Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,36 +1,34 @@
|
|
1 |
# app.py
|
2 |
-
from threading import Thread
|
3 |
import gradio as gr
|
4 |
-
from transformers import AutoTokenizer, AutoModelForCausalLM, TextIteratorStreamer
|
5 |
-
import torch
|
6 |
import spaces
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
|
8 |
-
#
|
9 |
# 1. 加载模型与 Tokenizer
|
10 |
-
#
|
11 |
-
# 如果你的模型需要加速/量化等特殊配置,可在 from_pretrained() 中添加相应参数
|
12 |
-
# 例如 device_map='auto' 或 trust_remote_code=True 等
|
13 |
model_name = "agentica-org/DeepScaleR-1.5B-Preview"
|
14 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
15 |
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto")
|
16 |
|
17 |
-
#
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
# model = model.quantize(4/8) # 如果你的模型和环境支持
|
22 |
|
23 |
-
#
|
24 |
-
# 2.
|
25 |
-
#
|
26 |
def preprocess_messages(history):
|
27 |
"""
|
28 |
-
|
29 |
-
|
30 |
-
User: ...
|
31 |
-
Assistant: ...
|
32 |
-
最后再接上 "Assistant: " 用于提示模型继续生成。
|
33 |
-
你也可以修改为自己需要的对话模板。
|
34 |
"""
|
35 |
prompt = ""
|
36 |
for user_msg, assistant_msg in history:
|
@@ -38,27 +36,33 @@ def preprocess_messages(history):
|
|
38 |
prompt += f"User: {user_msg}\n"
|
39 |
if assistant_msg:
|
40 |
prompt += f"Assistant: {assistant_msg}\n"
|
41 |
-
|
42 |
-
# 继续生成时,让模型再续写 "Assistant:"
|
43 |
prompt += "Assistant: "
|
44 |
return prompt
|
45 |
|
46 |
-
|
47 |
-
#
|
48 |
-
#
|
49 |
-
|
|
|
50 |
def predict(history, max_length, top_p, temperature):
|
51 |
"""
|
52 |
-
|
53 |
-
|
54 |
"""
|
55 |
prompt = preprocess_messages(history)
|
56 |
|
57 |
-
|
58 |
-
|
|
|
|
|
|
|
|
|
|
|
59 |
input_ids = inputs["input_ids"].to(model.device)
|
|
|
60 |
|
61 |
-
#
|
62 |
streamer = TextIteratorStreamer(
|
63 |
tokenizer=tokenizer,
|
64 |
timeout=60,
|
@@ -68,103 +72,98 @@ def predict(history, max_length, top_p, temperature):
|
|
68 |
|
69 |
generate_kwargs = {
|
70 |
"input_ids": input_ids,
|
71 |
-
"
|
|
|
72 |
"do_sample": True,
|
73 |
"top_p": top_p,
|
74 |
"temperature": temperature,
|
75 |
"repetition_penalty": 1.2,
|
76 |
"streamer": streamer,
|
77 |
-
# 如果需要自定义一些特殊 token 或其他参数可在此补充
|
78 |
-
# "eos_token_id": ...
|
79 |
}
|
80 |
|
81 |
-
#
|
82 |
t = Thread(target=model.generate, kwargs=generate_kwargs)
|
83 |
t.start()
|
84 |
|
85 |
-
# history[-1][1]
|
86 |
partial_output = ""
|
87 |
for new_token in streamer:
|
88 |
partial_output += new_token
|
89 |
history[-1][1] = partial_output
|
90 |
yield history
|
91 |
|
92 |
-
|
93 |
-
#
|
94 |
-
#
|
|
|
95 |
def main():
|
96 |
with gr.Blocks() as demo:
|
97 |
-
gr.HTML("<h1 align='center'>DeepScaleR-1.5B
|
98 |
|
99 |
-
# 聊天窗口
|
100 |
chatbot = gr.Chatbot()
|
101 |
|
102 |
with gr.Row():
|
103 |
with gr.Column(scale=2):
|
104 |
user_input = gr.Textbox(
|
105 |
-
show_label=True,
|
106 |
-
placeholder="请输入您的问题...",
|
107 |
label="User Input"
|
108 |
)
|
109 |
submitBtn = gr.Button("Submit")
|
110 |
-
|
111 |
with gr.Column(scale=1):
|
112 |
max_length = gr.Slider(
|
113 |
-
minimum=0,
|
114 |
-
maximum=
|
115 |
-
value=512,
|
116 |
-
step=1,
|
117 |
-
label="Max New Tokens",
|
118 |
interactive=True
|
119 |
)
|
120 |
top_p = gr.Slider(
|
121 |
-
minimum=0,
|
122 |
-
maximum=1,
|
123 |
-
value=0.8,
|
124 |
-
step=0.01,
|
125 |
-
label="Top P",
|
126 |
interactive=True
|
127 |
)
|
128 |
temperature = gr.Slider(
|
129 |
-
minimum=0.
