github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-07 12:01:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix missing tags parameter

Browse Source
This commit is contained in:
YeAnbang
2025-05-21 10:51:32 +08:00
parent 88e3b09c79
commit 78a06f5ce3
4 changed files with 27 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
applications/ColossalChat/coati/distributed/grpo_consumer.py
View File

@@ -120,16 +120,7 @@ class GRPOConsumer(BaseConsumer):
"either max_tokens (vllm) or max_new_tokens (transformers) must be set in generate_config."
)
# Initialize verifiable reward.
response_format_tags = (
{
"think_start": {"text": "<think>", "num_occur": 1},
"think_end": {"text": "</think>", "num_occur": 1},
"answer_start": {"text": "<answer>", "num_occur": 1},
"answer_end": {"text": "</answer>", "num_occur": 1},
}
if grpo_config.get("reward_fn_type") == "think_answer_tags"
else None
)
response_format_tags = grpo_config.get("response_format_tags", None)
reward_model_kwargs = {
k: v
for k, v in grpo_config.items()