Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -13,25 +13,19 @@ print(f"Model downloaded to: {model_path}")
|
|
13 |
|
14 |
# --- 2. Model Loading (Optimized for HF Space CPU) ---
|
15 |
print("===== Loading model... =====")
|
16 |
-
|
17 |
-
# This will correctly get the number of CPUs allocated to your Space (often 2 on free tier)
|
18 |
n_threads = os.cpu_count()
|
19 |
-
|
20 |
llm = Llama(
|
21 |
model_path=model_path,
|
22 |
n_ctx=2048,
|
23 |
n_threads=n_threads,
|
24 |
-
# --- !! KEY CHANGE FOR HUGGING FACE SPACES !! ---
|
25 |
-
# Set n_gpu_layers to 0. Free Spaces run on CPUs.
|
26 |
-
# Setting this to a non-zero value will cause errors without a GPU upgrade.
|
27 |
n_gpu_layers=0
|
28 |
)
|
|
|
29 |
print(f"Model loaded for CPU execution with {n_threads} threads.")
|
30 |
|
31 |
|
32 |
-
# --- 3. Chat Function with Streaming (
|
33 |
def chat(message, history):
|
34 |
-
# This function remains the same, as streaming is even MORE important on a slower CPU!
|
35 |
history_prompt = ""
|
36 |
for user_msg, assistant_msg in history:
|
37 |
history_prompt += f"### User:\n{user_msg}\n\n### Assistant:\n{assistant_msg}\n\n"
|
@@ -57,7 +51,9 @@ You are Dolphin 3.0, a helpful and friendly AI assistant.
|
|
57 |
partial_message += token
|
58 |
yield partial_message
|
59 |
|
60 |
-
# --- 4. The Enhanced Chatbot UI (
|
|
|
|
|
61 |
iface = gr.ChatInterface(
|
62 |
fn=chat,
|
63 |
title="🐬 Dolphin 3.0 on Hugging Face Spaces",
|
@@ -67,18 +63,9 @@ iface = gr.ChatInterface(
|
|
67 |
theme="soft",
|
68 |
examples=[["Hello!"], ["Write a short poem about the stars."], ["What is the capital of India?"]],
|
69 |
cache_examples=False,
|
70 |
-
|
71 |
-
undo_btn="Delete Previous",
|
72 |
-
clear_btn="Clear Chat",
|
73 |
)
|
74 |
|
75 |
-
# --- Pro-Tip: Create a requirements.txt file for your Space! ---
|
76 |
-
# Your Space needs to know what libraries to install. Create a file
|
77 |
-
# named `requirements.txt` in your repository with the following lines:
|
78 |
-
#
|
79 |
-
# gradio
|
80 |
-
# llama-cpp-python
|
81 |
-
# huggingface_hub
|
82 |
|
83 |
if __name__ == "__main__":
|
84 |
iface.launch()
|
|
|
13 |
|
14 |
# --- 2. Model Loading (Optimized for HF Space CPU) ---
|
15 |
print("===== Loading model... =====")
|
|
|
|
|
16 |
n_threads = os.cpu_count()
|
|
|
17 |
llm = Llama(
|
18 |
model_path=model_path,
|
19 |
n_ctx=2048,
|
20 |
n_threads=n_threads,
|
|
|
|
|
|
|
21 |
n_gpu_layers=0
|
22 |
)
|
23 |
+
# I see from your logs it found 16 threads! That's a powerful CPU Space.
|
24 |
print(f"Model loaded for CPU execution with {n_threads} threads.")
|
25 |
|
26 |
|
27 |
+
# --- 3. Chat Function with Streaming (No changes) ---
|
28 |
def chat(message, history):
|
|
|
29 |
history_prompt = ""
|
30 |
for user_msg, assistant_msg in history:
|
31 |
history_prompt += f"### User:\n{user_msg}\n\n### Assistant:\n{assistant_msg}\n\n"
|
|
|
51 |
partial_message += token
|
52 |
yield partial_message
|
53 |
|
54 |
+
# --- 4. The Enhanced Chatbot UI (Corrected!) ---
|
55 |
+
# We've removed the 'retry_btn' and 'undo_btn' arguments to match the
|
56 |
+
# version of Gradio running on Hugging Face Spaces.
|
57 |
iface = gr.ChatInterface(
|
58 |
fn=chat,
|
59 |
title="🐬 Dolphin 3.0 on Hugging Face Spaces",
|
|
|
63 |
theme="soft",
|
64 |
examples=[["Hello!"], ["Write a short poem about the stars."], ["What is the capital of India?"]],
|
65 |
cache_examples=False,
|
66 |
+
clear_btn="Clear Chat", # This one is universally loved and supported
|
|
|
|
|
67 |
)
|
68 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
if __name__ == "__main__":
|
71 |
iface.launch()
|