Skip to content

GitLab

Unverified Commit 6b7038ba authored by fzyzcjy's avatar fzyzcjy Committed by GitHub
Browse files

Speedup warmup when DP > 1 (#4695)

parent 57eec0bf
Show whitespace changes
Inline Side-by-side
Showing with 9 additions and 10 deletions
python/sglang/srt/entrypoints/http_server.py
View file @ 6b7038ba
...@@ -730,9 +730,9 @@ def _wait_and_warmup( ...@@ -730,9 +730,9 @@ def _wait_and_warmup(
}, },
} }
if server_args.skip_tokenizer_init: if server_args.skip_tokenizer_init:
json_data["input_ids"] = [10, 11, 12] json_data["input_ids"] = [[10, 11, 12] for _ in range(server_args.dp_size)]
else: else:
json_data["text"] = "The capital city of France is" json_data["text"] = ["The capital city of France is"] * server_args.dp_size
# Debug dumping # Debug dumping
if server_args.debug_tensor_dump_input_file: if server_args.debug_tensor_dump_input_file:
...@@ -743,7 +743,6 @@ def _wait_and_warmup( ...@@ -743,7 +743,6 @@ def _wait_and_warmup(
json_data["sampling_params"]["max_new_tokens"] = 0 json_data["sampling_params"]["max_new_tokens"] = 0
try: try:
for i in range(server_args.dp_size):
res = requests.post( res = requests.post(
url + request_name, url + request_name,
json=json_data, json=json_data,
......
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