File size: 2,505 Bytes
63fba53 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 | import torch
from transformers import AutoModel, AutoTokenizer
from .utils import load_image
class InternVL:
def __init__(self,model_path,args):
super().__init__()
self.llm = AutoModel.from_pretrained(
model_path,
torch_dtype=torch.bfloat16,
low_cpu_mem_usage=True,
trust_remote_code=True,
device_map="cuda",
attn_implementation="flash_attention_2"
)
self.tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True, use_fast=False)
self.generation_config ={
'max_new_tokens': args.max_new_tokens,
'repetition_penalty': args.repetition_penalty,
'temperature' : args.temperature,
'top_p' : args.top_p
}
def process_messages(self,messages):
prompt = ""
if "system" in messages:
prompt = messages["system"]
if "image" in messages:
text = messages["prompt"]
prompt = prompt + "\n" + "<image>" + '\n' + text
image = messages["image"]
pixel_value = load_image(image).to(torch.bfloat16).to("cuda")
elif "images" in messages:
text = messages["prompt"]
images = messages["images"]
for i,image in enumerate(images):
prompt = prompt + f"<image_{i+1}>: <image>" + "\n"
prompt = prompt + '\n' + text
pixel_value = [load_image(image).to(torch.bfloat16).to("cuda") for image in images]
pixel_value = torch.cat(pixel_value,dim=0)
else:
text = messages["prompt"]
prompt = prompt + "\n" + text
pixel_value = None
llm_inputs = {
"prompt": prompt,
"pixel_values": pixel_value
}
return llm_inputs
def generate_output(self,messages):
llm_inputs = self.process_messages(messages)
question = llm_inputs["prompt"]
pixel_values = llm_inputs["pixel_values"]
response, history = self.llm.chat(self.tokenizer, pixel_values, question, self.generation_config,
history=None, return_history=True)
return response
def generate_outputs(self,messages_list):
res = []
for messages in messages_list:
result = self.generate_output(messages)
res.append(result)
return res
|