|
130 |
-
maximum=2.0,
|
131 |
-
value=0.7,
|
132 |
-
step=0.01,
|
133 |
-
label="Temperature",
|
134 |
interactive=True
|
135 |
)
|
136 |
|
137 |
-
#
|
138 |
def user(query, history):
|
139 |
return "", history + [[query, ""]]
|
140 |
|
141 |
-
# Submit:
|
142 |
-
# 1) user() -> 新增一条 (user输入,"") 的对话记录
|
143 |
-
# 2) predict() -> 基于更新后的 history 进行生成
|
144 |
submitBtn.click(
|
145 |
-
fn=user,
|
146 |
-
inputs=[user_input, chatbot],
|
147 |
outputs=[user_input, chatbot],
|
148 |
-
queue=False
|
149 |
).then(
|
150 |
-
fn=predict,
|
151 |
-
inputs=[chatbot, max_length, top_p, temperature],
|
152 |
outputs=chatbot
|
153 |
)
|
154 |
|
155 |
-
#
|
156 |
def clear_history():
|
157 |
return [], []
|
158 |
-
emptyBtn.click(
|
159 |
-
fn=clear_history,
|
160 |
-
inputs=[],
|
161 |
-
outputs=[chatbot, user_input],
|
162 |
-
queue=False
|
163 |
-
)
|
164 |
|
165 |
-
|
166 |
-
|
|
|
|
|
167 |
demo.launch()
|
168 |
|
|
|
|
|
|
|
169 |
if __name__ == "__main__":
|
170 |
main()
|
|
|
1 |
# app.py
|
|
|
2 |
import gradio as gr
|
|
|
|
|
3 |
import spaces
|
4 |
+
from threading import Thread
|
5 |
+
import torch
|
6 |
+
|
7 |
+
from transformers import (
|
8 |
+
AutoTokenizer,
|
9 |
+
AutoModelForCausalLM,
|
10 |
+
TextIteratorStreamer,
|
11 |
+
)
|
12 |
|
13 |
+
# ------------------------------
|
14 |
# 1. 加载模型与 Tokenizer
|
15 |
+
# ------------------------------
|
|
|
|
|
16 |
model_name = "agentica-org/DeepScaleR-1.5B-Preview"
|
17 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
18 |
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto")
|
19 |
|
20 |
+
# 如果 tokenizer 没有设置 pad_token_id,则显式指定为 eos_token_id
|
21 |
+
if tokenizer.pad_token_id is None:
|
22 |
+
tokenizer.pad_token_id = tokenizer.eos_token_id
|
23 |
+
|
|
|
24 |
|
25 |
+
# ------------------------------
|
26 |
+
# 2. 对话历史 -> Prompt 格式
|
27 |
+
# ------------------------------
|
28 |
def preprocess_messages(history):
|
29 |
"""
|
30 |
+
将聊天记录拼成一个最简单的 Prompt。
|
31 |
+
你可以自定义更适合该模型的提示格式或特殊 Token。
|
|
|
|
|
|
|
|
|
32 |
"""
|
33 |
prompt = ""
|
34 |
for user_msg, assistant_msg in history:
|
|
|
36 |
prompt += f"User: {user_msg}\n"
|
37 |
if assistant_msg:
|
38 |
prompt += f"Assistant: {assistant_msg}\n"
|
39 |
+
# 继续生成时,提示 "Assistant:"
|
|
|
40 |
prompt += "Assistant: "
|
41 |
return prompt
|
42 |
|
43 |
+
|
44 |
+
# ------------------------------
|
45 |
+
# 3. 预测 / 推理函数
|
46 |
+
# ------------------------------
|
47 |
+
@spaces.GPU() # 让 huggingface spaces 调用 GPU
|
48 |
def predict(history, max_length, top_p, temperature):
|
49 |
"""
|
50 |
+
基于当前的 history 做文本生成。
|
51 |
+
使用 HF 提供的 TextIteratorStreamer 实现流式生成。
|
52 |
"""
|
53 |
prompt = preprocess_messages(history)
|
54 |
|
55 |
+
inputs = tokenizer(
|
56 |
+
prompt,
|
57 |
+
return_tensors="pt",
|
58 |
+
padding=True, # 自动 padding
|
59 |
+
truncation=True, # 超长截断
|
60 |
+
max_length=2048 # 你可根据显存大小或模型上限做调整
|
61 |
+
)
|
62 |
input_ids = inputs["input_ids"].to(model.device)
