Skip to content

GitLab

Unverified Commit 0967110e authored by Woosuk Kwon's avatar Woosuk Kwon Committed by GitHub
Browse files

[Minor] Update the tqdm bar for parallel sampling (#14571) 


Signed-off-by: default avatarWoosuk Kwon <woosuk.kwon@berkeley.edu>
parent fb0acb6c
Show whitespace changes
Inline Side-by-side
Showing with 3 additions and 2 deletions
vllm/entrypoints/llm.py
View file @ 0967110e
...@@ -1384,8 +1384,9 @@ class LLM: ...@@ -1384,8 +1384,9 @@ class LLM:
if use_tqdm: if use_tqdm:
if isinstance(output, RequestOutput): if isinstance(output, RequestOutput):
# Calculate tokens only for RequestOutput # Calculate tokens only for RequestOutput
n = len(output.outputs)
assert output.prompt_token_ids is not None assert output.prompt_token_ids is not None
total_in_toks += len(output.prompt_token_ids) total_in_toks += len(output.prompt_token_ids) * n
in_spd = total_in_toks / pbar.format_dict["elapsed"] in_spd = total_in_toks / pbar.format_dict["elapsed"]
total_out_toks += sum( total_out_toks += sum(
len(stp.token_ids) for stp in output.outputs) len(stp.token_ids) for stp in output.outputs)
...@@ -1394,7 +1395,7 @@ class LLM: ...@@ -1394,7 +1395,7 @@ class LLM:
pbar.postfix = ( pbar.postfix = (
f"est. speed input: {in_spd:.2f} toks/s, " f"est. speed input: {in_spd:.2f} toks/s, "
f"output: {out_spd:.2f} toks/s") f"output: {out_spd:.2f} toks/s")
pbar.update(len(output.outputs)) pbar.update(n)
else: else:
pbar.update(1) pbar.update(1)
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment