alibaba / alibaba/ROLL

code_dapo agentic task fails

Open
#309 2 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
3.4k
Forks
312
Avg merge
1h 2m
Merged PRs (30d)
2

Description

I encountered an issue of being too long and uncut while running your `examples/qwen3_agentic_gem/gem_math_dapo.yaml`configuration: index 0 is out of bounds for dimension 1 with size 0:

```
(EnvironmentWorker(val_env-0) pid=2125841) File "/code/roll/distributed/scheduler/generate_scheduler.py", line 913, in generate_one_request [repeated 6x across cluster]

  | 16:52:17.393 | (EnvironmentWorker(val_env-0) pid=2125841) output: DataProto = postprocess_generate( [repeated 6x across cluster]
  | 16:52:17.393 | (EnvironmentWorker(val_env-0) pid=2125841) File "/code/roll/utils/functionals.py", line 815, in postprocess_generate [repeated 6x across cluster]
  | 16:52:17.393 | (EnvironmentWorker(val_env-0) pid=2125841) response_mask = get_pad_mask(response_id=response, pad_token=pad_token_id, dtype=attention_mask.dtype) [repeated 6x across cluster]
  | 16:52:17.393 | (EnvironmentWorker(val_env-0) pid=2125841) File "/code/roll/utils/functionals.py", line 318, in get_pad_mask [repeated 6x across cluster]
  | 16:52:17.393 | (EnvironmentWorker(val_env-0) pid=2125841) not (pad_mask[:, 0] == 0).logical_and(pad_mask.sum(-1) != 0).any() [repeated 6x across cluster]
  | 16:52:17.393 | (EnvironmentWorker(val_env-0) pid=2125841) **IndexError: index 0 is out of bounds for dimension 1 with size 0** [repeated 6x across cluster]

Contributor guide

No contributing guide indexed for this repository

Research direction

Reproduce the failure with examples/qwen3_agentic_gem/gem_math_dapo.yaml. Start by tracing the reported call from distributed/scheduler/generate_scheduler.py:913 through roll/utils/functionals.py:815 and get_pad_mask at line 318. Done means the reported configuration no longer raises the empty-dimension IndexError and the behavior is verified against this reproduction.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
ai, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.