Elgato_dark/unsloth
1
0
Fork 0
mirror of https://github.com/unslothai/unsloth synced 2026-04-21 13:37:39 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

EOL LF (unix line endings) normalization (#3478)

Browse source
This commit is contained in:
Dan Saunders 2025-10-17 19:22:42 -04:00 committed by GitHub
parent f62c454a86
commit f845cf964f
2 changed files with 230 additions and 228 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.gitattributes vendored Normal file
View file

@ -0,0 +1,2 @@
# Normalize Python files to LF line endings
*.py text eol=lf

456
unsloth-cli.py
View file

@ -1,228 +1,228 @@
#!/usr/bin/env python3 #!/usr/bin/env python3
""" """
🦥 Starter Script for Fine-Tuning FastLanguageModel with Unsloth 🦥 Starter Script for Fine-Tuning FastLanguageModel with Unsloth
This script is designed as a starting point for fine-tuning your models using unsloth. This script is designed as a starting point for fine-tuning your models using unsloth.
It includes configurable options for model loading, PEFT parameters, training arguments, It includes configurable options for model loading, PEFT parameters, training arguments,
and model saving/pushing functionalities. and model saving/pushing functionalities.
You will likely want to customize this script to suit your specific use case You will likely want to customize this script to suit your specific use case
and requirements. and requirements.
Here are a few suggestions for customization: Here are a few suggestions for customization:
- Modify the dataset loading and preprocessing steps to match your data. - Modify the dataset loading and preprocessing steps to match your data.
- Customize the model saving and pushing configurations. - Customize the model saving and pushing configurations.
Usage: (most of the options have valid default values this is an extended example for demonstration purposes) Usage: (most of the options have valid default values this is an extended example for demonstration purposes)
python unsloth-cli.py --model_name "unsloth/llama-3-8b" --max_seq_length 8192 --dtype None --load_in_4bit \ python unsloth-cli.py --model_name "unsloth/llama-3-8b" --max_seq_length 8192 --dtype None --load_in_4bit \
--r 64 --lora_alpha 32 --lora_dropout 0.1 --bias "none" --use_gradient_checkpointing "unsloth" \ --r 64 --lora_alpha 32 --lora_dropout 0.1 --bias "none" --use_gradient_checkpointing "unsloth" \
--random_state 3407 --use_rslora --per_device_train_batch_size 4 --gradient_accumulation_steps 8 \ --random_state 3407 --use_rslora --per_device_train_batch_size 4 --gradient_accumulation_steps 8 \
--warmup_steps 5 --max_steps 400 --learning_rate 2e-6 --logging_steps 1 --optim "adamw_8bit" \ --warmup_steps 5 --max_steps 400 --learning_rate 2e-6 --logging_steps 1 --optim "adamw_8bit" \
--weight_decay 0.005 --lr_scheduler_type "linear" --seed 3407 --output_dir "outputs" \ --weight_decay 0.005 --lr_scheduler_type "linear" --seed 3407 --output_dir "outputs" \
--report_to "tensorboard" --save_model --save_path "model" --quantization_method "f16" \ --report_to "tensorboard" --save_model --save_path "model" --quantization_method "f16" \
--push_model --hub_path "hf/model" --hub_token "your_hf_token" --push_model --hub_path "hf/model" --hub_token "your_hf_token"
To see a full list of configurable options, use: To see a full list of configurable options, use:
python unsloth-cli.py --help python unsloth-cli.py --help
Happy fine-tuning! Happy fine-tuning!
