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

feat: add llama3.2-11b-vision in all in one (#12207)

Browse source 
* feat: add llama3.2-11b-vision in all in one

* fix: change model

* fix: change name

* fix: add a space

* fix: switch import
This commit is contained in:
Chu,Youcheng 2024-10-16 10:32:11 +08:00 committed by GitHub
parent c9ac39fc1e
commit f17cc4fdee
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 9 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
python/llm/dev/benchmark/all-in-one/run.py
View file

@ -41,6 +41,8 @@ LLAMA_IDS = ['meta-llama/Llama-2-7b-chat-hf','meta-llama/Llama-2-13b-chat-hf',
'decapoda-research/llama-65b-hf','lmsys/vicuna-7b-v1.5',
'lmsys/vicuna-13b-v1.3','lmsys/vicuna-33b-v1.3','project-baize/merged-baize-30b']
LLAMA3_VISION_IDS = ['meta-llama/Llama-3.2-11B-Vision-Instruct']
CHATGLM_IDS = ['THUDM/chatglm-6b', 'THUDM/chatglm2-6b', 'THUDM/chatglm3-6b']
LLAVA_IDS = ['liuhaotian/llava-v1.5-7b']
@ -770,6 +772,13 @@ def run_optimize_model_gpu(repo_id,
model = optimize_model(model, low_bit=low_bit)
tokenizer = LlamaTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = model.to('xpu')
elif repo_id in LLAMA3_VISION_IDS:
from transformers import MllamaForConditionalGeneration
model = MllamaForConditionalGeneration.from_pretrained(model_path, trust_remote_code=True,
low_cpu_mem_usage=True).eval()
model = optimize_model(model, low_bit=low_bit, modules_to_not_convert=["multi_modal_projector"])
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = model.to('xpu')
else:
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype='auto', low_cpu_mem_usage=True,
trust_remote_code=True, use_cache=True).eval()