init inference

app.py

@@ -1,29 +1,12 @@
-import os
-os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
-
 import gradio as gr
-from unsloth import FastLanguageModel
-import torch
-
-max_seq_length = 2048  # Choose any! We auto support RoPE Scaling internally!
-dtype = None  # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
-load_in_4bit = True  # Use 4bit quantization to reduce memory usage. Can be False.
+from huggingface_hub import InferenceClient
 
-# Force the model to run on CPU only by setting the device to "cpu"
-device = "cpu"
+"""
+For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
+"""
+client = InferenceClient("Mat17892/llama_lora_G14")
 
-# Load model and tokenizer with the device set to "cpu"
-model, tokenizer = FastLanguageModel.from_pretrained(
-    model_name="llama_lora_model_1",
-    max_seq_length=max_seq_length,
-    dtype=dtype,
-    load_in_4bit=load_in_4bit,
-)
 
-# Move the model to CPU (even if it was initially loaded with GPU support)
-model.to(device)
-
-# Respond function
 def respond(
     message,
     history: list[tuple[str, str]],
@@ -32,44 +15,34 @@ def respond(
     temperature,
     top_p,
 ):
-    # Prepare the system message
     messages = [{"role": "system", "content": system_message}]
 
-    # Add history to the messages
     for val in history:
         if val[0]:
             messages.append({"role": "user", "content": val[0]})
         if val[1]:
             messages.append({"role": "assistant", "content": val[1]})
 
-    # Add the current message from the user
     messages.append({"role": "user", "content": message})
 
-    # Prepare the inputs for the model
-    inputs = tokenizer.apply_chat_template(
-        messages,
-        tokenize=True,
-        add_generation_prompt=True,
-        return_tensors="pt",
-    )
+    response = ""
 
-    # Generate the response using your model on CPU
-    outputs = model.generate(
-        input_ids=inputs["input_ids"].to(device),  # Ensure input is on the CPU
-        max_new_tokens=max_tokens,
+    for message in client.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
         temperature=temperature,
         top_p=top_p,
-        use_cache=True,
-    )
-
-    # Decode the generated output
-    response = tokenizer.batch_decode(outputs, skip_special_tokens=True)
+    ):
+        token = message.choices[0].delta.content
 
-    # Return the response
-    return response[0]
+        response += token
+        yield response
 
 
-# Gradio interface setup
+"""
+For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
+"""
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
@@ -86,6 +59,6 @@ demo = gr.ChatInterface(
     ],
 )
 
-if __name__ == "__main__":
-    demo.launch()
 
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -1,2 +1 @@
 huggingface_hub==0.25.2
-unsloth