""" """
import argparse import argparse
import os import os
def run(args): def run(args):
import torch import torch
from unsloth import FastLanguageModel from unsloth import FastLanguageModel
from datasets import load_dataset from datasets import load_dataset
from transformers.utils import strtobool from transformers.utils import strtobool
from trl import SFTTrainer, SFTConfig from trl import SFTTrainer, SFTConfig
from transformers import TrainingArguments from transformers import TrainingArguments
from unsloth import is_bfloat16_supported from unsloth import is_bfloat16_supported
import logging import logging
logging.getLogger('hf-to-gguf').setLevel(logging.WARNING) logging.getLogger('hf-to-gguf').setLevel(logging.WARNING)
# Load model and tokenizer # Load model and tokenizer
model, tokenizer = FastLanguageModel.from_pretrained( model, tokenizer = FastLanguageModel.from_pretrained(
model_name=args.model_name, model_name=args.model_name,
max_seq_length=args.max_seq_length, max_seq_length=args.max_seq_length,
dtype=args.dtype, dtype=args.dtype,
load_in_4bit=args.load_in_4bit, load_in_4bit=args.load_in_4bit,
) )
# Configure PEFT model # Configure PEFT model
model = FastLanguageModel.get_peft_model( model = FastLanguageModel.get_peft_model(
model, model,
r=args.r, r=args.r,
target_modules=["q_proj", "k_proj", "v_proj", "o_proj", target_modules=["q_proj", "k_proj", "v_proj", "o_proj",
"gate_proj", "up_proj", "down_proj"], "gate_proj", "up_proj", "down_proj"],
lora_alpha=args.lora_alpha, lora_alpha=args.lora_alpha,
lora_dropout=args.lora_dropout, lora_dropout=args.lora_dropout,
bias=args.bias, bias=args.bias,
use_gradient_checkpointing=args.use_gradient_checkpointing, use_gradient_checkpointing=args.use_gradient_checkpointing,
random_state=args.random_state, random_state=args.random_state,
use_rslora=args.use_rslora, use_rslora=args.use_rslora,
loftq_config=args.loftq_config, loftq_config=args.loftq_config,
) )
alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request. alpaca_prompt = """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: ### Instruction:
{} {}
### Input: ### Input:
{} {}
### Response: ### Response:
{}""" {}"""
EOS_TOKEN = tokenizer.eos_token # Must add EOS_TOKEN EOS_TOKEN = tokenizer.eos_token # Must add EOS_TOKEN
def formatting_prompts_func(examples): def formatting_prompts_func(examples):
instructions = examples["instruction"] instructions = examples["instruction"]
inputs = examples["input"] inputs = examples["input"]
outputs = examples["output"] outputs = examples["output"]
texts = [] texts = []
for instruction, input, output in zip(instructions, inputs, outputs): for instruction, input, output in zip(instructions, inputs, outputs):
text = alpaca_prompt.format(instruction, input, output) + EOS_TOKEN text = alpaca_prompt.format(instruction, input, output) + EOS_TOKEN
texts.append(text) texts.append(text)
return {"text": texts} return {"text": texts}
use_modelscope = strtobool(os.environ.get('UNSLOTH_USE_MODELSCOPE', 'False')) use_modelscope = strtobool(os.environ.get('UNSLOTH_USE_MODELSCOPE', 'False'))
if use_modelscope: if use_modelscope:
from modelscope import MsDataset from modelscope import MsDataset
dataset = MsDataset.load(args.dataset, split="train") dataset = MsDataset.load(args.dataset, split="train")
else: else:
# Load and format dataset # Load and format dataset
dataset = load_dataset(args.dataset, split="train") dataset = load_dataset(args.dataset, split="train")
dataset = dataset.map(formatting_prompts_func, batched=True) dataset = dataset.map(formatting_prompts_func, batched=True)
print("Data is formatted and ready!") print("Data is formatted and ready!")
