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

Add MiniCPM-V-2_6 to iGPU Perf (#11810)

Browse source 
* Add MiniCPM-V-2_6 to iGPU Perf

* keep last model in yaml

* fix MINICPM_V_IDS

* Restore tested model list

* Small fix

---------

Co-authored-by: Yuwen Hu <yuwen.hu@intel.com>
This commit is contained in:
Jin, Qiao 2024-08-16 18:41:21 +08:00 committed by GitHub
parent 96796f95cb
commit 9f17234f3b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 21 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
python/llm/dev/benchmark/all-in-one/run.py
View file

@ -992,6 +992,13 @@ def run_transformer_int4_gpu_win(repo_id,
trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding).eval()
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = model.to('xpu')
elif repo_id in MINICPM_V_IDS:
model = AutoModel.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
modules_to_not_convert=["vpm", "resampler"],
trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding).eval()
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = model.to('xpu')
model = model.llm
else:
model = AutoModelForCausalLM.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding).eval()
@ -1108,6 +1115,14 @@ def run_transformer_int4_fp16_gpu_win(repo_id,
torch_dtype=torch.float16).eval()
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = model.to('xpu')
elif repo_id in MINICPM_V_IDS:
model = AutoModel.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
modules_to_not_convert=["vpm", "resampler"],
trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding,
torch_dtype=torch.float16).eval()
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = model.to('xpu')
model = model.llm
else:
model = AutoModelForCausalLM.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding,

1
python/llm/test/benchmark/igpu-perf/1024-128_437.yaml
View file

@ -5,6 +5,7 @@ repo_id:
- 'microsoft/Phi-3-mini-4k-instruct'
- 'microsoft/Phi-3-mini-128k-instruct'
- 'microsoft/phi-3-vision-128k-instruct'
- 'openbmb/MiniCPM-V-2_6'
local_model_hub: 'path to your local model hub'
warm_up: 1
num_trials: 3

1
python/llm/test/benchmark/igpu-perf/1024-128_int4_fp16_437.yaml
View file

@ -5,6 +5,7 @@ repo_id:
- 'microsoft/Phi-3-mini-4k-instruct'
- 'microsoft/Phi-3-mini-128k-instruct'
- 'microsoft/phi-3-vision-128k-instruct'
- 'openbmb/MiniCPM-V-2_6'
local_model_hub: 'path to your local model hub'
warm_up: 1
num_trials: 3

1
python/llm/test/benchmark/igpu-perf/2048-256_int4_fp16_437.yaml
View file

@ -5,6 +5,7 @@ repo_id:
- 'microsoft/Phi-3-mini-4k-instruct'
- 'microsoft/Phi-3-mini-128k-instruct'
- 'microsoft/phi-3-vision-128k-instruct'
- 'openbmb/MiniCPM-V-2_6'
local_model_hub: 'path to your local model hub'
warm_up: 1
num_trials: 3

1
python/llm/test/benchmark/igpu-perf/3072-384_int4_fp16_437.yaml
View file

@ -5,6 +5,7 @@ repo_id:
- 'microsoft/Phi-3-mini-4k-instruct'
- 'microsoft/Phi-3-mini-128k-instruct'
- 'microsoft/phi-3-vision-128k-instruct'
- 'openbmb/MiniCPM-V-2_6'
local_model_hub: 'path to your local model hub'
warm_up: 1
num_trials: 3

1
python/llm/test/benchmark/igpu-perf/32-32_int4_fp16_437.yaml
View file

@ -5,6 +5,7 @@ repo_id:
- 'microsoft/Phi-3-mini-4k-instruct'
- 'microsoft/Phi-3-mini-128k-instruct'
- 'microsoft/phi-3-vision-128k-instruct'
- 'openbmb/MiniCPM-V-2_6'
local_model_hub: 'path to your local model hub'
warm_up: 3
num_trials: 5

1
python/llm/test/benchmark/igpu-perf/4096-512_int4_fp16_437.yaml
View file

@ -5,6 +5,7 @@ repo_id:
- 'microsoft/Phi-3-mini-4k-instruct'
- 'microsoft/Phi-3-mini-128k-instruct'
- 'microsoft/phi-3-vision-128k-instruct'
- 'openbmb/MiniCPM-V-2_6'
local_model_hub: 'path to your local model hub'
warm_up: 1
num_trials: 3