Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,71 +1,57 @@
|
|
1 |
-
from flask import Flask, request, jsonify
|
2 |
-
import
|
3 |
-
import io
|
4 |
-
import random
|
5 |
-
import os
|
6 |
-
from PIL import Image
|
7 |
-
from datetime import datetime
|
8 |
|
9 |
app = Flask(__name__)
|
|
|
10 |
|
11 |
-
|
12 |
-
|
13 |
-
headers = {"Authorization": f"Bearer {API_TOKEN}"}
|
14 |
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
-
|
19 |
-
|
20 |
-
"inputs": prompt,
|
21 |
-
"is_negative": is_negative,
|
22 |
-
"steps": steps,
|
23 |
-
"cfg_scale": cfg_scale,
|
24 |
-
"seed": seed if seed is not None else random.randint(-1, 2147483647)
|
25 |
-
}
|
26 |
|
27 |
-
|
28 |
-
|
29 |
-
return image
|
30 |
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
filename = f"{prompt}-{random.randint(1, 100000)}-{timestamp}.png"
|
35 |
-
filepath = os.path.join(TEMP_DIR, filename)
|
36 |
-
image.save(filepath, format='PNG')
|
37 |
-
return filename
|
38 |
|
39 |
|
40 |
@app.route("/generate", methods=["POST"])
|
41 |
-
def
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
response = {
|
57 |
-
"success": False,
|
58 |
-
"error": str(e)
|
59 |
-
}
|
60 |
-
|
61 |
-
return jsonify(response)
|
62 |
-
|
63 |
-
|
64 |
-
@app.route("/temp/<filename>")
|
65 |
-
def show_image(filename):
|
66 |
-
return send_from_directory(TEMP_DIR, filename)
|
67 |
-
|
68 |
-
|
69 |
if __name__ == "__main__":
|
70 |
if not os.path.exists(TEMP_DIR):
|
71 |
os.makedirs(TEMP_DIR)
|
|
|
1 |
+
from flask import Flask, request, jsonify
|
2 |
+
from huggingface_hub import InferenceClient
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
app = Flask(__name__)
|
5 |
+
app.config["DEBUG"] = True # Enable for debugging
|
6 |
|
7 |
+
# Load model client
|
8 |
+
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
|
|
|
9 |
|
10 |
+
# Function for text generation with enhanced prompt formatting
|
11 |
+
def generate(
|
12 |
+
prompt, history, system_prompt, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0
|
13 |
+
):
|
14 |
+
temperature = float(temperature)
|
15 |
+
if temperature < 1e-2:
|
16 |
+
temperature = 1e-2
|
17 |
+
top_p = float(top_p)
|
18 |
|
19 |
+
generate_kwargs = dict(
|
20 |
+
temperature=temperature,
|
21 |
+
max_new_tokens=max_new_tokens,
|
22 |
+
top_p=top_p,
|
23 |
+
repetition_penalty=repetition_penalty,
|
24 |
+
do_sample=True,
|
25 |
+
seed=42,
|
26 |
+
)
|
27 |
|
28 |
+
# Enhanced prompt formatting for better context
|
29 |
+
formatted_prompt = f"{system_prompt}\n" + ''.join(f"{user_prompt} ||| {bot_response}\n" for user_prompt, bot_response in history) + f"\n{prompt}"
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
|
31 |
+
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
32 |
+
output = ""
|
|
|
33 |
|
34 |
+
for response in stream:
|
35 |
+
output += response.token.text
|
36 |
+
return output
|
|
|
|
|
|
|
|
|
37 |
|
38 |
|
39 |
@app.route("/generate", methods=["POST"])
|
40 |
+
def generate_text():
|
41 |
+
data = request.json
|
42 |
+
prompt = data.get("prompt")
|
43 |
+
history = data.get("history", [])
|
44 |
+
system_prompt = data.get("system_prompt")
|
45 |
+
temperature = data.get("temperature", 0.9)
|
46 |
+
max_new_tokens = data.get("max_new_tokens", 256)
|
47 |
+
top_p = data.get("top_p", 0.95)
|
48 |
+
repetition_penalty = data.get("repetition_penalty", 1.0)
|
49 |
+
|
50 |
+
response = generate(
|
51 |
+
prompt, history, system_prompt, temperature, max_new_tokens, top_p, repetition_penalty
|
52 |
+
)
|
53 |
+
|
54 |
+
return jsonify({"response": response})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
if __name__ == "__main__":
|
56 |
if not os.path.exists(TEMP_DIR):
|
57 |
os.makedirs(TEMP_DIR)
|