Update app.py

app.py

@@ -1,61 +1,50 @@
+## Imports
+from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
 import gradio as gr
-from peft import PeftModel, PeftConfig
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from huggingface_hub import login
-import torch
-import re
-import os
-
-model_name = "google/gemma-2b"
-peft_model = "kazuma313/gemma-dokter-ft"
-device_map = "auto"
-hf_token = os.getenv('hftoken')
-login(token=hf_token, add_to_git_credential=True)
-
-from accelerate import disk_offload
-save_dir="gemma-dokter-ft"
-# disk_offload(model=peft_model, offload_dir=save_dir)
-
-# config = PeftConfig.from_pretrained(peft_model)
-
-base_model = AutoModelForCausalLM.from_pretrained(
-    model_name,
-    token=hf_token,
-    low_cpu_mem_usage=True,
-    return_dict=True,
-    torch_dtype=torch.float16,
-    device_map=device_map,
-)
-model = PeftModel.from_pretrained(base_model, peft_model)
-model = model.merge_and_unload()
-
-# Reload tokenizer to save it
-tokenizer = AutoTokenizer.from_pretrained(model_name, 
-                                          token=hf_token,
-                                          trust_remote_code=True)
-tokenizer.pad_token = tokenizer.eos_token
-tokenizer.padding_side = "right"
-
-def echo(message, history, tokens):
-    pattern = r'Step \d+/\d+|^\d+\.\s*'
-    input_ids = tokenizer(message, return_tensors="pt")
-    outputs = model.generate(**input_ids, max_length=tokens)
-    answer = tokenizer.decode(outputs[0], skip_special_tokens=True).split('Answer:')[-1]
-    clean_answer = re.sub(pattern, '', answer)
-
-    return clean_answer
-
-
-demo = gr.ChatInterface(echo, 
-                        examples = [["what is the negative effect of alcohol?"],
-                                    ["i have lack of sleep, what happend if continously do this?"]],
-                        title="dokter Bot",
-                        retry_btn=None,
-                        undo_btn="Delete Previous",
-                        clear_btn="Clear",
-                        additional_inputs=[
-                            gr.Slider(64, 256, value=80)
-                        ], 
-    
-                       )
-demo.launch()
+
+## Download the GGUF model
+model_name = "kazuma313/lora_model_dokter_consultasi_q4_k_m"
+model_file = "lora_model_dokter_consultasi_q4_k_m-unsloth.Q4_K_M.gguf" # this is the specific model file we'll use in this example. It's a 4-bit quant, but other levels of quantization are available in the model repo if preferred
+model_path = hf_hub_download(model_name, filename=model_file)
+
+
+prompt_template="""<|begin_of_text|>Dibawah ini adalah percakapan antara dokter dengan pasiennya yang ingin berkonsultasi terkait kesehatan. Tuliskan jawaban yang tepat dan lengkap sesuai sesuai pertanyaan dari pasien.<|end_of_text|>
+
+### Pertanyaan:
+{ask}
+
+### Jawaban:
+"""
+
+def output_inference(tanya, history):
+    prompt = prompt_template.format(ask=tanya)
+
+    output = llm(
+      prompt,
+      stop=["<|end_of_text|>"],
+      max_tokens=512,
+      temperature=0.3,
+      top_p=0.95,
+      top_k=40,
+      min_p=0.05,
+      typical_p=1.0,
+      stream=False,
+
+    )
+    return output['choices'][0]['text']
+
+
+gr.ChatInterface(
+    output_inference,
+    chatbot=gr.Chatbot(height=300),
+    textbox=gr.Textbox(placeholder="Tanya saya kesehatan anda", container=False, scale=7),
+    title="Konsultasi dokter",
+    description="Tanya saja semua keluhan mu",
+    theme="soft",
+    examples=["apa saja tips hidup sehat?", "apa penyebab dari minum alkohol berlebihan?", "apa yang terjadi jika pola tidur tidak teratur?", "berapa hasil dari 10 + 5?"],
+    cache_examples=True,
+    retry_btn=None,
+    undo_btn="Delete Previous",
+    clear_btn="Clear",
+).launch()