# Configure training arguments # Configure training arguments
training_args = SFTConfig( training_args = SFTConfig(
per_device_train_batch_size=args.per_device_train_batch_size, per_device_train_batch_size=args.per_device_train_batch_size,
gradient_accumulation_steps=args.gradient_accumulation_steps, gradient_accumulation_steps=args.gradient_accumulation_steps,
warmup_steps=args.warmup_steps, warmup_steps=args.warmup_steps,
max_steps=args.max_steps, max_steps=args.max_steps,
learning_rate=args.learning_rate, learning_rate=args.learning_rate,
fp16=not is_bfloat16_supported(), fp16=not is_bfloat16_supported(),
bf16=is_bfloat16_supported(), bf16=is_bfloat16_supported(),
logging_steps=args.logging_steps, logging_steps=args.logging_steps,
optim=args.optim, optim=args.optim,
weight_decay=args.weight_decay, weight_decay=args.weight_decay,
lr_scheduler_type=args.lr_scheduler_type, lr_scheduler_type=args.lr_scheduler_type,
seed=args.seed, seed=args.seed,
output_dir=args.output_dir, output_dir=args.output_dir,
report_to=args.report_to, report_to=args.report_to,
max_length=args.max_seq_length, max_length=args.max_seq_length,
dataset_num_proc=2, dataset_num_proc=2,
packing=False, packing=False,
) )
# Initialize trainer # Initialize trainer
trainer = SFTTrainer( trainer = SFTTrainer(
model=model, model=model,
processing_class=tokenizer, processing_class=tokenizer,
train_dataset=dataset, train_dataset=dataset,
args=training_args, args=training_args,
) )
# Train model # Train model
trainer_stats = trainer.train() trainer_stats = trainer.train()
# Save model # Save model
if args.save_model: if args.save_model:
# if args.quantization_method is a list, we will save the model for each quantization method # if args.quantization_method is a list, we will save the model for each quantization method
if args.save_gguf: if args.save_gguf:
if isinstance(args.quantization, list): if isinstance(args.quantization, list):
for quantization_method in args.quantization: for quantization_method in args.quantization:
print(f"Saving model with quantization method: {quantization_method}") print(f"Saving model with quantization method: {quantization_method}")
model.save_pretrained_gguf( model.save_pretrained_gguf(
args.save_path, args.save_path,
tokenizer, tokenizer,
quantization_method=quantization_method, quantization_method=quantization_method,
) )
if args.push_model: if args.push_model:
model.push_to_hub_gguf( model.push_to_hub_gguf(
hub_path=args.hub_path, hub_path=args.hub_path,
hub_token=args.hub_token, hub_token=args.hub_token,
quantization_method=quantization_method, quantization_method=quantization_method,
) )
else: else:
print(f"Saving model with quantization method: {args.quantization}") print(f"Saving model with quantization method: {args.quantization}")
model.save_pretrained_gguf(args.save_path, tokenizer, quantization_method=args.quantization) model.save_pretrained_gguf(args.save_path, tokenizer, quantization_method=args.quantization)
if args.push_model: if args.push_model:
model.push_to_hub_gguf( model.push_to_hub_gguf(
hub_path=args.hub_path, hub_path=args.hub_path,
hub_token=args.hub_token, hub_token=args.hub_token,
quantization_method=quantization_method, quantization_method=quantization_method,
) )
else: else:
model.save_pretrained_merged(args.save_path, tokenizer, args.save_method) model.save_pretrained_merged(args.save_path, tokenizer, args.save_method)
if args.push_model: if args.push_model:
model.push_to_hub_merged(args.save_path, tokenizer, args.hub_token) model.push_to_hub_merged(args.save_path, tokenizer, args.hub_token)
else: else:
print("Warning: The model is not saved!") print("Warning: The model is not saved!")
if __name__ == "__main__": if __name__ == "__main__":
# Define argument parser # Define argument parser
parser = argparse.ArgumentParser(description="🦥 Fine-tune your llm faster using unsloth!") parser = argparse.ArgumentParser(description="🦥 Fine-tune your llm faster using unsloth!")
model_group = parser.add_argument_group("🤖 Model Options") model_group = parser.add_argument_group("🤖 Model Options")
model_group.add_argument('--model_name', type=str, default="unsloth/llama-3-8b", help="Model name to load") model_group.add_argument('--model_name', type=str, default="unsloth/llama-3-8b", help="Model name to load")
model_group.add_argument('--max_seq_length', type=int, default=2048, help="Maximum sequence length, default is 2048. We auto support RoPE Scaling internally!") model_group.add_argument('--max_seq_length', type=int, default=2048, help="Maximum sequence length, default is 2048. We auto support RoPE Scaling internally!")
model_group.add_argument('--dtype', type=str, default=None, help="Data type for model (None for auto detection)") model_group.add_argument('--dtype', type=str, default=None, help="Data type for model (None for auto detection)")
model_group.add_argument('--load_in_4bit', action='store_true', help="Use 4bit quantization to reduce memory usage") model_group.add_argument('--load_in_4bit', action='store_true', help="Use 4bit quantization to reduce memory usage")
model_group.add_argument('--dataset', type=str, default="yahma/alpaca-cleaned", help="Huggingface dataset to use for training") model_group.add_argument('--dataset', type=str, default="yahma/alpaca-cleaned", help="Huggingface dataset to use for training")
lora_group = parser.add_argument_group("🧠 LoRA Options", "These options are used to configure the LoRA model.") lora_group = parser.add_argument_group("🧠 LoRA Options", "These options are used to configure the LoRA model.")
