Skip to content

Update args_train.sh #78

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 4 commits into
base: megatron_script_tecent
Choose a base branch
from
Open

Update args_train.sh #78

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8d4398c
Update args_train.sh
strint Apr 30, 2023
6d0c582
Update args_train.sh
strint Apr 30, 2023
f08490d
rm sed
strint May 2, 2023
b2d6dd8
Update args_train.sh
strint May 2, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
67 changes: 33 additions & 34 deletions onebench/libai/args_train.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,10 +30,10 @@ UNSET_DROPOUT=${23:-false}

ONEFLOW_COMMIT=$(python3 -c 'import oneflow; print(oneflow.__git_commit__)')

sed -i '/import time/a\import os' ./libai/engine/trainer.py
sed -i '/for self.iter in range(start_iter, max_iter):/a\ if self.iter == 99: \
cmd = "nvidia-smi --query-gpu=timestamp,name,driver_version,utilization.gpu,utilization.memory,memory.total,memory.free,memory.used --format=csv" \
os.system(cmd)' ./libai/engine/trainer.py
#sed -i '/import time/a\import os' ./libai/engine/trainer.py
#sed -i '/for self.iter in range(start_iter, max_iter):/a\ if self.iter == 99: \
# cmd = "nvidia-smi --query-gpu=timestamp,name,driver_version,utilization.gpu,utilization.memory,memory.total,memory.free,memory.used --format=csv" \
# os.system(cmd)' ./libai/engine/trainer.py

GPU_NAME="$(nvidia-smi -i 0 --query-gpu=gpu_name --format=csv,noheader)"
GPU_NAME="${GPU_NAME// /_}"
Expand Down Expand Up @@ -74,7 +74,7 @@ LOG_FILENAME=${TRAN_MODEL}_${RUN_TYPE}_nl${NUM_LAYER}_nah${NUM_ATT_HEADS}_hs${HI

if [[ $UNSET_DROPOUT = "true" ]]; then
#sed -i 's/persistent_workers=True/#persistent_workers=True/g' ./libai/data/build.py
sed -i 's/shuffle=True/shuffle=False/g' ./libai/data/build.py
#sed -i 's/shuffle=True/shuffle=False/g' ./libai/data/build.py
hidden_dropout_prob=0.0
attention_probs_dropout_prob=0.0
bias_dropout_fusion=false
Expand All @@ -83,7 +83,7 @@ fi

if [[ $SAVE_MODEL = "false" ]]; then
#sed -i 's/hooks.PeriodicCheckpointer/#&/' ./libai/engine/default.py
sed -i '/if self.cfg.train.evaluation.enabled:/i\ ret.pop()' ./libai/engine/default.py
#sed -i '/if self.cfg.train.evaluation.enabled:/i\ ret.pop()' ./libai/engine/default.py
LOG_FOLDER=$LOG_FOLDER/${ONEFLOW_COMMIT}
fi

Expand All @@ -96,35 +96,34 @@ echo LOG_FILENAME=$LOG_FILENAME
python3 -m oneflow.distributed.launch \
--nproc_per_node $GPUS_PER_NODE --nnodes $NNODES --node_rank $NODE_RANK --master_addr $MASTER_ADDR --master_port $MASTER_PORT \
tools/train_net.py \
--resume \
--config-file $CONFIG \
model.cfg.hidden_dropout_prob=$hidden_dropout_prob \
model.cfg.attention_probs_dropout_prob=$attention_probs_dropout_prob \
model.cfg.bias_dropout_fusion=$bias_dropout_fusion \
model.cfg.hidden_layers=$NUM_LAYER \
model.cfg.hidden_size=$HIDDEN_SIZE \
model.cfg.num_attention_heads=$NUM_ATT_HEADS \
model.cfg.intermediate_size=$INTERMEDIATE_SIZE \
model.cfg.ffn_hidden_size=$INTERMEDIATE_SIZE \
--config-file $CONFIG \
model.cfg.hidden_dropout_prob=$hidden_dropout_prob \
model.cfg.attention_probs_dropout_prob=$attention_probs_dropout_prob \
model.cfg.bias_dropout_fusion=$bias_dropout_fusion \
model.cfg.hidden_layers=$NUM_LAYER \
model.cfg.hidden_size=$HIDDEN_SIZE \
model.cfg.num_attention_heads=$NUM_ATT_HEADS \
model.cfg.intermediate_size=$INTERMEDIATE_SIZE \
model.cfg.ffn_hidden_size=$INTERMEDIATE_SIZE \
model.cfg.head_size=$HEAD_SIZE \
graph.enabled=$GRAPH_ENABLED \
train.dist.pipeline_num_layers=$NUM_LAYER \
train.train_micro_batch_size=$MICRO_BATCH_SIZE \
train.global_batch_size=$GLOBAL_BATCH_SIZE \
train.dist.tensor_parallel_size=$MP \
train.dist.pipeline_parallel_size=$PP \
train.amp.enabled=$USE_FP16 \
train.activation_checkpoint.enabled=$ACTIVATION_CHECKPOINT \
train.num_accumulation_steps=$ACC \
train.evaluation.enabled=$EVALUATION_ENABLED \
train.evaluation.eval_iter=$EVAL_ITER \
train.train_iter=$TRAIN_ITERS \
train.train_epoch=$TRAIN_EPOCH \
train.log_period=$LOG_PERIOD \
train.zero_optimization.enabled=$ZERO_ENABLE \
train.zero_optimization.stage=$ZERO_STAGE \
train.load_weight=$LOAD_WEIGHT \
train.checkpoint.period=$save_checkpoint_period \
graph.enabled=$GRAPH_ENABLED \
train.dist.pipeline_num_layers=$NUM_LAYER \
train.train_micro_batch_size=$MICRO_BATCH_SIZE \
train.global_batch_size=$GLOBAL_BATCH_SIZE \
train.dist.tensor_parallel_size=$MP \
train.dist.pipeline_parallel_size=$PP \
train.amp.enabled=$USE_FP16 \
train.activation_checkpoint.enabled=$ACTIVATION_CHECKPOINT \
train.num_accumulation_steps=$ACC \
train.evaluation.enabled=$EVALUATION_ENABLED \
train.evaluation.eval_iter=$EVAL_ITER \
train.train_iter=$TRAIN_ITERS \
train.train_epoch=$TRAIN_EPOCH \
train.log_period=$LOG_PERIOD \
train.zero_optimization.enabled=$ZERO_ENABLE \
train.zero_optimization.stage=$ZERO_STAGE \
train.load_weight=$LOAD_WEIGHT \
train.checkpoint.period=$save_checkpoint_period \
train.output_dir=$LOG_FILENAME 2>&1 | tee ${LOG_FILENAME}/output.log


Expand Down