-
Notifications
You must be signed in to change notification settings - Fork 4
/
app.py
311 lines (236 loc) · 10.5 KB
/
app.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
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
import gradio as gr
import os, gc, copy, torch
os.environ["RWKV_CUDA_ON"] = '1' # if '1' then use CUDA kern0el for seq mode (much faster)
from datetime import datetime
import yaml
import time
from rwkv.model import RWKV
from rwkv.utils import PIPELINE, PIPELINE_ARGS
from easynmt import EasyNMT
#from minirwkv4 import blipcaption
from minirwkv4 import vitgptcaption
#from minirwkv4 import vitvqa
from minirwkv4 import blipvqa
translatemodel = EasyNMT('opus-mt')
ctx_limit = 2048 #3B模型最大值为4096，7B为8192
title = "MiniRWKV-4：基于RWKV-4 + BLIP/VIT-GPT的多模态图文对话大模型"
description = """<h3>MiniRWKV-4的例子，上传你的图片并开始聊天!</h3>"""
article = """<p>开源地址：<a href='https://github.com/StarRing2022/MiniRWKV-4'>StarRing2022/MiniRWKV-4</a></p>"""
def readcog(path):
with open(path, 'r',encoding='UTF-8') as file:
data = file.read()
result = yaml.safe_load(data)
return result
LMyamlres = readcog("./config/minirwkv4.yaml")
#model_path = LMyamlres['model-language']['3Bpath']
model_path = LMyamlres['model-language']['7Bpath']
model = RWKV(model=model_path, strategy='cuda fp16i8 *8 -> cuda fp16') #加载模型
tokenizer_path = LMyamlres['model-language']['tokenizer']
pipeline = PIPELINE(model, tokenizer_path) #加载tokenizer
def upload_file(chatbot, upload_btn):
chat_history = chatbot
file = upload_btn
lipres=""
#vcaption = blipcaption.get_blipcap(file.name)
vcaption = vitgptcaption.get_vitgptcap(file.name)
lipres = translatemodel.translate(vcaption, target_lang='zh')
lipres = str(lipres)
time.sleep(1)
rwkvres = get_answer(botmode = 0,message = lipres)
#print(rwkvres)
chatres = str(lipres+"。"+rwkvres)
#chat_history = chat_history + [((file.name,), lipres)]
chat_history = chat_history + [((file.name,), chatres)]
return chat_history
def reset_chat(input_txt,chatbot):
return None, None
def dispic(upload_btn):
try:
if not upload_btn:
return upload_btn
else:
#print(upload_btn.name)
upload_btn.name=""
upload_btn=None
except:
pass
return upload_btn
def generate_prompt(prompt,cardiogenic_prompt=None,operability_prompt=None,exogenous_prompt=None):
promptalter = ""
if cardiogenic_prompt:
promptalter = promptalter + cardiogenic_prompt
if operability_prompt:
promptalter = promptalter + operability_prompt
if exogenous_prompt:
promptalter = promptalter + exogenous_prompt
promptalter = promptalter + prompt
#print(promptalter)
return f"Human: {promptalter} \nAssistant:"
def get_answer(botmode,message,token_count=500,temperature=0.8,top_p=0.7,presencePenalty=0.1,countPenalty=0.1):
args = PIPELINE_ARGS(temperature = max(0.2, float(temperature)), top_p = float(top_p),
alpha_frequency = float(presencePenalty),
alpha_presence = float(countPenalty),
token_ban = [], # ban the generation of some tokens
token_stop = [0]) # stop generation whenever you see any token here
message = message.strip().replace('\r\n','\n')
#prompt种类：cardiogenic,operability,exogenous
CPyamlres = readcog("./prompts/cardiogenic.yaml")
cardiogenic_prompt=CPyamlres['promptwords']['nature']
#print(cardiogenic_prompt) #心源性
OPyamlres = readcog("./prompts/operability.yaml")
operability_prompt=OPyamlres['promptwords']['task']
#print(operability_prompt) #操作性
EXyamlres = readcog("./prompts/exogenous.yaml")
exogenous_prompt=EXyamlres['promptwords']['instruction'] #外因性
#print(exogenous_prompt)
# 判断提示模式
if(botmode==1):
# 提示模式1
ctx = generate_prompt(message,cardiogenic_prompt=cardiogenic_prompt).strip()
#print(ctx)
elif(botmode==2):
# 提示模式2
ctx = generate_prompt(message,cardiogenic_prompt=cardiogenic_prompt,operability_prompt=operability_prompt).strip()
#print(ctx)
elif(botmode==3):
# 提示模式3
ctx = generate_prompt(message,cardiogenic_prompt=cardiogenic_prompt,operability_prompt=operability_prompt,exogenous_prompt=exogenous_prompt).strip()
#print(ctx)
elif(botmode==0):
# 不使用提示
ctx = generate_prompt(message).strip()
#print(ctx)
all_tokens = []
out_last = 0
out_str = ''
occurrence = {}
state = None
for i in range(int(token_count)):
out, state = model.forward(pipeline.encode(ctx)[-ctx_limit:] if i == 0 else [token], state)
for n in occurrence:
out[n] -= (args.alpha_presence + occurrence[n] * args.alpha_frequency)
token = pipeline.sample_logits(out, temperature=args.temperature, top_p=args.top_p)
if token in args.token_stop:
break
all_tokens += [token]
