Spaces:
Running
Running
Create app.py with ctransformers
Browse files
app.py
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from ctransformers import AutoModelForCausalLM
|
| 3 |
+
|
| 4 |
+
MODEL_REPO = "bartowski/Hermes-3-Llama-3.1-8B-GGUF"
|
| 5 |
+
MODEL_FILE = "Hermes-3-Llama-3.1-8B-Q4_K_M.gguf"
|
| 6 |
+
|
| 7 |
+
print("Loading model...")
|
| 8 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 9 |
+
MODEL_REPO,
|
| 10 |
+
model_file=MODEL_FILE,
|
| 11 |
+
hf_model=MODEL_REPO,
|
| 12 |
+
gpu_layers=0,
|
| 13 |
+
context_length=4096,
|
| 14 |
+
)
|
| 15 |
+
|
| 16 |
+
def respond(message, history):
|
| 17 |
+
prompt = ""
|
| 18 |
+
for user_msg, bot_msg in history:
|
| 19 |
+
prompt += f"<|im_start|>user\n{user_msg}\n<|im_end|>\n"
|
| 20 |
+
prompt += f"<|im_start|>assistant\n{bot_msg}\n<|im_end|>\n"
|
| 21 |
+
prompt += f"<|im_start|>user\n{message}\n<|im_end|>\n<|im_start|>assistant\n"
|
| 22 |
+
output = model(prompt, max_new_tokens=512, temperature=0.7, top_p=0.9)
|
| 23 |
+
return output
|
| 24 |
+
|
| 25 |
+
iface = gr.ChatInterface(respond)
|
| 26 |
+
|
| 27 |
+
if __name__ == "__main__":
|
| 28 |
+
iface.launch()
|