lora_group.add_argument('--r', type=int, default=16, help="Rank for Lora model, default is 16. (common values: 8, 16, 32, 64, 128)") lora_group.add_argument('--r', type=int, default=16, help="Rank for Lora model, default is 16. (common values: 8, 16, 32, 64, 128)")
lora_group.add_argument('--lora_alpha', type=int, default=16, help="LoRA alpha parameter, default is 16. (common values: 8, 16, 32, 64, 128)") lora_group.add_argument('--lora_alpha', type=int, default=16, help="LoRA alpha parameter, default is 16. (common values: 8, 16, 32, 64, 128)")
lora_group.add_argument('--lora_dropout', type=float, default=0.0, help="LoRA dropout rate, default is 0.0 which is optimized.") lora_group.add_argument('--lora_dropout', type=float, default=0.0, help="LoRA dropout rate, default is 0.0 which is optimized.")
lora_group.add_argument('--bias', type=str, default="none", help="Bias setting for LoRA") lora_group.add_argument('--bias', type=str, default="none", help="Bias setting for LoRA")
lora_group.add_argument('--use_gradient_checkpointing', type=str, default="unsloth", help="Use gradient checkpointing") lora_group.add_argument('--use_gradient_checkpointing', type=str, default="unsloth", help="Use gradient checkpointing")
lora_group.add_argument('--random_state', type=int, default=3407, help="Random state for reproducibility, default is 3407.") lora_group.add_argument('--random_state', type=int, default=3407, help="Random state for reproducibility, default is 3407.")
lora_group.add_argument('--use_rslora', action='store_true', help="Use rank stabilized LoRA") lora_group.add_argument('--use_rslora', action='store_true', help="Use rank stabilized LoRA")
lora_group.add_argument('--loftq_config', type=str, default=None, help="Configuration for LoftQ") lora_group.add_argument('--loftq_config', type=str, default=None, help="Configuration for LoftQ")
training_group = parser.add_argument_group("🎓 Training Options") training_group = parser.add_argument_group("🎓 Training Options")
training_group.add_argument('--per_device_train_batch_size', type=int, default=2, help="Batch size per device during training, default is 2.") training_group.add_argument('--per_device_train_batch_size', type=int, default=2, help="Batch size per device during training, default is 2.")
training_group.add_argument('--gradient_accumulation_steps', type=int, default=4, help="Number of gradient accumulation steps, default is 4.") training_group.add_argument('--gradient_accumulation_steps', type=int, default=4, help="Number of gradient accumulation steps, default is 4.")
training_group.add_argument('--warmup_steps', type=int, default=5, help="Number of warmup steps, default is 5.") training_group.add_argument('--warmup_steps', type=int, default=5, help="Number of warmup steps, default is 5.")
training_group.add_argument('--max_steps', type=int, default=400, help="Maximum number of training steps.") training_group.add_argument('--max_steps', type=int, default=400, help="Maximum number of training steps.")
training_group.add_argument('--learning_rate', type=float, default=2e-4, help="Learning rate, default is 2e-4.") training_group.add_argument('--learning_rate', type=float, default=2e-4, help="Learning rate, default is 2e-4.")
training_group.add_argument('--optim', type=str, default="adamw_8bit", help="Optimizer type.") training_group.add_argument('--optim', type=str, default="adamw_8bit", help="Optimizer type.")
training_group.add_argument('--weight_decay', type=float, default=0.01, help="Weight decay, default is 0.01.") training_group.add_argument('--weight_decay', type=float, default=0.01, help="Weight decay, default is 0.01.")
training_group.add_argument('--lr_scheduler_type', type=str, default="linear", help="Learning rate scheduler type, default is 'linear'.") training_group.add_argument('--lr_scheduler_type', type=str, default="linear", help="Learning rate scheduler type, default is 'linear'.")
training_group.add_argument('--seed', type=int, default=3407, help="Seed for reproducibility, default is 3407.") training_group.add_argument('--seed', type=int, default=3407, help="Seed for reproducibility, default is 3407.")
