falcon-finetune-personachat.py

import torch, einops
from datasets import load_dataset
from peft import LoraConfig
from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    BitsAndBytesConfig,
    AutoTokenizer,
    TrainingArguments
)
from peft.tuners.lora import LoraLayer

from trl import SFTTrainer

template = """### Personality:
{personality}

### History:
{history}

### Response:
"""

model_name = "tiiuae/falcon-7b"
dataset_name = "bavard/personachat_truecased"

def create_and_prepare_model():
    compute_dtype = getattr(torch, "float16")

    bnb_config = BitsAndBytesConfig(
        load_in_4bit=True,
        bnb_4bit_quant_type="nf4",
        bnb_4bit_compute_dtype=compute_dtype,
        bnb_4bit_use_double_quant=True,
    )

    # device_map={"": 0}
    device_map="auto"

    model = AutoModelForCausalLM.from_pretrained(
        model_name, quantization_config=bnb_config, device_map=device_map, trust_remote_code=True
    )
    model = AutoModelForCausalLM.from_pretrained(model_name, device_map=device_map, trust_remote_code=True)

    peft_config = LoraConfig(
        lora_alpha=16,
        lora_dropout=0.1,
        r=64,
        bias="none",
        task_type="CAUSAL_LM",
        target_modules=[
            "query_key_value"
        ],
    )

    tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
    tokenizer.pad_token = tokenizer.eos_token

    return model, peft_config, tokenizer


training_arguments = TrainingArguments(
    output_dir="./results",
    per_device_train_batch_size=1,
    gradient_accumulation_steps=4,
    optim="paged_adamw_32bit",
    save_steps=1000,
    logging_steps=10,
    learning_rate=2e-4,
    fp16=True,
    max_grad_norm=0.3,
    max_steps=10000,
    warmup_ratio=0.03,
    group_by_length=False,
    lr_scheduler_type="constant",
)

dataset = load_dataset(dataset_name, split="train")
model, peft_config, tokenizer = create_and_prepare_model()
model.config.use_cache = False

def formatting_func(example):
    return template.format(
        personality = "\n".join(example["personality"]),
        history = "\n".join(example["history"]),
        response = example["candidates"][-1]
    )

trainer = SFTTrainer(
    model=model,
    train_dataset=dataset,
    peft_config=peft_config,
    max_seq_length=512,
    tokenizer=tokenizer,
    args=training_arguments,
    packing=True,
    formatting_func=formatting_func
)

trainer.train()