Supervised Fine-tuning Trainer

Supervised fine-tuning (or SFT for short) is a crucial step in RLHF. In TRL we provide an easy-to-use API to create your SFT models and train them with few lines of code on your dataset.

Check out a complete flexible example at examples/scripts/sft.py. Experimental support for Vision Language Models is also included in the example examples/scripts/vsft_llava.py.

Quickstart

If you have a dataset hosted on the 🤗 Hub, you can easily fine-tune your SFT model using SFTTrainer from TRL. Let us assume your dataset is imdb, the text you want to predict is inside the text field of the dataset, and you want to fine-tune the facebook/opt-350m model. The following code-snippet takes care of all the data pre-processing and training for you:

from datasets import load_dataset
from trl import SFTTrainer

dataset = load_dataset("imdb", split="train")

trainer = SFTTrainer(
    "facebook/opt-350m",
    train_dataset=dataset,
    dataset_text_field="text",
    max_seq_length=512,
)
trainer.train()

Make sure to pass a correct value for max_seq_length as the default value will be set to min(tokenizer.model_max_length, 1024).

You can also construct a model outside of the trainer and pass it as follows:

from transformers import AutoModelForCausalLM
from datasets import load_dataset
from trl import SFTTrainer

dataset = load_dataset("imdb", split="train")

model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")

trainer = SFTTrainer(
    model,
    train_dataset=dataset,
    dataset_text_field="text",
    max_seq_length=512,
)

trainer.train()

The above snippets will use the default training arguments from the transformers.TrainingArguments class. If you want to modify that, make sure to create your own TrainingArguments object and pass it to the SFTTrainer constructor as it is done on the supervised_finetuning.py script on the stack-llama example.

Advanced usage

Train on completions only

You can use the DataCollatorForCompletionOnlyLM to train your model on the generated prompts only. Note that this works only in the case when packing=False. To instantiate that collator for instruction data, pass a response template and the tokenizer. Here is an example of how it would work to fine-tune opt-350m on completions only on the CodeAlpaca dataset:

from transformers import AutoModelForCausalLM, AutoTokenizer
from datasets import load_dataset
from trl import SFTTrainer, DataCollatorForCompletionOnlyLM

dataset = load_dataset("lucasmccabe-lmi/CodeAlpaca-20k", split="train")

model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")

def formatting_prompts_func(example):
    output_texts = []
    for i in range(len(example['instruction'])):
        text = f"### Question: {example['instruction'][i]}\n ### Answer: {example['output'][i]}"
        output_texts.append(text)
    return output_texts

response_template = " ### Answer:"
collator = DataCollatorForCompletionOnlyLM(response_template, tokenizer=tokenizer)

trainer = SFTTrainer(
    model,
    train_dataset=dataset,
    formatting_func=formatting_prompts_func,
    data_collator=collator,
)

trainer.train()

To instantiate that collator for assistant style conversation data, pass a response template, an instruction template and the tokenizer. Here is an example of how it would work to fine-tune opt-350m on assistant completions only on the Open Assistant Guanaco dataset:

from transformers import AutoModelForCausalLM, AutoTokenizer
from datasets import load_dataset
from trl import SFTTrainer, DataCollatorForCompletionOnlyLM

dataset = load_dataset("timdettmers/openassistant-guanaco", split="train")

model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")

instruction_template = "### Human:"
response_template = "### Assistant:"
collator = DataCollatorForCompletionOnlyLM(instruction_template=instruction_template, response_template=response_template, tokenizer=tokenizer, mlm=False)

trainer = SFTTrainer(
    model,
    train_dataset=dataset,
    dataset_text_field="text",
    data_collator=collator,
)

trainer.train()

Make sure to have a pad_token_id which is different from eos_token_id which can result in the model not properly predicting EOS (End of Sentence) tokens during generation.

