from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
import torch
from peft import PeftModel, PeftConfig


class GoralConversation:
    def __init__(
        self,
        message_template=" <s> {role}\n{content} </s>\n",
        system_prompt="Ты — Горал, русскоязычный автоматический ассистент. Ты разговариваешь с людьми и помогаешь им.",
        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,
            ]
        )
        final_text += " "
        final_text += tokenizer.decode([self.bot_token_id])
        return final_text.strip()


def generate(model, tokenizer, prompt, generation_config):
    data = tokenizer(
        prompt,
        return_tensors="pt",
        truncation=True,
        max_length=2048,
    )
    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()


weights_path = "dim/llama2_7b_dolly_oasst1_chip2"
access_token = ""

config = PeftConfig.from_pretrained(weights_path)
model = AutoModelForCausalLM.from_pretrained(
    config.base_model_name_or_path,
    load_in_8bit=True,
    torch_dtype=torch.float16,
    device_map={"": 0},
    token=access_token,
)
model = PeftModel.from_pretrained(
    model,
    weights_path,
    torch_dtype=torch.float16,
)
model.eval()

tokenizer = AutoTokenizer.from_pretrained(weights_path)
generation_config = GenerationConfig.from_pretrained(weights_path)
generation_config.do_sample = False


inp = "Напишите интересный пост в блоге о недавней поездке на Гавайи, рассказывая о культурном опыте и достопримечательностях, которые обязательно нужно увидеть."
conversation = GoralConversation(
    start_token_id=1,
    bot_token_id=9225,
)
conversation.add_user_message(inp)
prompt = conversation.get_prompt(tokenizer)

output = generate(model, tokenizer, prompt, generation_config)
print(inp)
print(output)

Training procedure

The following bitsandbytes quantization config was used during training:

Framework versions