[Feat] Return hidden states (experimental) (#3364)
Co-authored-by: Chayenne <zhaochen20@outlook.com>
This commit is contained in:
@@ -156,6 +156,10 @@ class TpModelWorkerClient:
|
||||
logits_output.input_token_logprobs = (
|
||||
logits_output.input_token_logprobs.to("cpu", non_blocking=True)
|
||||
)
|
||||
if logits_output.hidden_states is not None:
|
||||
logits_output.hidden_states = logits_output.hidden_states.to(
|
||||
"cpu", non_blocking=True
|
||||
)
|
||||
next_token_ids = next_token_ids.to("cpu", non_blocking=True)
|
||||
copy_done.record()
|
||||
|
||||
|
||||
Reference in New Issue
Block a user