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

[BugFix] Fix reward model backend #593

Merged
merged 2 commits into from
Oct 28, 2024
Merged

[BugFix] Fix reward model backend #593

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a43bd2c
[BugFix] Fix reward model backend
sufubao Oct 28, 2024
9cefe8a
[misc] Assert conflicting modes
sufubao Oct 28, 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
9 changes: 8 additions & 1 deletion lightllm/server/api_server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -524,7 +524,14 @@ def main():
assert args.disable_cudagraph

# 这些模式不能同时设置。
assert [args.splitfuse_mode, args.beam_mode, args.diverse_mode, args.token_healing_mode].count(True) <= 1
assert [
args.splitfuse_mode,
args.beam_mode,
args.diverse_mode,
args.token_healing_mode,
args.use_reward_model,
args.return_all_prompt_logprobs,
].count(True) <= 1
# 部分模式目前还无法与dynamic_prompt_cache一起跑，to do。
if args.use_dynamic_prompt_cache:
assert args.beam_mode is False
Expand Down
4 changes: 2 additions & 2 deletions lightllm/server/router/model_infer/mode_backend/continues_batch/impl_for_reward_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@ def forward(self, batch_id, is_prefill):
kwargs, run_reqs = prepare_prefill_inputs(batch, self.radix_cache, self.is_multimodal)

scores: torch.Tensor = self.model.forward(**kwargs)
scores = scores.detach().cpu().numpy()
scores = scores[0].detach().cpu().numpy()

next_token_id = 1
next_token_logprob = 1.0
Expand All @@ -36,7 +36,7 @@ def forward(self, batch_id, is_prefill):
req_obj.out_token_id_count[next_token_id] += 1
req_obj.finish_status = FinishStatus.FINISHED_STOP

metadata = {"id": int(next_token_id), "logprob": float(next_token_logprob), "score": float(score[0])}
metadata = {"id": int(next_token_id), "logprob": float(next_token_logprob), "score": float(score)}

output_dict[req_obj.r_id] = (
req_obj.req_status,
Expand Down
Loading