Skip to content

Length gen experiments #40

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

Merged
Benjamin-Walker merged 2 commits into from
Oct 24, 2025
Merged

Length gen experiments #40

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ea7e0e8
Length gen experiments
Benjamin-Walker Oct 23, 2025
295b09a
Removed padding length for mLSTM on A5 length gen
Benjamin-Walker Oct 24, 2025
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
1 change: 1 addition & 0 deletions length_gen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,6 +94,7 @@ def build_model(model_name, config, data_dim, label_dim, device):
dropout_rate=config.get("dropout_rate", 0.01),
use_glu=config.get("use_glu", False),
diagonal=config.get("diagonal", False),
diagonal_dense=config.get("diagonal_dense", False),
fwht=config.get("fwht", False),
second_embedding=config.get("second_embedding", False),
rank=config.get("rank", 0),
Expand Down
20 changes: 15 additions & 5 deletions train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,6 +118,7 @@ def train_model(
criterion = nn.CrossEntropyLoss()
step = 0
total_loss = 0
best_val_acc = 0
steps = []
val_accs = []
start_time = time.time()
Expand Down Expand Up @@ -232,8 +233,10 @@ def train_model(

checkpoint_filename = f"checkpoint_{task}_{model_name}_{time_str}.pt"
checkpoint_path = os.path.join("checkpoints", checkpoint_filename)
torch.save(checkpoint, checkpoint_path)
print(f"Saved model checkpoint to: {checkpoint_path}")
if accuracy > best_val_acc:
best_val_acc = accuracy
torch.save(checkpoint, checkpoint_path)
print(f"Saved model checkpoint to: {checkpoint_path}")

early_stop = accuracy > early_stop_threshold

Expand Down Expand Up @@ -336,9 +339,15 @@ def train_dataloader_multilength():
dataloader = {"train": train_dataloader_multilength(), "val": val_dataloader}

elif task == "A5_generalise":
train_padding_length = 128
if model_name == "lcde":
train_padding_length = 20
if model_name[:8] == "deltanet" or model_name == "deltaproduct":
train_padding_length = 65
elif model_name == "xlstm":
Copy link
Copy Markdown
Owner Author

@Benjamin-Walker Benjamin-Walker Oct 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Remove as don't perform mLSTM on length generalisation due to needing fixed input length

if slstm_at == []:
train_padding_length = 128
else:
train_padding_length = 40
else:
train_padding_length = 40
val_padding_length = 128
train_dataloader, _, data_dim, label_dim = create_group_dataloaders(
group="A5",
Expand All @@ -360,6 +369,7 @@ def train_dataloader_multilength():
train_split=1.0,
seed=2 * seed,
)

dataloader = {"train": train_dataloader, "val": val_dataloader}

else:
Expand Down