Using token_ids directly for response_template

Some tokenizers like Llama 2 (meta-llama/Llama-2-XXb-hf) tokenize sequences differently depending whether they have context or not. For example:

from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")

def print_tokens_with_ids(txt):
    tokens = tokenizer.tokenize(txt, add_special_tokens=False)
    token_ids = tokenizer.encode(txt, add_special_tokens=False)
    print(list(zip(tokens, token_ids)))

prompt = """### User: Hello\n\n### Assistant: Hi, how can I help you?"""
print_tokens_with_ids(prompt)  # [..., ('▁Hello', 15043), ('<0x0A>', 13), ('<0x0A>', 13), ('##', 2277), ('#', 29937), ('▁Ass', 4007), ('istant', 22137), (':', 29901), ...]

response_template = "### Assistant:"
print_tokens_with_ids(response_template)  # [('▁###', 835), ('▁Ass', 4007), ('istant', 22137), (':', 29901)]

In this case, and due to lack of context in response_template, the same string (”### Assistant:”) is tokenized differently:

• Text (with context): [2277, 29937, 4007, 22137, 29901]
• response_template (without context): [835, 4007, 22137, 29901]

This will lead to an error when the DataCollatorForCompletionOnlyLM does not find the response_template in the dataset example text:

RuntimeError: Could not find response key [835, 4007, 22137, 29901] in token IDs tensor([    1,   835,  ...])

To solve this, you can tokenize the response_template with the same context than in the dataset, truncate it as needed and pass the token_ids directly to the response_template argument of the DataCollatorForCompletionOnlyLM class. For example:

response_template_with_context = "\n### Assistant:"  # We added context here: "\n". This is enough for this tokenizer
response_template_ids = tokenizer.encode(response_template_with_context, add_special_tokens=False)[2:]  # Now we have it like in the dataset texts: `[2277, 29937, 4007, 22137, 29901]`

data_collator = DataCollatorForCompletionOnlyLM(response_template_ids, tokenizer=tokenizer)

Add Special Tokens for Chat Format

Adding special tokens to a language model is crucial for training chat models. These tokens are added between the different roles in a conversation, such as the user, assistant, and system and help the model recognize the structure and flow of a conversation. This setup is essential for enabling the model to generate coherent and contextually appropriate responses in a chat environment. The setup_chat_format() function in trl easily sets up a model and tokenizer for conversational AI tasks. This function:

• Adds special tokens to the tokenizer, e.g. <|im_start|> and <|im_end|>, to indicate the start and end of a conversation.
• Resizes the model’s embedding layer to accommodate the new tokens.
• Sets the chat_template of the tokenizer, which is used to format the input data into a chat-like format. The default is chatml from OpenAI.
• optionally you can pass resize_to_multiple_of to resize the embedding layer to a multiple of the resize_to_multiple_of argument, e.g. 64. If you want to see more formats being supported in the future, please open a GitHub issue on trl

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")

# Set up the chat format with default 'chatml' format
model, tokenizer = setup_chat_format(model, tokenizer)

With our model and tokenizer set up, we can now fine-tune our model on a conversational dataset. Below is an example of how a dataset can be formatted for fine-tuning.

Dataset format support

The SFTTrainer supports popular dataset formats. This allows you to pass the dataset to the trainer without any pre-processing directly. The following formats are supported:

• conversational format

{"messages": [{"role": "system", "content": "You are helpful"}, {"role": "user", "content": "What's the capital of France?"}, {"role": "assistant", "content": "..."}]}
{"messages": [{"role": "system", "content": "You are helpful"}, {"role": "user", "content": "Who wrote 'Romeo and Juliet'?"}, {"role": "assistant", "content": "..."}]}
{"messages": [{"role": "system", "content": "You are helpful"}, {"role": "user", "content": "How far is the Moon from Earth?"}, {"role": "assistant", "content": "..."}]}

• instruction format

{"prompt": "<prompt text>", "completion": "<ideal generated text>"}
{"prompt": "<prompt text>", "completion": "<ideal generated text>"}
{"prompt": "<prompt text>", "completion": "<ideal generated text>"}

If your dataset uses one of the above formats, you can directly pass it to the trainer without pre-processing. The SFTTrainer will then format the dataset for you using the defined format from the model’s tokenizer with the apply_chat_template method.

from datasets import load_dataset
from trl import SFTTrainer

...

# load jsonl dataset
dataset = load_dataset("json", data_files="path/to/dataset.jsonl", split="train")
# load dataset from the HuggingFace Hub
dataset = load_dataset("philschmid/dolly-15k-oai-style", split="train")

...

trainer = SFTTrainer(
    "facebook/opt-350m",
    args=training_args,
    train_dataset=dataset,
    packing=True,
)

If the dataset is not in one those format you can either preprocess the dataset to match the formatting or pass a formatting function to the SFTTrainer to do it for you. Let’s have a look.

