Upload 2 files

app.py

@@ -0,0 +1,59 @@
+import torch
+import gradio as gr
+from peft import PeftModel, PeftConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
+from conversation import Conversation
+
+MODEL_NAME = "warleagle/medical_chat_saiga"
+
+config = PeftConfig.from_pretrained(MODEL_NAME)
+model = AutoModelForCausalLM.from_pretrained(
+    config.base_model_name_or_path,
+    load_in_8bit=True,
+    torch_dtype=torch.float16,
+    device_map="auto"
+)
+
+model = PeftModel.from_pretrained(
+    model,
+    MODEL_NAME,
+    torch_dtype=torch.float16
+)
+model.eval()
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=False)
+generation_config = GenerationConfig.from_pretrained(MODEL_NAME)
+generation_config.max_new_tokens = 70
+
+
+def generate(model, tokenizer, prompt, generation_config):
+    data = tokenizer(prompt, return_tensors="pt", add_special_tokens=False)
+    data = {k: v.to(model.device) for k, v in data.items()}
+    output_ids = model.generate(
+        **data,
+        generation_config=generation_config
+    )[0]
+    output_ids = output_ids[len(data["input_ids"][0]):]
+    output = tokenizer.decode(output_ids, skip_special_tokens=True)
+    return output.strip()
+
+def predict(input_data, temp):
+    generation_config.temperature = temp
+    
+    conversation = Conversation()
+    conversation.add_user_message(input_data)
+    prompt = conversation.get_prompt(tokenizer)
+
+    output = generate(model, tokenizer, prompt, generation_config)
+    return output
+    
+io = gr.Interface(predict,
+                  inputs=[gr.Text(value="Как записаться к стоматологу?"),
+                          gr.Slider(minimum=0.01,
+                                    maximum=1,
+                                    value=0.3,
+                                    step=0.1)],
+                  outputs=[gr.Text()])
+
+if __name__ == 'main':
+    io.launch()

conversation.py

@@ -0,0 +1,43 @@
+class Conversation:
+    def __init__(
+        self,
+    ):
+        
+        self.message_template = "<s>{role}\n{content}</s>"
+        self.response_template = "<s>bot\n"
+        self.messages = [{
+            "role": "system",
+            "content": "Предложи ответ оператора технической поддержки на вопрос пользователя из чата."
+        }]
+
+    def add_user_message(self, message):
+        self.messages.append({
+            "role": "user",
+            "content": message
+        })
+
+    def add_bot_message(self, message):
+        self.messages.append({
+            "role": "bot",
+            "content": message
+        })
+
+    def get_prompt(self):
+        final_text = ""
+        for message in self.messages:
+            message_text = self.message_template.format(**message)
+            final_text += message_text
+        final_text += "<s>bot\n"
+        return final_text.strip()
+
+
+def generate(model, tokenizer, prompt, generation_config):
+    data = tokenizer(prompt, return_tensors="pt", add_special_tokens=False)
+    data = {k: v.to(model.device) for k, v in data.items()}
+    output_ids = model.generate(
+        **data,
+        generation_config=generation_config
+    )[0]
+    output_ids = output_ids[len(data["input_ids"][0]):]
+    output = tokenizer.decode(output_ids, skip_special_tokens=True)
+    return output.strip()