ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

LLM: fix input length condition in deepspeed all-in-one benchmark. (#11185)

Browse source
This commit is contained in:
Cengguang Zhang 2024-06-03 10:05:43 +08:00 committed by GitHub
parent 401013a630
commit 3eb13ccd8c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
python/llm/dev/benchmark/all-in-one/run.py
View file

@ -1518,7 +1518,7 @@ def run_deepspeed_optimize_model_gpu(repo_id,
# in_len.txt maybe shorter than we need, # in_len.txt maybe shorter than we need,
# use much longer context to make sure input length # use much longer context to make sure input length
test_length = min(in_len*2, 8192) test_length = min(in_len*2, 8192)
while test_length not in [32, 256, 1024, 2048, 8192]: while test_length not in [32, 256, 1024, 2048, 8192] and test_length < 8192:
test_length = test_length * 2 test_length = test_length * 2
input_str = open(f"prompt/{test_length}.txt", 'r').read() input_str = open(f"prompt/{test_length}.txt", 'r').read()
# As different tokenizer has different encodings, # As different tokenizer has different encodings,