Format your input prompts

For instruction fine-tuning, it is quite common to have two columns inside the dataset: one for the prompt & the other for the response. This allows people to format examples like Stanford-Alpaca did as follows:

Below is an instruction ...

### Instruction
{prompt}

### Response:
{completion}

Let us assume your dataset has two fields, question and answer. Therefore you can just run:

...
def formatting_prompts_func(example):
    output_texts = []
    for i in range(len(example['question'])):
        text = f"### Question: {example['question'][i]}\n ### Answer: {example['answer'][i]}"
        output_texts.append(text)
    return output_texts

trainer = SFTTrainer(
    model,
    train_dataset=dataset,
    formatting_func=formatting_prompts_func,
)

trainer.train()

To properly format your input make sure to process all the examples by looping over them and returning a list of processed text. Check out a full example on how to use SFTTrainer on alpaca dataset here

Packing dataset ( ConstantLengthDataset )

SFTTrainer supports example packing, where multiple short examples are packed in the same input sequence to increase training efficiency. This is done with the ConstantLengthDataset utility class that returns constant length chunks of tokens from a stream of examples. To enable the usage of this dataset class, simply pass packing=True to the SFTTrainer constructor.

...

trainer = SFTTrainer(
    "facebook/opt-350m",
    train_dataset=dataset,
    dataset_text_field="text",
    packing=True
)

trainer.train()

Note that if you use a packed dataset and if you pass max_steps in the training arguments you will probably train your models for more than few epochs, depending on the way you have configured the packed dataset and the training protocol. Double check that you know and understand what you are doing. If you don’t want to pack your eval_dataset, you can pass eval_packing=False to the SFTTrainer init method.

Customize your prompts using packed dataset

If your dataset has several fields that you want to combine, for example if the dataset has question and answer fields and you want to combine them, you can pass a formatting function to the trainer that will take care of that. For example:

def formatting_func(example):
    text = f"### Question: {example['question']}\n ### Answer: {example['answer']}"
    return text

trainer = SFTTrainer(
    "facebook/opt-350m",
    train_dataset=dataset,
    packing=True,
    formatting_func=formatting_func
)

trainer.train()

You can also customize the ConstantLengthDataset much more by directly passing the arguments to the SFTTrainer constructor. Please refer to that class’ signature for more information.

Control over the pretrained model

You can directly pass the kwargs of the from_pretrained() method to the SFTTrainer. For example, if you want to load a model in a different precision, analogous to

model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m", torch_dtype=torch.bfloat16)

...

trainer = SFTTrainer(
    "facebook/opt-350m",
    train_dataset=dataset,
    dataset_text_field="text",
    model_init_kwargs={
        "torch_dtype": torch.bfloat16,
    },
)

trainer.train()

Note that all keyword arguments of from_pretrained() are supported.

Training adapters

We also support a tight integration with 🤗 PEFT library so that any user can conveniently train adapters and share them on the Hub instead of training the entire model

from datasets import load_dataset
from trl import SFTTrainer
from peft import LoraConfig

dataset = load_dataset("imdb", split="train")

peft_config = LoraConfig(
    r=16,
    lora_alpha=32,
    lora_dropout=0.05,
    bias="none",
    task_type="CAUSAL_LM",
)

trainer = SFTTrainer(
    "EleutherAI/gpt-neo-125m",
    train_dataset=dataset,
    dataset_text_field="text",
    peft_config=peft_config
)

trainer.train()

You can also continue training your PeftModel. For that, first load a PeftModel outside SFTTrainer and pass it directly to the trainer without the peft_config argument being passed.

Training adapters with base 8 bit models

For that you need to first load your 8bit model outside the Trainer and pass a PeftConfig to the trainer. For example:

...

peft_config = LoraConfig(
    r=16,
    lora_alpha=32,
    lora_dropout=0.05,
    bias="none",
    task_type="CAUSAL_LM",
)

model = AutoModelForCausalLM.from_pretrained(
    "EleutherAI/gpt-neo-125m",
    load_in_8bit=True,
    device_map="auto",
)

trainer = SFTTrainer(
    model,
    train_dataset=dataset,
    dataset_text_field="text",
    peft_config=peft_config,
)

trainer.train()

Using Flash Attention and Flash Attention 2

You can benefit from Flash Attention 1 & 2 using SFTTrainer out of the box with minimal changes of code. First, to make sure you have all the latest features from transformers, install transformers from source

pip install -U git+https://github.com/huggingface/transformers.git

Note that Flash Attention only works on GPU now and under half-precision regime (when using adapters, base model loaded in half-precision) Note also both features are perfectly compatible with other tools such as quantization.