if token not in occurrence:
occurrence[token] = 1
else:
occurrence[token] += 1
tmp = pipeline.decode(all_tokens[out_last:])
if '\ufffd' not in tmp:
out_str += tmp
out_last = i + 1
del out
del state
gc.collect()
torch.cuda.empty_cache()
answer = out_str.strip()
return answer
def gen_response(
input_txt,
chatbot,
upload_btn,
temperature=0.9,
top_p=0.7,
presencePenalty = 0.1,
countPenalty = 0.1,
):
usrmsg = input_txt
chat_history = chatbot
response = ""
#判断是否结合图片进行对话
BotMode = 1 # 1为只加载心源性提示；2为加载心源性提示和操作性提示；3为三种提示都加载
try:
if not upload_btn:
BotMode = 1
response = get_answer(botmode = BotMode,message=usrmsg,token_count=1024,temperature=temperature,top_p=top_p,presencePenalty=presencePenalty,countPenalty=countPenalty)
else:
BotMode = 3
#print(upload_btn.name)
file = upload_btn
imgquery = translatemodel.translate(input_txt, target_lang='en')
#print(imgquery)
#vqares = vitvqa.get_vqares(file.name,imgquery)
vqares = blipvqa.get_bqares(file.name,imgquery)
#print(vqares)
if vqares.isdigit():
pass
else:
vqares = translatemodel.translate(vqares, target_lang='zh')
#print(vqares)
msgvqa = f"已知问答题，对于问题：{usrmsg}，问题的答案是：{vqares}。请再次回答：{usrmsg}"
#二阶段推理
response_step1 = get_answer(botmode = 0,message=msgvqa,token_count=1024,temperature=temperature,top_p=top_p,presencePenalty=presencePenalty,countPenalty=countPenalty)
response_step2 = get_answer(botmode = 3,message=response_step1,token_count=1024,temperature=temperature,top_p=top_p,presencePenalty=presencePenalty,countPenalty=countPenalty)
response = response_step1+"\n"+response_step2
except:
BotMode = 2
response = get_answer(botmode = BotMode,message=usrmsg,token_count=1024,temperature=temperature,top_p=top_p,presencePenalty=presencePenalty,countPenalty=countPenalty)
#print(response)
chat_history.append((usrmsg, response))
return "",chat_history
with gr.Blocks(title = "MiniRWKV-4 Demo") as demo:
gr.HTML(f"<div style=\"text-align: center;\">\n<h1>🐦{title}</h1>\n</div>")
gr.Markdown(description)
gr.Markdown(article)
with gr.Row():
chatbot = gr.Chatbot(value=[], label = "MiniRWKV-4",elem_id="chatbot").style(height=500)
with gr.Row():
with gr.Column(scale=0.85):
input_txt = gr.Textbox(show_label=False,placeholder="输入内容，或上传一张图片")
with gr.Column(scale=0.15, min_width=0):
upload_btn = gr.UploadButton("📁", file_types=["image"])
disload_btn = gr.Button("清除图片")
with gr.Row():
temperature = gr.Slider(0.2, 2.0, label="创造力", step=0.1, value=1.2)
top_p = gr.Slider(0.0, 1.0, label="注意力参数", step=0.05, value=0.5)
presence_penalty = gr.Slider(0.0, 1.0, label="在场惩罚参数", step=0.1, value=0.4)
count_penalty = gr.Slider(0.0, 1.0, label="计数惩罚参数", step=0.1, value=0.4)
submit_btn = gr.Button("提交", variant="primary")
clear_btn = gr.Button("清空", variant="secondary")
input_txt.submit(gen_response, [input_txt, chatbot, upload_btn, temperature, top_p, presence_penalty, count_penalty], [input_txt, chatbot])
submit_btn.click(gen_response, [input_txt, chatbot, upload_btn, temperature, top_p, presence_penalty, count_penalty], [input_txt, chatbot])
clear_btn.click(reset_chat, [input_txt,chatbot], [input_txt,chatbot])
upload_btn.upload(upload_file, [chatbot, upload_btn], [chatbot])
disload_btn.click(dispic,[upload_btn],[upload_btn])
demo.queue(concurrency_count=1, max_size=10)
demo.launch(share=False)
# if __name__ == "__main__":
# token_count = 500
# args = PIPELINE_ARGS(temperature = max(0.2, float(0.8)), top_p = float(0.7),
# alpha_frequency = 0.1,
# alpha_presence = 0.1,
# token_ban = [], # ban the generation of some tokens
# token_stop = [0]) # stop generation whenever you see any token here
# message = "你好"
# message = message.strip().replace('\r\n','\n')
# ctx = generate_prompt(message).strip()
# #print(ctx)
# all_tokens = []
# out_last = 0
# out_str = ''
# occurrence = {}
# state = None
# for i in range(int(token_count)):
# out, state = model.forward(pipeline.encode(ctx)[-ctx_limit:] if i == 0 else [token], state)
# for n in occurrence:
# out[n] -= (args.alpha_presence + occurrence[n] * args.alpha_frequency)
# token = pipeline.sample_logits(out, temperature=args.temperature, top_p=args.top_p)
# if token in args.token_stop:
# break
# all_tokens += [token]
# if token not in occurrence:
# occurrence[token] = 1
# else:
# occurrence[token] += 1
# tmp = pipeline.decode(all_tokens[out_last:])
# if '\ufffd' not in tmp:
# out_str += tmp
# out_last = i + 1
# del out
# del state
# gc.collect()
# torch.cuda.empty_cache()
# answer = out_str.strip()
# print(answer)