Skip to content
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

Jump to bottom

handle tokenization more correctly, clean up CI #62

Merged
merged 9 commits into from
May 30, 2024
Merged

handle tokenization more correctly, clean up CI #62

Changes from 1 commit
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
2527625
handle tokenization more correctly, clean up CI
charlesfrye May 30, 2024
3f5f018
fix codellama module saving
charlesfrye May 30, 2024
2c65751
unify loss check now that pythia training is well-tuned
charlesfrye May 30, 2024
1606515
switch to a100s by default
charlesfrye May 30, 2024
1f3e482
adjust epoch counts
charlesfrye May 30, 2024
b7a7d10
resolve issue with codellama
charlesfrye May 30, 2024
fb7d9b7
save model only at end of training in ci
charlesfrye May 30, 2024
62e612d
25 epochs for all models except pythia
charlesfrye May 30, 2024
9f63e2c
give codellama more room breathe in CI
charlesfrye May 30, 2024
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
Prev Previous commit
Next Next commit
25 epochs for all models except pythia
  • Loading branch information
@charlesfrye
charlesfrye committed May 30, 2024
commit 62e612d8abbf02a65a0cf5aecc8ff8bb6d236058
14 changes: 5 additions & 9 deletions ci/prep_for_ci.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,18 +18,14 @@ def main(config: str, data: str):
cfg["weight_decay"] = 0

val_set_size = 0.5
num_epochs = 50
if (
"Meta-Llama-3-8B" in cfg["base_model"]
or "codellama/CodeLlama-7b-Instruct-hf" in cfg["base_model"]
):
num_epochs = num_epochs // 2
elif "pythia" in cfg["base_model"]:
num_epochs = num_epochs * 2

num_epochs = 25
if "pythia" in cfg["base_model"]:
num_epochs = num_epochs * 4

cfg["val_set_size"] = val_set_size
cfg["num_epochs"] = num_epochs
cfg["eval_steps"] = num_epochs // 10
cfg["eval_steps"] = num_epochs // 5
cfg.pop("evals_per_epoch", None) # incompatible with eval_steps
cfg.pop("sample_packing", False) # requires larger dataset

Expand Down
Loading