Implementation

Supervised Finetuning with LLMs¶

Large Language Models or LLMs have seen a lot of progress in the last year. We went from now ChatGPT competitor to a whole zoo of LLMs, including Meta AI's Llama 3, Mistrals Mistral & Mixtral models, TII Falcon, and many more. Those LLMs can be used for a variety of tasks, including chatbots, question answering, summarization without any additional training. However, if you want to customize a model for your application. You may need to fine-tune the model on your data to achieve higher quality results than prompting or saving cost by training smaller models more efficient model.

This blog post walks you thorugh how to fine-tune open LLMs using Hugging Face TRL, Transformers & datasets in 2024. In the blog, we are going to:

1. Define our use case
2. Setup development environment
3. Create and prepare the dataset
4. Fine-tune LLM using trl and the SFTTrainer
5. Test and evaluate the LLM
6. Deploy the LLM for Production

Note: This blog was created to run on consumer size GPUs (24GB), e.g. NVIDIA A10G or RTX 4090/3090, but can be easily adapted to run on bigger GPUs.

1. Define our use case¶

When fine-tuning LLMs, it is important you know your use case and the task you want to solve. This will help you to choose the right model or help you to create a dataset to fine-tune your model. If you haven't defined your use case yet. You might want to go back to the drawing board. I want to mention that not all use cases require fine-tuning and it is always recommended to evaluate and try out already fine-tuned models or API-based models before fine-tuning your own model.

As an example, we are going to use the following use case:

We want to fine-tune a model, which can generate SQL queries based on a natural language instruction, which can then be integrated into our BI tool. The goal is to reduce the time it takes to create a SQL query and make it easier for non-technical users to create SQL queries.

Text to SQL can be a good use case for fine-tuning LLMs, as it is a complex task that requires a lot of (internal) knowledge about the data and the SQL language.

2. Setup development environment¶

Our first step is to install Hugging Face Libraries and Pyroch, including trl, transformers and datasets. If you haven't heard of trl yet, don't worry. It is a new library on top of transformers and datasets, which makes it easier to fine-tune, rlhf, align open LLMs.

In [ ]:

# Install Pytorch & other libraries
%pip install "torch==2.4.0" tensorboard

# Install Hugging Face libraries
%pip install  --upgrade \
  "transformers==4.44.2" \
  "datasets==2.21.0" \
  "accelerate==0.33.0" \
  "evaluate==0.4.2" \
  "bitsandbytes==0.43.3" \
  "trl==0.9.6" \
  "peft==0.12.0" 

# Install Pytorch & other libraries %pip install "torch==2.4.0" tensorboard # Install Hugging Face libraries %pip install --upgrade \ "transformers==4.44.2" \ "datasets==2.21.0" \ "accelerate==0.33.0" \ "evaluate==0.4.2" \ "bitsandbytes==0.43.3" \ "trl==0.9.6" \ "peft==0.12.0"

If you are using a GPU with Ampere architecture (e.g. NVIDIA A10G or RTX 4090/3090) or newer you can use Flash attention. Flash Attention is a an method that reorders the attention computation and leverages classical techniques (tiling, recomputation) to significantly speed it up and reduce memory usage from quadratic to linear in sequence length. The TL;DR; accelerates training up to 3x. Learn more at FlashAttention.

Note: If your machine has less than 96GB of RAM and lots of CPU cores, reduce the number of MAX_JOBS. On the g6.2xlarge we used 4.

In [ ]:

import torch; assert torch.cuda.get_device_capability()[0] >= 8, 'Hardware not supported for Flash Attention'
# install flash-attn
!pip install ninja packaging
!MAX_JOBS=4 pip install flash-attn --no-build-isolation

import torch; assert torch.cuda.get_device_capability()[0] >= 8, 'Hardware not supported for Flash Attention' # install flash-attn !pip install ninja packaging !MAX_JOBS=4 pip install flash-attn --no-build-isolation

Installing flash attention can take quite a bit of time (10-45 minutes).

We will use the Hugging Face Hub as a remote model versioning service. This means we will automatically push our model, logs and information to the Hub during training. You must register on the Hugging Face for this. After you have an account, we will use the login util from the huggingface_hub package to log into our account and store our token (access key) on the disk.

In [16]:

from huggingface_hub import login

login(
  token="", # ADD YOUR TOKEN HERE
  add_to_git_credential=True
)

from huggingface_hub import login login( token="", # ADD YOUR TOKEN HERE add_to_git_credential=True )

Token is valid (permission: write).
Cannot authenticate through git-credential as no helper is defined on your machine.
You might have to re-authenticate when pushing to the Hugging Face Hub.
Run the following command in your terminal in case you want to set the 'store' credential helper as default.

git config --global credential.helper store

Read https://git-scm.com/book/en/v2/Git-Tools-Credential-Storage for more details.
Token has not been saved to git credential helper.
Your token has been saved to /home/ubuntu/.cache/huggingface/token
Login successful

