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

Small fix to all-in-one benchmark scripts (#12317)

Browse source
This commit is contained in:
Yuwen Hu 2024-11-01 19:16:25 +08:00 committed by GitHub
parent 48123af463
commit 20755e8077
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 22 additions and 20 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

42
python/llm/dev/benchmark/all-in-one/run.py
View file

@ -215,7 +215,7 @@ def run_model(repo_id, test_api, in_out_pairs, local_model_hub=None, warm_up=1,
result[in_out_pair][-1][6] if any(keyword in test_api for keyword in ['int4_gpu', 'int4_fp16_gpu_win', 'int4_loadlowbit_gpu', 'int4_fp16_loadlowbit_gpu', 'fp16_gpu', 'deepspeed_optimize_model_gpu']) and not lookahead else 'N/A', result[in_out_pair][-1][6] if any(keyword in test_api for keyword in ['int4_gpu', 'int4_fp16_gpu_win', 'int4_loadlowbit_gpu', 'int4_fp16_loadlowbit_gpu', 'fp16_gpu', 'deepspeed_optimize_model_gpu']) and not lookahead else 'N/A',
streaming if 'win' in test_api else 'N/A', streaming if 'win' in test_api else 'N/A',
use_fp16_torch_dtype if 'pipeline_parallel_gpu' in test_api else 'N/A', use_fp16_torch_dtype if 'pipeline_parallel_gpu' in test_api else 'N/A',
group_size], group_size if 'transformers_int4_npu_win' in test_api else 'N/A'],
) )
@ -604,16 +604,16 @@ def run_transformer_int4_gpu(repo_id,
def transformers_int4_npu_win(repo_id, def transformers_int4_npu_win(repo_id,
local_model_hub, local_model_hub,
in_out_pairs, in_out_pairs,
warm_up, warm_up,
num_trials, num_trials,
num_beams, num_beams,
low_bit, low_bit,
batch_size, batch_size,
optimize_model, optimize_model,
transpose_value_cache, transpose_value_cache,
npu_group_size): npu_group_size):
from ipex_llm.transformers.npu_model import AutoModel, AutoModelForCausalLM from ipex_llm.transformers.npu_model import AutoModel, AutoModelForCausalLM
from transformers import AutoTokenizer, LlamaTokenizer from transformers import AutoTokenizer, LlamaTokenizer
@ -681,15 +681,15 @@ def transformers_int4_npu_win(repo_id,
return result return result
def run_transformer_int4_loadlowbit_npu_win(repo_id, def run_transformer_int4_loadlowbit_npu_win(repo_id,
local_model_hub, local_model_hub,
in_out_pairs, in_out_pairs,
warm_up, warm_up,
num_trials, num_trials,
num_beams, num_beams,
low_bit, low_bit,
batch_size, batch_size,
optimize_model, optimize_model,
transpose_value_cache): transpose_value_cache):
from ipex_llm.transformers.npu_model import AutoModel, AutoModelForCausalLM from ipex_llm.transformers.npu_model import AutoModel, AutoModelForCausalLM
from transformers import AutoTokenizer, LlamaTokenizer from transformers import AutoTokenizer, LlamaTokenizer
@ -2196,6 +2196,8 @@ if __name__ == '__main__':
task = conf['task'] task = conf['task']
if 'optimize_model' in conf: if 'optimize_model' in conf:
optimize_model = conf['optimize_model'] optimize_model = conf['optimize_model']
if 'group_size' in conf:
group_size = conf['group_size']
if 'npu_group_size' in conf: if 'npu_group_size' in conf:
group_size = conf['npu_group_size'] group_size = conf['npu_group_size']
lookahead = False lookahead = False