|
63 |
+
attention_mask = inputs["attention_mask"].to(model.device)
|
64 |
|
65 |
+
# 流式输出器
|
66 |
streamer = TextIteratorStreamer(
|
67 |
tokenizer=tokenizer,
|
68 |
timeout=60,
|
|
|
72 |
|
73 |
generate_kwargs = {
|
74 |
"input_ids": input_ids,
|
75 |
+
"attention_mask": attention_mask,
|
76 |
+
"max_new_tokens": max_length, # 新生成的 token 数
|
77 |
"do_sample": True,
|
78 |
"top_p": top_p,
|
79 |
"temperature": temperature,
|
80 |
"repetition_penalty": 1.2,
|
81 |
"streamer": streamer,
|
|
|
|
|
82 |
}
|
83 |
|
84 |
+
# 在后台线程中执行 generate,主线程循环读取新 token
|
85 |
t = Thread(target=model.generate, kwargs=generate_kwargs)
|
86 |
t.start()
|
87 |
|
88 |
+
# 将最新生成的 token 依次拼接到 history[-1][1]
|
89 |
partial_output = ""
|
90 |
for new_token in streamer:
|
91 |
partial_output += new_token
|
92 |
history[-1][1] = partial_output
|
93 |
yield history
|
94 |
|
95 |
+
|
96 |
+
# ------------------------------
|
97 |
+
# 4. Gradio UI
|
98 |
+
# ------------------------------
|
99 |
def main():
|
100 |
with gr.Blocks() as demo:
|
101 |
+
gr.HTML("<h1 align='center'>DeepScaleR-1.5B Chat Demo</h1>")
|
102 |
|
|
|
103 |
chatbot = gr.Chatbot()
|
104 |
|
105 |
with gr.Row():
|
106 |
with gr.Column(scale=2):
|
107 |
user_input = gr.Textbox(
|
108 |
+
show_label=True,
|
109 |
+
placeholder="请输入您的问题...",
|
110 |
label="User Input"
|
111 |
)
|
112 |
submitBtn = gr.Button("Submit")
|
113 |
+
clearBtn = gr.Button("Clear History")
|
114 |
with gr.Column(scale=1):
|
115 |
max_length = gr.Slider(
|
116 |
+
minimum=0,
|
117 |
+
maximum=1024, # 可根据需要调大/调小
|
118 |
+
value=512,
|
119 |
+
step=1,
|
120 |
+
label="Max New Tokens",
|
121 |
interactive=True
|
122 |
)
|
123 |
top_p = gr.Slider(
|
124 |
+
minimum=0,
|
125 |
+
maximum=1,
|
126 |
+
value=0.8,
|
127 |
+
step=0.01,
|
128 |
+
label="Top P",
|
129 |
interactive=True
|
130 |
)
|
131 |
temperature = gr.Slider(
|
132 |
+
minimum=0.0,
|
133 |
+
maximum=2.0,
|
134 |
+
value=0.7,
|
135 |
+
step=0.01,
|
136 |
+
label="Temperature",
|
137 |
interactive=True
|
138 |
)
|
139 |
|
140 |
+
# 用户点击 Submit 时,先将输入添加到 history,然后再调用 predict 生成
|
141 |
def user(query, history):
|
142 |
return "", history + [[query, ""]]
|
143 |
|
|
|
|
|
|
|
144 |
submitBtn.click(
|
145 |
+
fn=user,
|
146 |
+
inputs=[user_input, chatbot],
|
147 |
outputs=[user_input, chatbot],
|
148 |
+
queue=False # 不排队
|
149 |
).then(
|
150 |
+
fn=predict,
|
151 |
+
inputs=[chatbot, max_length, top_p, temperature],
|
152 |
outputs=chatbot
|
153 |
)
|
154 |
|
155 |
+
# 清空聊天记录
|
156 |
def clear_history():
|
157 |
return [], []
|
|
|
|
|
|
|
|
|
|
|
|
|
158 |
|
159 |
+
clearBtn.click(fn=clear_history, inputs=[], outputs=[chatbot, user_input], queue=False)
|
160 |
+
|
161 |
+
# 可选:启用队列防止并发冲突
|
162 |
+
demo.queue(concurrency_count=1)
|
163 |
demo.launch()
|
164 |
|
165 |
+
# ------------------------------
|
166 |
+
# 入口
|
167 |
+
# ------------------------------
|
168 |
if __name__ == "__main__":
|
169 |
main()
|