Using Flash-Attention 1

For Flash Attention 1 you can use the BetterTransformer API and force-dispatch the API to use Flash Attention kernel. First, install the latest optimum package:

pip install -U optimum

Once you have loaded your model, wrap the trainer.train() call under the with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=False, enable_mem_efficient=False): context manager:

...

+ with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=False, enable_mem_efficient=False):
    trainer.train()

Note that you cannot train your model using Flash Attention 1 on an arbitrary dataset as torch.scaled_dot_product_attention does not support training with padding tokens if you use Flash Attention kernels. Therefore you can only use that feature with packing=True. If your dataset contains padding tokens, consider switching to Flash Attention 2 integration.

Below are some numbers you can get in terms of speedup and memory efficiency, using Flash Attention 1, on a single NVIDIA-T4 16GB.

Using Flash Attention-2

To use Flash Attention 2, first install the latest flash-attn package:

pip install -U flash-attn

And add attn_implementation="flash_attention_2" when calling from_pretrained:

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    load_in_4bit=True,
    attn_implementation="flash_attention_2"
)

If you don’t use quantization, make sure your model is loaded in half-precision and dispatch your model on a supported GPU device. After loading your model, you can either train it as it is, or attach adapters and train adapters on it in case your model is quantized.

In contrary to Flash Attention 1, the integration makes it possible to train your model on an arbitrary dataset that also includes padding tokens.

Using model creation utility

We included a utility function to create your model.

from trl import ModelConfig, SFTTrainer, get_kbit_device_map, get_peft_config, get_quantization_config
model_config = ModelConfig(
    model_name_or_path="facebook/opt-350m"
    attn_implementation=None, # or "flash_attention_2"
)
torch_dtype = (
    model_config.torch_dtype
    if model_config.torch_dtype in ["auto", None]
    else getattr(torch, model_config.torch_dtype)
)
quantization_config = get_quantization_config(model_config)
model_kwargs = dict(
    revision=model_config.model_revision,
    trust_remote_code=model_config.trust_remote_code,
    attn_implementation=model_config.attn_implementation,
    torch_dtype=torch_dtype,
    use_cache=False if training_args.gradient_checkpointing else True,
    device_map=get_kbit_device_map() if quantization_config is not None else None,
    quantization_config=quantization_config,
)
model = AutoModelForCausalLM.from_pretrained(model_config.model_name_or_path, **model_kwargs)
trainer = SFTTrainer(
    ...,
    model=model_config.model_name_or_path,
    peft_config=get_peft_config(model_config),
)

Enhance model’s performances using NEFTune

NEFTune is a technique to boost the performance of chat models and was introduced by the paper “NEFTune: Noisy Embeddings Improve Instruction Finetuning” from Jain et al. it consists of adding noise to the embedding vectors during training. According to the abstract of the paper:

Standard finetuning of LLaMA-2-7B using Alpaca achieves 29.79% on AlpacaEval, which rises to 64.69% using noisy embeddings. NEFTune also improves over strong baselines on modern instruction datasets. Models trained with Evol-Instruct see a 10% improvement, with ShareGPT an 8% improvement, and with OpenPlatypus an 8% improvement. Even powerful models further refined with RLHF such as LLaMA-2-Chat benefit from additional training with NEFTune.

To use it in SFTTrainer simply pass neftune_noise_alpha when creating your SFTTrainer instance. Note that to avoid any surprising behaviour, NEFTune is disabled after training to retrieve back the original behaviour of the embedding layer.

from datasets import load_dataset
from trl import SFTTrainer

dataset = load_dataset("imdb", split="train")

trainer = SFTTrainer(
    "facebook/opt-350m",
    train_dataset=dataset,
    dataset_text_field="text",
    max_seq_length=512,
    neftune_noise_alpha=5,
)
trainer.train()

We have tested NEFTune by training mistralai/Mistral-7B-v0.1 on the OpenAssistant dataset and validated that using NEFTune led to a performance boost of ~25% on MT Bench.

