No more Inference Client bullshit

app.py

@@ -1,11 +1,15 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
 
-"""
-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("mergekit-community/Anti-Qwen2.5-Coder-0.5B-Instruct")
+# Load the model and tokenizer locally
+model_name = "mergekit-community/Anti-Qwen2.5-Coder-0.5B-Instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16)
 
+# Ensure the model runs on GPU if available
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model = model.to(device)
 
 def respond(
     message,
@@ -15,34 +19,36 @@ def respond(
     temperature,
     top_p,
 ):
-    messages = [{"role": "system", "content": system_message}]
+    # Prepare the conversation history for the model
+    messages = [f"System: {system_message}"]
 
     for val in history:
         if val[0]:
-            messages.append({"role": "user", "content": val[0]})
+            messages.append(f"User: {val[0]}")
         if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+            messages.append(f"Assistant: {val[1]}")
 
-    messages.append({"role": "user", "content": message})
+    messages.append(f"User: {message}")
+    context = "\n".join(messages)
 
-    response = ""
+    # Tokenize input
+    input_ids = tokenizer(context, return_tensors="pt", truncation=True, max_length=2048).input_ids.to(device)
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
+    # Generate response
+    output_ids = model.generate(
+        input_ids,
+        max_new_tokens=max_tokens,
         temperature=temperature,
         top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+        do_sample=True,
+        pad_token_id=tokenizer.eos_token_id,
+    )
 
-        response += token
-        yield response
+    # Decode and yield response
+    response = tokenizer.decode(output_ids[0][input_ids.shape[-1]:], skip_special_tokens=True)
+    yield response
 
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
+# Gradio interface setup
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
@@ -59,6 +65,5 @@ demo = gr.ChatInterface(
     ],
 )
 
-
 if __name__ == "__main__":
     demo.launch()