Skip to content

fix ioi paths and check add_bos_token on tokenizer #9

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
oliveradk wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix ioi paths and check add_bos_token on tokenizer #9

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b2e3afe
fix ioi paths and check add_bos_token on tokenizer
oliveradk Jul 26, 2024
03b045b
comment on add_bos_token
oliveradk Aug 22, 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
6 changes: 4 additions & 2 deletions auto_circuit/tasks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,6 +205,8 @@ def init_task(self):
bs_2 = bs[1] if isinstance(bs, tuple) else bs
count_2 = b_count[1] if isinstance(b_count, tuple) else b_count
has_tokenizer = hasattr(model, "tokenizer") and model.tokenizer is not None
if has_tokenizer and hasattr(model.tokenizer, "add_bos_token"): # ensures bos token not added to answers/wrong answers for pythia models
model.tokenizer.add_bos_token = False
train_loader, test_loader = load_datasets_from_json(
model=model if has_tokenizer else None,
path=repo_path_to_abs_path(f"datasets/{self._dataset_name}.json"),
Expand Down Expand Up @@ -296,7 +298,7 @@ def init_task(self):
key="Indirect Object Identification Component Circuit",
name="Indirect Object Identification",
_model_def="gpt2-small",
_dataset_name="ioi_prompts",
_dataset_name="ioi/ioi_prompts",
batch_size=64,
batch_count=2,
_true_edge_func=ioi_true_edges,
Expand All @@ -306,7 +308,7 @@ def init_task(self):
key="Indirect Object Identification GPT2 Autoencoder Component Circuit",
name="Indirect Object Identification",
_model_def="gpt2-small",
_dataset_name="ioi_prompts",
_dataset_name="ioi/ioi_prompts",
batch_size=1,
batch_count=2,
_true_edge_func=None,
Expand Down