Update README.md

README.md

@@ -16,81 +16,45 @@ GITHUB_ACTIONS=true pip install auto-gptq
 Пример кода для использования модели в генерации ответа:
 
 ```python
-from transformers import AutoTokenizer, TextGenerationPipeline
-from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
-
-
-class Conversation:
-    def __init__(
-        self,
-        message_template=DEFAULT_MESSAGE_TEMPLATE,
-        system_prompt=DEFAULT_SYSTEM_PROMPT,
-        start_token_id=1,
-        bot_token_id=9225
-    ):
-        self.message_template = message_template
-        self.start_token_id = start_token_id
-        self.bot_token_id = bot_token_id
-        self.messages = [{
-            "role": "system",
-            "content": system_prompt
-        }]
-
-    def get_start_token_id(self):
-        return self.start_token_id
-
-    def get_bot_token_id(self):
-        return self.bot_token_id
-
-    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, tokenizer):
-        final_text = ""
-        for message in self.messages:
-            message_text = self.message_template.format(**message)
-            final_text += message_text
-        final_text += tokenizer.decode([self.start_token_id, self.bot_token_id])
-        return final_text.strip()
-
-
-def generate(model, tokenizer, prompt, generation_config):
-    data = tokenizer(prompt, return_tensors="pt")
-    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()
-
-
-MODEL_NAME = "gurgutan/saiga2-13b-4bit"
-DEFAULT_MESSAGE_TEMPLATE = "<s>{role}\n{content}</s>\n"
-DEFAULT_SYSTEM_PROMPT = "Ты — Сайга, русскоязычный автоматический ассистент. Ты разговариваешь с людьми и помогаешь им."
-
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=True)
-model = AutoGPTQForCausalLM.from_quantized(MODEL_NAME, device="cuda:0", use_safetensors=True, use_triton=False)
-generation_config = GenerationConfig.from_pretrained(MODEL_NAME)
+from transformers import AutoTokenizer
+from auto_gptq import AutoGPTQForCausalLM
+
+device = "cuda:0"
+quantized_model_dir = "saiga2-13b-4bit"
+tokenizer = AutoTokenizer.from_pretrained(quantized_model_dir, use_fast=True)
+model = AutoGPTQForCausalLM.from_quantized(quantized_model_dir, device=device)
+
+def generate_answer(model, tokenizer, request: str):
+    s = f"system\n{config.system_prompt}</s>\n" + \
+        f"<s>user\n{request}</s>\n" + \
+        f"<s>bot\n"
+    request_tokens = tokenizer(s, return_tensors="pt")    
+    del request_tokens['token_type_ids']
+    del request_tokens['attention_mask']
+    request_tokens = request_tokens.to(model.device)
+    answer_tokens = model.generate(**request_tokens,
+                    num_beams=4,
+                    top_k=32,
+                    temperature=0.6,
+                    repetition_penalty=1.2,
+                    no_repeat_ngram_size=15,
+                    max_new_tokens=1536,
+                    pad_token_id=tokenizer.eos_token_id)[0]
+    print(request)
+    answer_tokens = answer_tokens[len(request_tokens[0]):-1]
+    answer = tokenizer.decode(answer_tokens).strip()
+    print(answer)
+    return answer
+
+model_name = "gurgutan/saiga2-13b-4bit"
+system_prompt = "Ты — Сайга, русскоязычный автоматический ассистент. Ты разговариваешь с людьми и помогаешь им."
+tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=True)
+model = AutoGPTQForCausalLM.from_quantized(model_name, device=device)
 model.eval()
 
-input = "Сочини стих, который начинается словами: Буря мглою небо кроет"
-conversation = Conversation()
-conversation.add_user_message(input)
-prompt = conversation.get_prompt(tokenizer)
-output = generate(model, tokenizer, prompt, generation_config)
-print(inp)
-print(output)
+user_text = "Сочини стих, который начинается словами: Буря мглою небо кроет"
+answer_text = generate_answer(model, tokenizer, user_text)  
+print(answer_text)
 
 ```
 # Original model: [saiga2-13B-lora](https://huggingface.co/IlyaGusev/saiga2_13b_lora)