Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -11,19 +11,21 @@ import gradio as gr
|
|
11 |
login(os.getenv("HUGGINGFACEHUB_API_TOKEN"))
|
12 |
|
13 |
# Token authentication for requests
|
14 |
-
API_TOKEN = os.getenv("HF_API_TOKEN")
|
15 |
|
16 |
# Set up model loading and pipeline
|
17 |
torch_dtype = torch.bfloat16 if torch.cuda.is_available() else torch.float32
|
18 |
os.environ['HF_HOME'] = '/tmp/cache'
|
19 |
-
|
20 |
model_name = "cerebras/btlm-3b-8k-chat"
|
21 |
-
|
|
|
|
|
22 |
model = AutoModelForCausalLM.from_pretrained(
|
23 |
model_name,
|
24 |
torch_dtype=torch_dtype,
|
25 |
device_map="auto",
|
26 |
-
trust_remote_code=True
|
|
|
27 |
)
|
28 |
|
29 |
generator = pipeline(
|
@@ -36,7 +38,6 @@ generator = pipeline(
|
|
36 |
trust_remote_code=True
|
37 |
)
|
38 |
|
39 |
-
# Flask app
|
40 |
app = Flask(__name__)
|
41 |
|
42 |
@app.route("/")
|
@@ -45,7 +46,6 @@ def home():
|
|
45 |
|
46 |
@app.route("/v1/chat/completions", methods=["POST"])
|
47 |
def chat():
|
48 |
-
# Token auth: require Bearer token
|
49 |
auth_header = request.headers.get("Authorization", "")
|
50 |
if not auth_header.startswith("Bearer ") or auth_header.split(" ")[1] != API_TOKEN:
|
51 |
return jsonify({"error": "Unauthorized"}), 401
|
@@ -56,7 +56,6 @@ def chat():
|
|
56 |
temperature = data.get("temperature", 0.7)
|
57 |
stream = data.get("stream", False)
|
58 |
|
59 |
-
# Build the prompt from chat history
|
60 |
prompt = ""
|
61 |
for msg in messages:
|
62 |
role = msg.get("role", "user").capitalize()
|
@@ -64,7 +63,6 @@ def chat():
|
|
64 |
prompt += f"{role}: {content}\n"
|
65 |
prompt += "Assistant:"
|
66 |
|
67 |
-
# If stream = True, stream response like OpenAI
|
68 |
if stream:
|
69 |
def generate_stream():
|
70 |
output = generator(
|
@@ -97,7 +95,6 @@ def chat():
|
|
97 |
|
98 |
return Response(generate_stream(), content_type="text/event-stream")
|
99 |
|
100 |
-
# Non-streamed response
|
101 |
output = generator(
|
102 |
prompt,
|
103 |
max_new_tokens=max_tokens,
|
@@ -109,24 +106,21 @@ def chat():
|
|
109 |
reply = output[0]["generated_text"].replace(prompt, "").strip()
|
110 |
|
111 |
return jsonify({
|
112 |
-
"choices": [
|
113 |
-
{
|
114 |
-
"
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
}
|
121 |
-
]
|
122 |
})
|
123 |
|
124 |
-
# Optional Gradio frontend to keep
|
125 |
with gr.Blocks() as demo:
|
126 |
gr.Markdown("### LLM backend is running and ready for API calls.")
|
127 |
|
128 |
demo.launch()
|
129 |
|
130 |
if __name__ == "__main__":
|
131 |
-
# Listen on port 8080 as required by HF Spaces
|
132 |
app.run(host="0.0.0.0", port=8080)
|
|
|
11 |
login(os.getenv("HUGGINGFACEHUB_API_TOKEN"))
|
12 |
|
13 |
# Token authentication for requests
|
14 |
+
API_TOKEN = os.getenv("HF_API_TOKEN")
|
15 |
|
16 |
# Set up model loading and pipeline
|
17 |
torch_dtype = torch.bfloat16 if torch.cuda.is_available() else torch.float32
|
18 |
os.environ['HF_HOME'] = '/tmp/cache'
|
|
|
19 |
model_name = "cerebras/btlm-3b-8k-chat"
|
20 |
+
revision = "main" # Pin to stable revision
|
21 |
+
|
22 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True, revision=revision)
|
23 |
model = AutoModelForCausalLM.from_pretrained(
|
24 |
model_name,
|
25 |
torch_dtype=torch_dtype,
|
26 |
device_map="auto",
|
27 |
+
trust_remote_code=True,
|
28 |
+
revision=revision
|
29 |
)
|
30 |
|
31 |
generator = pipeline(
|
|
|
38 |
trust_remote_code=True
|
39 |
)
|
40 |
|
|
|
41 |
app = Flask(__name__)
|
42 |
|
43 |
@app.route("/")
|
|
|
46 |
|
47 |
@app.route("/v1/chat/completions", methods=["POST"])
|
48 |
def chat():
|
|
|
49 |
auth_header = request.headers.get("Authorization", "")
|
50 |
if not auth_header.startswith("Bearer ") or auth_header.split(" ")[1] != API_TOKEN:
|
51 |
return jsonify({"error": "Unauthorized"}), 401
|
|
|
56 |
temperature = data.get("temperature", 0.7)
|
57 |
stream = data.get("stream", False)
|
58 |
|
|
|
59 |
prompt = ""
|
60 |
for msg in messages:
|
61 |
role = msg.get("role", "user").capitalize()
|
|
|
63 |
prompt += f"{role}: {content}\n"
|
64 |
prompt += "Assistant:"
|
65 |
|
|
|
66 |
if stream:
|
67 |
def generate_stream():
|
68 |
output = generator(
|
|
|
95 |
|
96 |
return Response(generate_stream(), content_type="text/event-stream")
|
97 |
|
|
|
98 |
output = generator(
|
99 |
prompt,
|
100 |
max_new_tokens=max_tokens,
|
|
|
106 |
reply = output[0]["generated_text"].replace(prompt, "").strip()
|
107 |
|
108 |
return jsonify({
|
109 |
+
"choices": [{
|
110 |
+
"message": {
|
111 |
+
"role": "assistant",
|
112 |
+
"content": reply
|
113 |
+
},
|
114 |
+
"finish_reason": "stop",
|
115 |
+
"index": 0
|
116 |
+
}]
|
|
|
|
|
117 |
})
|
118 |
|
119 |
+
# Optional Gradio frontend to keep Space alive
|
120 |
with gr.Blocks() as demo:
|
121 |
gr.Markdown("### LLM backend is running and ready for API calls.")
|
122 |
|
123 |
demo.launch()
|
124 |
|
125 |
if __name__ == "__main__":
|
|
|
126 |
app.run(host="0.0.0.0", port=8080)
|