Remove monkey_patch_vllm_dummy_weight_loader (#2064)
This commit is contained in:
@@ -895,7 +895,7 @@ class Scheduler:
|
||||
logits_output, next_token_ids, bid = result
|
||||
|
||||
if self.enable_overlap:
|
||||
logits_output, next_token_ids = self.tp_worker.resulve_batch_result(bid)
|
||||
logits_output, next_token_ids = self.tp_worker.resolve_batch_result(bid)
|
||||
else:
|
||||
# Move next_token_ids and logprobs to cpu
|
||||
if batch.return_logprob:
|
||||
@@ -970,7 +970,7 @@ class Scheduler:
|
||||
self.num_generated_tokens += len(batch.reqs)
|
||||
|
||||
if self.enable_overlap:
|
||||
logits_output, next_token_ids = self.tp_worker.resulve_batch_result(bid)
|
||||
logits_output, next_token_ids = self.tp_worker.resolve_batch_result(bid)
|
||||
next_token_logprobs = logits_output.next_token_logprobs
|
||||
else:
|
||||
# Move next_token_ids and logprobs to cpu
|
||||
|
||||
Reference in New Issue
Block a user