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

Add Sequence-Level KD #2220

Merged
merged 8 commits into from
Oct 11, 2024
Merged

Add Sequence-Level KD #2220

Changes from 1 commit
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
4d368ec
Fix templates for dpo, etc.
mst272 Aug 15, 2024
36b6a6b
Update dpo.py
mst272 Aug 16, 2024
496a0e0
make this a utility.
mst272 Aug 18, 2024
697c164
Add Sequence-Level KD
mst272 Oct 11, 2024
d44af53
Merge branch 'main' of https://github.com/mst272/trl
mst272 Oct 11, 2024
f139c47
add to the docs-strings and the documentation
mst272 Oct 11, 2024
514cfe9
reviewed
mst272 Oct 11, 2024
875928b
Update docs/source/gkd_trainer.md
kashif Oct 11, 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
Update dpo.py 
Add the third issue fixs
  • Loading branch information
@mst272
mst272 authored Aug 16, 2024
commit 36b6a6b922aa0972c72ba402197f36b2af0b5c96
10 changes: 8 additions & 2 deletions examples/scripts/dpo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -155,14 +155,20 @@
ds[key] = ds[key].select(range(50))

def process(row):
# judge the third issue.
chat_right = True
# judge chat template
prompt = row["chosen"][:-1]
chosen = [row["chosen"][-1]]
test_prompt = tokenizer.apply_chat_template(prompt, tokenize=False)
test_chosen = tokenizer.apply_chat_template(chosen, tokenize=False)
try:
test_chosen = tokenizer.apply_chat_template(chosen, tokenize=False)
except Exception as e:
chat_right = False
conversation = copy.deepcopy(prompt)
conversation.append(row["chosen"][-1])
if tokenizer.apply_chat_template(conversation) == test_prompt + test_chosen:
# the chat_right is for the third issue, the subsequent checks is for the first and second issues.
if chat_right and tokenizer.apply_chat_template(conversation) == test_prompt + test_chosen:
row["prompt"] = tokenizer.apply_chat_template(row["chosen"][:-1], tokenize=False)
row["chosen"] = tokenizer.apply_chat_template([row["chosen"][-1]], tokenize=False)
row["rejected"] = tokenizer.apply_chat_template([row["rejected"][-1]], tokenize=False)
Expand Down