finetune.py

import os
import sys

import torch
import torch.nn as nn
import bitsandbytes as bnb
from datasets import load_dataset
import transformers

assert (
    "LlamaTokenizer" in transformers._import_structure["models.llama"]
), "LLaMA is now in HuggingFace's main branch.\nPlease reinstall it: pip uninstall transformers && pip install git+https://github.com/huggingface/transformers.git"
from transformers import LlamaForCausalLM, LlamaTokenizer
from peft import (
    prepare_model_for_int8_training,
    LoraConfig,
    get_peft_model,
    get_peft_model_state_dict,
)

import argparse

parser = argparse.ArgumentParser(description='Process some integers.')
parser.add_argument('--size', type=str, help='the size of llama model')
parser.add_argument('--data', type=str, help='the data used for instructing tuning')
parser.add_argument('--local_rank', default=-1, type=int,help='node rank for distributed training')

args = parser.parse_args()


# optimized for RTX 4090. for larger GPUs, increase some of these?
MICRO_BATCH_SIZE = 4  # this could actually be 5 but i like powers of 2
BATCH_SIZE = 128 
GRADIENT_ACCUMULATION_STEPS = BATCH_SIZE // MICRO_BATCH_SIZE
EPOCHS = 3  # we don't always need 3 tbh
LEARNING_RATE = 3e-4  # the Karpathy constant
CUTOFF_LEN = 512  # 256 accounts for about 96% of the Alpaca data
LORA_R = 8
LORA_ALPHA = 16
LORA_DROPOUT = 0.05
VAL_SET_SIZE = 2000
TARGET_MODULES = [
    "q_proj",
    "v_proj",
]

if args.data == "alpaca":
    DATA_PATH = "data/alpaca_data_cleaned.json"
elif args.data == "belle":
    DATA_PATH = "data/belle_data_cn.json"
elif args.data == "alpaca-belle":
    DATA_PATH = "data/alpaca_plus_belle_data.json" 
elif args.data == "cot":
    DATA_PATH = "data/CoT_data.json"
elif args.data == "alpaca-cot":
    DATA_PATH = "data/alcapa_plus_cot.json"
elif args.data == "alpaca-belle-cot":
    DATA_PATH = "data/alcapa_plus_belle_plus_cot.json"
else:
    print("Error: Wrong type of data.")
    assert 1==2


# Your own local path for saving model weights
OUTPUT_DIR = "saved-"+args.data+args.size+"b"


device_map = "auto"
world_size = int(os.environ.get("WORLD_SIZE", 1))
ddp = world_size != 1
if ddp:
    device_map = {"": int(os.environ.get("LOCAL_RANK") or 0)}
    GRADIENT_ACCUMULATION_STEPS = GRADIENT_ACCUMULATION_STEPS // world_size


model = LlamaForCausalLM.from_pretrained(
    "decapoda-research/llama-"+args.size+"b-hf",
    load_in_8bit=True,
    device_map=device_map,
)
tokenizer = LlamaTokenizer.from_pretrained(
    "decapoda-research/llama-"+args.size+"b-hf", add_eos_token=True
)

model = prepare_model_for_int8_training(model)

config = LoraConfig(
    r=LORA_R,
    lora_alpha=LORA_ALPHA,
    target_modules=TARGET_MODULES,
    lora_dropout=LORA_DROPOUT,
    bias="none",
    task_type="CAUSAL_LM",
)
model = get_peft_model(model, config)
tokenizer.pad_token_id = 0  # unk. we want this to be different from the eos token
data = load_dataset("json", data_files=DATA_PATH)


def generate_prompt(data_point):
    # sorry about the formatting disaster gotta move fast
    if data_point["input"]:
        return f"""Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{data_point["instruction"]}

### Input:
{data_point["input"]}

### Response:
{data_point["output"]}"""
    else:
        return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{data_point["instruction"]}

### Response:
{data_point["output"]}"""


def tokenize(prompt):
    # there's probably a way to do this with the tokenizer settings
    # but again, gotta move fast
    result = tokenizer(
        prompt,
        truncation=True,
        max_length=CUTOFF_LEN + 1,
        padding="max_length",
    )
    return {
        "input_ids": result["input_ids"][:-1],
        "attention_mask": result["attention_mask"][:-1],
    }


def generate_and_tokenize_prompt(data_point):
    # This function masks out the labels for the input,
    # so that our loss is computed only on the response.
    user_prompt = (
        (
            f"""Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{data_point["instruction"]}

### Input:
{data_point["input"]}

### Response:
"""
        )
        if data_point["input"]
        else (
            f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{data_point["instruction"]}

### Response:
"""
        )
    )
    len_user_prompt_tokens = (
        len(
            tokenizer(
                user_prompt,
                truncation=True,
                max_length=CUTOFF_LEN + 1,
            )["input_ids"]
        )
        - 1
    )  # no eos token
    full_tokens = tokenizer(
        user_prompt + data_point["output"],
        truncation=True,
        max_length=CUTOFF_LEN + 1,
        padding="max_length",
    )["input_ids"][:-1]
    return {
        "input_ids": full_tokens,
        "labels": [-100] * len_user_prompt_tokens
        + full_tokens[len_user_prompt_tokens:],
        "attention_mask": [1] * (len(full_tokens)),
    }


if VAL_SET_SIZE > 0:
    train_val = data["train"].train_test_split(
        test_size=VAL_SET_SIZE, shuffle=True, seed=42
    )
    train_data = train_val["train"].shuffle().map(generate_and_tokenize_prompt)
    val_data = train_val["test"].shuffle().map(generate_and_tokenize_prompt)
else:
    train_data = data["train"].shuffle().map(generate_and_tokenize_prompt)
    val_data = None
    
# from https://github.com/tloen/alpaca-lora/fintune.py    
if not ddp and torch.cuda.device_count() > 1:
        # keeps Trainer from trying its own DataParallelism when more than 1 gpu is available
        model.is_parallelizable = True
        model.model_parallel = True

trainer = transformers.Trainer(
    model=model,
    train_dataset=train_data,
    eval_dataset=val_data,
    args=transformers.TrainingArguments(
        per_device_train_batch_size=MICRO_BATCH_SIZE,
        gradient_accumulation_steps=GRADIENT_ACCUMULATION_STEPS,
        warmup_steps=100,
        num_train_epochs=EPOCHS,
        learning_rate=LEARNING_RATE,
        fp16=True,
        logging_steps=20,
        evaluation_strategy="steps" if VAL_SET_SIZE > 0 else "no",
        save_strategy="steps",
        eval_steps=200 if VAL_SET_SIZE > 0 else None,
        save_steps=200,
        output_dir=OUTPUT_DIR,
        save_total_limit=3,
        load_best_model_at_end=True if VAL_SET_SIZE > 0 else False,
        ddp_find_unused_parameters=False if ddp else None,
    ),
    data_collator=transformers.DataCollatorForLanguageModeling(tokenizer, mlm=False),
)
model.config.use_cache = False

old_state_dict = model.state_dict
model.state_dict = (
    lambda self, *_, **__: get_peft_model_state_dict(self, old_state_dict())
).__get__(model, type(model))

if torch.__version__ >= "2" and sys.platform != "win32":
    model = torch.compile(model)

trainer.train()

model.save_pretrained(OUTPUT_DIR)

print("\n If there's a warning about missing keys above, please disregard :)")