xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-07 11:12:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

Normalize lm-eval command between baseline and correctness test (#18560)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin 2025-07-10 22:27:32 +09:00 committed by GitHub
parent be1e128dfb
commit 1a4f35e2ea
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh
View File

@ -46,6 +46,6 @@ while getopts "m:b:l:f:t:" OPT; do
done done
lm_eval --model vllm \ lm_eval --model vllm \
--model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,distributed_executor_backend=ray,trust_remote_code=true,max_model_len=4096" \ --model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,add_bos_token=true,trust_remote_code=true,max_model_len=4096" \
--tasks gsm8k --num_fewshot "$FEWSHOT" --limit "$LIMIT" \ --tasks gsm8k --num_fewshot "$FEWSHOT" --limit "$LIMIT" \
--batch_size "$BATCH_SIZE" --batch_size "$BATCH_SIZE"

4
.buildkite/lm-eval-harness/test_lm_eval_correctness.py
View File

@ -18,12 +18,14 @@ RTOL = 0.08
def launch_lm_eval(eval_config, tp_size): def launch_lm_eval(eval_config, tp_size):
trust_remote_code = eval_config.get("trust_remote_code", False) trust_remote_code = eval_config.get("trust_remote_code", False)
max_model_len = eval_config.get("max_model_len", 4096)
model_args = ( model_args = (
f"pretrained={eval_config['model_name']}," f"pretrained={eval_config['model_name']},"
f"tensor_parallel_size={tp_size}," f"tensor_parallel_size={tp_size},"
f"enforce_eager=true," f"enforce_eager=true,"
f"add_bos_token=true," f"add_bos_token=true,"
f"trust_remote_code={trust_remote_code}" f"trust_remote_code={trust_remote_code},"
f"max_model_len={max_model_len}"
) )
results = lm_eval.simple_evaluate( results = lm_eval.simple_evaluate(
model="vllm", model="vllm",