-
Notifications
You must be signed in to change notification settings - Fork 4
/
llava15_transformers_engine.py
210 lines (162 loc) · 7.13 KB
/
llava15_transformers_engine.py
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
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
import os
import numpy as np
import argparse
import torch
import gradio as gr
from typing import Any, Iterator
from typing import Iterator, List, Optional, Tuple
import filelock
import glob
import json
import time
from gradio.routes import Request
from gradio.utils import SyncToAsyncIterator, async_iteration
from gradio.helpers import special_args
import anyio
from typing import AsyncGenerator, Callable, Literal, Union, cast
from gradio_client.documentation import document, set_documentation_group
from typing import List, Optional, Union, Dict, Tuple
from tqdm.auto import tqdm
from huggingface_hub import snapshot_download
from gradio.components import Button
from gradio.events import Dependency, EventListenerMethod
from transformers import AutoConfig, AutoTokenizer, PreTrainedTokenizer
import types
import sys
from .base_engine import BaseEngine
from .transformers_engine import TransformersEngine, NewGenerationMixin
from ..configs import (
STREAM_CHECK_MULTIPLE,
STREAM_YIELD_MULTIPLE,
)
from ..configs import (
STREAM_CHECK_MULTIPLE,
STREAM_YIELD_MULTIPLE,
IMAGE_TOKEN,
IMAGE_TOKEN_INTERACTIVE,
IMAGE_TOKEN_LENGTH,
MAX_PACHES,
DTYPE,
DEVICE,
)
CODE_PATH = os.environ.get("CODE_PATH", "")
MODEL_PATH = os.environ.get("MODEL_PATH", "")
# IMAGE_TOKEN = "<image"
# IMAGE_LENGTH = 576
# MAX_PACHES = 1
# ! Still working on it....
# Should only do with
"""
This is a chat between an inquisitive human and an AI assistant. Assume the role of the AI assistant. Read all the images carefully, and respond to the human's questions with informative, helpful, detailed and polite answers. 这是一个好奇的人类和一个人工智能助手之间的对话。假设你扮演这个AI助手的角色。仔细阅读所有的图像,并对人类的问题做出信息丰富、有帮助、详细的和礼貌的回答。
### Human: <image_placeholder>
Describe the cats and what they are doing in detail.
### Assistant:
"""
# prompt = "USER: <image>\nWhat are these?\nASSISTANT:"
# image_file = "http://images.cocodataset.org/val2017/000000039769.jpg"
# conv_llava_llama_2 = Conversation(
# system="You are a helpful language and vision assistant. "
# "You are able to understand the visual content that the user provides, "
# "and assist the user with a variety of tasks using natural language.",
# roles=("USER", "ASSISTANT"),
# version="llama_v2",
# messages=(),
# offset=0,
# sep_style=SeparatorStyle.LLAMA_2,
# sep="<s>",
# sep2="</s>",
# )
LLAVA_CHAT_TEMPLATE = """"""
# "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{{ bos_token }}{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '</s>'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}"
if IMAGE_TOKEN != "<image>":
print(f'WARNING!!!! {IMAGE_TOKEN=} is not <image>, this can lead to problems')
class Llava15TransformersEngine(TransformersEngine):
"""
Llava 1.5 hardcoded
"""
@property
def image_token(self):
return IMAGE_TOKEN
@property
def max_position_embeddings(self) -> int:
return self._model.config.text_config.max_position_embeddings
@property
def tokenizer(self):
return self._tokenizer
@property
def processor(self):
return self._processor
def apply_chat_template(self, conversations, add_generation_prompt: bool, add_special_tokens=False, **kwargs) -> str:
"""
return string convo, add_special_tokens should be added later
"""
prompt = ""
for turn in conversations:
if turn['role'] == 'system':
prompt += turn['content'] + "\n\n"
elif turn['role'] == 'user':
prompt += f"USER: {turn['content']}\n"
elif turn['role'] == 'assistant':
prompt += f"ASSISTANT: {turn['content']}\n"
if add_generation_prompt:
prompt += f"ASSISTANT:"
return prompt
def load_model(self):
import requests
from PIL import Image
from transformers import AutoProcessor, LlavaForConditionalGeneration
self.model_path = model_path = MODEL_PATH
self.torch_dtype = torch.bfloat16 if DTYPE == 'bfloat16' else torch.float16
self.device_map = DEVICE
print(f'Loading model from {model_path} on {self.device_map} with {self.torch_dtype} | LlavaForConditionalGeneration')
self._processor = AutoProcessor.from_pretrained(self.model_path)
self._model = LlavaForConditionalGeneration.from_pretrained(
MODEL_PATH,
torch_dtype=self.torch_dtype, device_map=self.device_map, trust_remote_code=True
).eval()
self._model.sample_old = self._model.sample
# self._model.sample = types.MethodType(NewGenerationMixin.sample_stream, self._model)
self._model._sample = types.MethodType(NewGenerationMixin.sample_stream, self._model)
self._tokenizer = self._processor.tokenizer
print(self._model)
print(f"{self.max_position_embeddings=}")
def get_multimodal_tokens(self, full_prompt, image_paths=None):
num_tokens = len(self.tokenizer.encode(full_prompt))
for image_path in image_paths:
num_tokens += IMAGE_TOKEN_LENGTH * MAX_PACHES
return num_tokens
def generate_yield_string(self, prompt, temperature, max_tokens, stop_strings: Optional[Tuple[str]] = None, **kwargs):
from transformers.generation.utils import GenerationConfig
from PIL import Image
image_paths = kwargs.get("image_paths", None)
image_paths = image_paths or []
images = [Image.open(x) for x in image_paths] if len(image_paths) > 0 else None
with torch.no_grad():
inputs = self.processor(prompt, images, return_tensors='pt')
# inputs = inputs.to("cuda", torch.bfloat16)
inputs = {k: v.to(self.device_map) for k, v in inputs.items() if v is not None}
num_tokens = self.get_multimodal_tokens(prompt, image_paths)
# non-streaming generation
# # ! streaming
generator = self._model.generate(
**inputs,
do_sample=True,
temperature=temperature,
max_new_tokens=max_tokens,
pad_token_id=self.tokenizer.pad_token_id,
)
out_tokens = []
response = None
for index, token in enumerate(generator):
out_tokens.append(token.item())
response = self.tokenizer.decode(out_tokens, skip_special_tokens=True)
if STREAM_YIELD_MULTIPLE > 0:
if index % STREAM_YIELD_MULTIPLE == 0 and index > 0:
yield response, num_tokens
else:
yield response, num_tokens
del generator
if response is not None:
full_text = prompt + response
num_tokens = self.get_multimodal_tokens(full_text, image_paths)
yield response, num_tokens