
@@ -51,12 +88,13 @@ def publish_model(model_name, trained_model_dir):
 
     # prepare config for use in official models dir
     logger.info("Preparing official config file.")
-    config_yaml = MODELS_DIRECTORY / model_name / "config.yaml"
 
-    with config_yaml.open() as f:
-        config_dict = yaml.safe_load(f)
+    # start with full train configuration
+    with (MODELS_DIRECTORY / model_name / "train_configuration.yaml").open() as f:
+        train_configuration_full_dict = yaml.safe_load(f)
 
-    train_config = TrainConfig.construct(**config_dict["train_config"]).get_model_only_params()
+    # get limited train config
+    train_config = get_model_only_params(train_configuration_full_dict, subset="train_config")
 
     # e.g. european model is trained from a checkpoint; we want to expose final model
     # (model_name: european) not the base checkpoint
@@ -66,18 +104,21 @@ def publish_model(model_name, trained_model_dir):
 
     official_config = dict(
         train_config=train_config,
-        video_loader_config=config_dict["video_loader_config"],
+        video_loader_config=get_model_only_params(
+            train_configuration_full_dict, subset="video_loader_config"
+        ),
         predict_config=dict(model_name=model_name),
     )
 
     # hash train_configuration to generate public filename for model
-    hash_str = hashlib.sha1(str(config_dict["train_config"]).encode("utf-8")).hexdigest()
+    hash_str = hashlib.sha1(str(train_configuration_full_dict).encode("utf-8")).hexdigest()[:10]
     public_file_name = f"{model_name}_{hash_str}.ckpt"
 
     # add that to official config
     official_config["public_checkpoint"] = public_file_name
 
     # write out official config
+    config_yaml = MODELS_DIRECTORY / model_name / "config.yaml"
     logger.info(f"Writing out to {config_yaml}")
     with config_yaml.open("w") as f:
         yaml.dump(official_config, f, sort_keys=False)
@@ -91,8 +132,11 @@ def upload_to_all_public_buckets(file, public_file_name):
         public_checkpoint = S3Path(
             f"s3://drivendata-public-assets{bucket}/zamba_official_models/{public_file_name}"
         )
-        logger.info(f"Uploading {file} to {public_checkpoint}")
-        public_checkpoint.upload_from(file, force_overwrite_to_cloud=True)
+        if public_checkpoint.exists():
+            logger.info(f"Skipping since {public_checkpoint} exists.")
+        else:
+            logger.info(f"Uploading {file} to {public_checkpoint}")
+            public_checkpoint.upload_from(file, force_overwrite_to_cloud=True)
 
 
 if __name__ == "__main__":
diff --git a/zamba/models/utils.py b/zamba/models/utils.py
index 7fc3112f..52042843 100644
--- a/zamba/models/utils.py
+++ b/zamba/models/utils.py
@@ -36,7 +36,10 @@ def download_weights(
 
 
 def get_model_checkpoint_filename(model_name):
-    config_file = MODELS_DIRECTORY / f"{model_name}/config.yaml"
+    if isinstance(model_name, Enum):
+        model_name = model_name.value
+
+    config_file = MODELS_DIRECTORY / model_name / "config.yaml"
     with config_file.open() as f:
         config_dict = yaml.safe_load(f)
-    return config_dict["public_checkpoint"]
+    return Path(config_dict["public_checkpoint"])