Note however, that the amount of performance gain is dataset dependent and in particular, applying NEFTune on synthetic datasets like UltraChat typically produces smaller gains.

Accelerate fine-tuning 2x using unsloth

You can further accelerate QLoRA / LoRA (2x faster, 60% less memory) using the unsloth library that is fully compatible with SFTTrainer. Currently unsloth supports only Llama (Yi, TinyLlama, Qwen, Deepseek etc) and Mistral architectures. Some benchmarks on 1x A100 listed below:

First install unsloth according to the official documentation. Once installed, you can incorporate unsloth into your workflow in a very simple manner; instead of loading AutoModelForCausalLM, you just need to load a FastLanguageModel as follows:

import torch
from transformers import TrainingArguments
from trl import SFTTrainer
from unsloth import FastLanguageModel

max_seq_length = 2048 # Supports automatic RoPE Scaling, so choose any number

# Load model
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "unsloth/mistral-7b",
    max_seq_length = max_seq_length,
    dtype = None, # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
    load_in_4bit = True, # Use 4bit quantization to reduce memory usage. Can be False
    # token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf
)

# Do model patching and add fast LoRA weights
model = FastLanguageModel.get_peft_model(
    model,
    r = 16,
    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
                      "gate_proj", "up_proj", "down_proj",],
    lora_alpha = 16,
    lora_dropout = 0, # Dropout = 0 is currently optimized
    bias = "none",    # Bias = "none" is currently optimized
    use_gradient_checkpointing = True,
    random_state = 3407,
)

args = TrainingArguments(output_dir = "./output")

trainer = SFTTrainer(
    model = model,
    args = args,
    train_dataset = dataset,
    dataset_text_field = "text",
    max_seq_length = max_seq_length,
)
trainer.train()

The saved model is fully compatible with Hugging Face’s transformers library. Learn more about unsloth in their official repository.

Best practices

Pay attention to the following best practices when training a model with that trainer:

• SFTTrainer always pads by default the sequences to the max_seq_length argument of the SFTTrainer. If none is passed, the trainer will retrieve that value from the tokenizer. Some tokenizers do not provide default value, so there is a check to retrieve the minimum between 2048 and that value. Make sure to check it before training.
• For training adapters in 8bit, you might need to tweak the arguments of the prepare_model_for_kbit_training method from PEFT, hence we advise users to use prepare_in_int8_kwargs field, or create the PeftModel outside the SFTTrainer and pass it.
• For a more memory-efficient training using adapters, you can load the base model in 8bit, for that simply add load_in_8bit argument when creating the SFTTrainer, or create a base model in 8bit outside the trainer and pass it.
• If you create a model outside the trainer, make sure to not pass to the trainer any additional keyword arguments that are relative to from_pretrained() method.

Multi-GPU Training

Trainer (and thus SFTTrainer) supports multi-GPU training. If you run your script with python script.py it will default to using DP as the strategy, which may be slower than expected. To use DDP (which is generally recommended, see here for more info) you must launch the script with python -m torch.distributed.launch script.py or accelerate launch script.py. For DDP to work you must also check the following:

• If you’re using gradient_checkpointing, add the following to the TrainingArguments: gradient_checkpointing_kwargs={'use_reentrant':False} (more info here
• Ensure that the model is placed on the correct device:

from accelerate import PartialState
device_string = PartialState().process_index
model = AutoModelForCausalLM.from_pretrained(
     ...
    device_map={'':device_string}
)

GPTQ Conversion

You may experience some issues with GPTQ Quantization after completing training. Lowering gradient_accumulation_steps to 4 will resolve most issues during the quantization process to GPTQ format.

SFTTrainer

Datasets

In the SFTTrainer we smartly support datasets.IterableDataset in addition to other style datasets. This is useful if you are using large corpora that you do not want to save all to disk. The data will be tokenized and processed on the fly, even when packing is enabled.

Additionally, in the SFTTrainer, we support pre-tokenized datasets if they are datasets.Dataset or datasets.IterableDataset. In other words, if such a dataset has a column of input_ids, no further processing (tokenization or packing) will be done, and the dataset will be used as-is. This can be useful if you have pretokenized your dataset outside of this script and want to re-use it directly.

ConstantLengthDataset