Skip to content

[Tunix] Update BaseAgent to accept observations with a "prompts" key rather than "question". #1164

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
copybara-service merged 1 commit into from
Feb 26, 2026
+8 −7
Merged

[Tunix] Update BaseAgent to accept observations with a "prompts" key rather than "question". #1164

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
7 changes: 1 addition & 6 deletions examples/deepscaler/train_deepscaler_nb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -304,11 +304,6 @@ def process_item(item):
"Let's think step by step, and put your final answer within \\boxed{}."
)
prompt = f"{question} {instruction}"
prompt = tokenizer.apply_chat_template(
[{"role": "user", "content": prompt}],
tokenize=False,
add_generation_prompt=True,
)

return {
"prompts": prompt,
Expand All @@ -326,7 +321,7 @@ def process_item(item):
tokenizer_source = MODEL_PATH if NOTEBOOK_ENV == "g3" else MODEL_VERSION
tokenizer = AutoTokenizer.from_pretrained(tokenizer_source)

chat_parser = parser.QwenChatTemplateParser(tokenizer)
chat_parser = parser.DefaultChatTemplateParser(tokenizer)

# %%
train_dataset, test_dataset = create_datasets()
Expand Down
8 changes: 7 additions & 1 deletion tunix/rl/agentic/agents/base_agent.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -168,7 +168,13 @@ def _observation_to_messages(
info: Additional information from the environment.
"""
del reward, done, info # Unused in default implementation.
if isinstance(observation, dict) and "question" in observation:
# prompts should not be applied with template beforehand to avoid double
# templating.
if isinstance(observation, dict) and "prompts" in observation:
self._messages.append(
{"role": "user", "content": observation["prompts"]}
)
elif isinstance(observation, dict) and "question" in observation:
self._messages.append(
{"role": "user", "content": observation["question"]}
)
Expand Down
Loading