File size: 7,234 Bytes
1fc662a |
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 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 |
import os
from .base import BaseModel
from ..smp import *
from ..dataset import DATASET_TYPE
from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
import torch
import json
os.environ["TOKENIZERS_PARALLELISM"] = "false"
def load_model_tokenizer(checkpoint_path):
tokenizer = AutoTokenizer.from_pretrained(
checkpoint_path, trust_remote_code=True,
)
device_map = 'auto'
model = AutoModelForCausalLM.from_pretrained(
checkpoint_path,
device_map=device_map,
trust_remote_code=True,
torch_dtype=torch.bfloat16,
)
return model, tokenizer
class Baichuan(BaseModel):
INSTALL_REQ = False
INTERLEAVE = False
def __init__(self, sft=True, model_path=None):
assert model_path is not None
self.device = "cuda"
self.model_path = model_path
self.model, self.tokenizer = load_model_tokenizer(model_path)
self.model.bind_processor(self.tokenizer, training=False)
torch.cuda.empty_cache()
self.use_reserve_qa_prompt = sft
self.reserve_qa_start_prompt = "<C_Q>"
self.reserve_qa_end_prompt = "<C_A>"
self.task_prompt=""
self.options_system_prompt = ('Carefully read the following question and select the letter corresponding '
'to the correct answer. Highlight the applicable choices without giving '
'explanations. ')
self.wo_options_system_prompt = 'Carefully read the following question Answer the question directly. '
self.detail_system_prompt = 'Answer this question in detail and step by step. '
self.vqa_prompt = 'Answer the question using a single word or phrase. '
def generate_inner(self, message, dataset=None):
image_str, question = '', ''
for s in message:
if s['type'] == 'image':
if len(s["value"].split(".")[-1]) > 2:
image_dict = {"local": s["value"]}
else:
image_dict = {"base64": s["value"]}
image_str += f"<img_start_baichuan>{json.dumps(image_dict)}<img_end_baichuan>\n"
elif s['type'] == 'text':
question += s['value']
# sft version: <C_Q>...<C_A>
if self.use_reserve_qa_prompt:
prompt = "{}{}{}{}{}".format(self.reserve_qa_start_prompt, image_str, question, self.task_prompt, self.reserve_qa_end_prompt)
else:
prompt = "{}{}{}".format(image_str, question, self.task_prompt)
print("****************************** prompt ******************************")
print(prompt)
print("********************************************************************")
with torch.inference_mode():
ret = self.model.processor(prompt)
input_ids = ret.input_ids
try:
ret = self.model.generate(
inputs=torch.LongTensor([input_ids]).cuda(),
images=[torch.tensor(img, dtype=torch.float32).cuda() for img in images] if ret.images is not None else None,
patch_nums=ret.patch_nums,
images_grid=ret.images_grid,
max_new_tokens=1024, do_sample=False, top_k=5, top_p=0.85, temperature=0,
num_return_sequences=1, repetition_penalty=1.05,
use_cache=False
)
ret = self.tokenizer.batch_decode(ret[:, torch.LongTensor([input_ids]).to(self.device).shape[1]:], skip_special_tokens=True)[0].strip()
except Exception as e:
print(e)
ret = ""
response = ret
print("=========================================== response ===========================================")
print(f"\033[32m{response}\033[0m")
print("================================================================================================")
return response
def use_custom_prompt(self, dataset):
if dataset is not None and listinstr(['M3GIA'], dataset):
return False
if listinstr(['MCQ', 'VQA'], DATASET_TYPE(dataset)):
return True
elif dataset is not None and listinstr(['HallusionBench'], dataset):
return True
return False
def build_prompt(self, line, dataset=None):
if isinstance(line, int):
line = self.data.iloc[line]
tgt_path = self.dump_image(line, dataset)
system_prompt = ''
question = line['question']
if DATASET_TYPE(dataset) == 'MCQ':
options = {
cand: line[cand]
for cand in string.ascii_uppercase
if cand in line and not pd.isna(line[cand])
}
options_prompt = 'Options:\n'
for key, item in options.items():
options_prompt += f'{key}. {item}\n'
hint = line['hint'] if ('hint' in line and not pd.isna(line['hint'])) else None
prompt = ''
if hint is not None:
prompt += f'Hint: {hint}\n'
prompt += f'Question: {question}\n'
if len(options):
prompt += options_prompt
if 'MMBench' in dataset:
prompt += 'Please select the correct answer from the options above. \n'
else:
system_prompt = self.options_system_prompt + '\nPlease just indicate your choice.'
else:
system_prompt = self.wo_options_system_prompt
if 'MMMU' in dataset: # Corner Case
prompt = system_prompt + '\n' + prompt
system_prompt = ''
elif dataset is not None and listinstr(['HallusionBench'], dataset):
question = line['question'] + ' Yes or No?'
prompt = question
elif dataset is not None and listinstr(['MME'], dataset):
question = line['question'] + ' Yes or No?'
prompt = question
elif dataset is not None and listinstr(['OCRBench'], dataset):
system_prompt = self.vqa_prompt
question = line['question']
prompt = question
elif DATASET_TYPE(dataset) == 'VQA':
if listinstr(['LLaVABench', 'MMLongBench_DOC'], dataset):
system_prompt = ''
prompt = question
elif listinstr(['MMVet'], dataset):
system_prompt = self.detail_system_prompt
prompt = question
elif listinstr(['ChartQA'], dataset):
system_prompt = 'Please answer the question using a single word. '
prompt = question
else:
system_prompt = self.vqa_prompt
prompt = question
msgs = []
if system_prompt:
msgs.append(dict(type='text', value=system_prompt))
if isinstance(tgt_path, list):
msgs.extend([dict(type='image', value=p) for p in tgt_path])
else:
msgs = [dict(type='image', value=tgt_path)]
msgs.append(dict(type='text', value=prompt))
return msgs
|