# Report/Logging arguments # Report/Logging arguments
report_group = parser.add_argument_group("📊 Report Options") report_group = parser.add_argument_group("📊 Report Options")
report_group.add_argument('--report_to', type=str, default="tensorboard", report_group.add_argument('--report_to', type=str, default="tensorboard",
choices=["azure_ml", "clearml", "codecarbon", "comet_ml", "dagshub", "dvclive", "flyte", "mlflow", "neptune", "tensorboard", "wandb", "all", "none"], choices=["azure_ml", "clearml", "codecarbon", "comet_ml", "dagshub", "dvclive", "flyte", "mlflow", "neptune", "tensorboard", "wandb", "all", "none"],
help="The list of integrations to report the results and logs to. Supported platforms are: \n\t\t 'azure_ml', 'clearml', 'codecarbon', 'comet_ml', 'dagshub', 'dvclive', 'flyte', 'mlflow', 'neptune', 'tensorboard', and 'wandb'. Use 'all' to report to all integrations installed, 'none' for no integrations.") help="The list of integrations to report the results and logs to. Supported platforms are: \n\t\t 'azure_ml', 'clearml', 'codecarbon', 'comet_ml', 'dagshub', 'dvclive', 'flyte', 'mlflow', 'neptune', 'tensorboard', and 'wandb'. Use 'all' to report to all integrations installed, 'none' for no integrations.")
report_group.add_argument('--logging_steps', type=int, default=1, help="Logging steps, default is 1") report_group.add_argument('--logging_steps', type=int, default=1, help="Logging steps, default is 1")
# Saving and pushing arguments # Saving and pushing arguments
save_group = parser.add_argument_group('💾 Save Model Options') save_group = parser.add_argument_group('💾 Save Model Options')
save_group.add_argument('--output_dir', type=str, default="outputs", help="Output directory") save_group.add_argument('--output_dir', type=str, default="outputs", help="Output directory")
save_group.add_argument('--save_model', action='store_true', help="Save the model after training") save_group.add_argument('--save_model', action='store_true', help="Save the model after training")
save_group.add_argument('--save_method', type=str, default="merged_16bit", choices=["merged_16bit", "merged_4bit", "lora"], help="Save method for the model, default is 'merged_16bit'") save_group.add_argument('--save_method', type=str, default="merged_16bit", choices=["merged_16bit", "merged_4bit", "lora"], help="Save method for the model, default is 'merged_16bit'")
save_group.add_argument('--save_gguf', action='store_true', help="Convert the model to GGUF after training") save_group.add_argument('--save_gguf', action='store_true', help="Convert the model to GGUF after training")
save_group.add_argument('--save_path', type=str, default="model", help="Path to save the model") save_group.add_argument('--save_path', type=str, default="model", help="Path to save the model")
save_group.add_argument('--quantization', type=str, default="q8_0", nargs="+", save_group.add_argument('--quantization', type=str, default="q8_0", nargs="+",
help="Quantization method for saving the model. common values ('f16', 'q4_k_m', 'q8_0'), Check our wiki for all quantization methods https://github.com/unslothai/unsloth/wiki#saving-to-gguf ") help="Quantization method for saving the model. common values ('f16', 'q4_k_m', 'q8_0'), Check our wiki for all quantization methods https://github.com/unslothai/unsloth/wiki#saving-to-gguf ")
push_group = parser.add_argument_group('🚀 Push Model Options') push_group = parser.add_argument_group('🚀 Push Model Options')
push_group.add_argument('--push_model', action='store_true', help="Push the model to Hugging Face hub after training") push_group.add_argument('--push_model', action='store_true', help="Push the model to Hugging Face hub after training")
push_group.add_argument('--push_gguf', action='store_true', help="Push the model as GGUF to Hugging Face hub after training") push_group.add_argument('--push_gguf', action='store_true', help="Push the model as GGUF to Hugging Face hub after training")
push_group.add_argument('--hub_path', type=str, default="hf/model", help="Path on Hugging Face hub to push the model") push_group.add_argument('--hub_path', type=str, default="hf/model", help="Path on Hugging Face hub to push the model")
push_group.add_argument('--hub_token', type=str, help="Token for pushing the model to Hugging Face hub") push_group.add_argument('--hub_token', type=str, help="Token for pushing the model to Hugging Face hub")
args = parser.parse_args() args = parser.parse_args()
run(args) run(args)