3. Create and prepare the dataset¶

Once you have determined that fine-tuning is the right solution we need to create a dataset to fine-tune our model. The dataset should be a diverse set of demonstrations of the task you want to solve. There are several ways to create such a dataset, including:

• Using existing open-source datasets, e.g., Spider
• Using LLMs to create synthetically datasets, e.g., Alpaca
• Using Humans to create datasets, e.g., Dolly.
• Using a combination of the above methods, e.g., Orca

Each of the methods has its own advantages and disadvantages and depends on the budget, time, and quality requirements. For example, using an existing dataset is the easiest but might not be tailored to your specific use case, while using humans might be the most accurate but can be time-consuming and expensive. It is also possible to combine several methods to create an instruction dataset, as shown in Orca: Progressive Learning from Complex Explanation Traces of GPT-4.

In our example we will use an already existing dataset called sql-create-context, which contains samples of natural language instructions, schema definitions and the corresponding SQL query.

With the latest release of trl we now support popular instruction and conversation dataset formats. This means we only need to convert our dataset to one of the supported formats and trl will take care of the rest. Those formats include:

• conversational format

{"messages": [{"role": "system", "content": "You are..."}, {"role": "user", "content": "..."}, {"role": "assistant", "content": "..."}]}
{"messages": [{"role": "system", "content": "You are..."}, {"role": "user", "content": "..."}, {"role": "assistant", "content": "..."}]}
{"messages": [{"role": "system", "content": "You are..."}, {"role": "user", "content": "..."}, {"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>"}

In our example we are going to load our open-source dataset using the 🤗 Datasets library and then convert it into the the conversational format, where we include the schema definition in the system message for our assistant. We'll then save the dataset as jsonl file, which we can then use to fine-tune our model. We are randomly downsampling the dataset to only 10,000 samples.

Note: This step can be different for your use case. For example, if you have already a dataset from, e.g. working with OpenAI, you can skip this step and go directly to the fine-tuning step.

In [17]:

from datasets import load_dataset

# Convert dataset to OAI messages
system_message = """You are an text to SQL query translator. Users will ask you questions in English and you will generate a SQL query based on the provided SCHEMA.
SCHEMA:
{schema}"""

def create_conversation(sample):
  return {
    "messages": [
      {"role": "system", "content": system_message.format(schema=sample["context"])},
      {"role": "user", "content": sample["question"]},
      {"role": "assistant", "content": sample["answer"]}
    ]
  }  

# Load dataset from the hub
dataset = load_dataset("b-mc2/sql-create-context", split="train")
dataset = dataset.shuffle().select(range(12500))

# Convert dataset to OAI messages
dataset = dataset.map(create_conversation, remove_columns=dataset.features,batched=False)
# split dataset into 10,000 training samples and 2,500 test samples
dataset = dataset.train_test_split(test_size=2500/12500)

print(dataset["train"][345]["messages"])

# save datasets to disk 
dataset["train"].to_json("train_dataset.json", orient="records")
dataset["test"].to_json("test_dataset.json", orient="records")

from datasets import load_dataset # Convert dataset to OAI messages system_message = """You are an text to SQL query translator. Users will ask you questions in English and you will generate a SQL query based on the provided SCHEMA. SCHEMA: {schema}""" def create_conversation(sample): return { "messages": [ {"role": "system", "content": system_message.format(schema=sample["context"])}, {"role": "user", "content": sample["question"]}, {"role": "assistant", "content": sample["answer"]} ] } # Load dataset from the hub dataset = load_dataset("b-mc2/sql-create-context", split="train") dataset = dataset.shuffle().select(range(12500)) # Convert dataset to OAI messages dataset = dataset.map(create_conversation, remove_columns=dataset.features,batched=False) # split dataset into 10,000 training samples and 2,500 test samples dataset = dataset.train_test_split(test_size=2500/12500) print(dataset["train"][345]["messages"]) # save datasets to disk dataset["train"].to_json("train_dataset.json", orient="records") dataset["test"].to_json("test_dataset.json", orient="records")

Downloading readme:   0%|          | 0.00/4.43k [00:00<?, ?B/s]

Map:   0%|          | 0/12500 [00:00<?, ? examples/s]

[{'content': 'You are an text to SQL query translator. Users will ask you questions in English and you will generate a SQL query based on the provided SCHEMA.\nSCHEMA:\nCREATE TABLE table_name_32 (date VARCHAR, attendance VARCHAR)', 'role': 'system'}, {'content': 'On what Date was the Attendance 73,405?', 'role': 'user'}, {'content': 'SELECT date FROM table_name_32 WHERE attendance = "73,405"', 'role': 'assistant'}]

Creating json from Arrow format:   0%|          | 0/10 [00:00<?, ?ba/s]

Creating json from Arrow format:   0%|          | 0/3 [00:00<?, ?ba/s]

Out[17]:

1187015

4. Fine-tune LLM using trl and the SFTTrainer¶

We are now ready to fine-tune our model. We will use the SFTTrainer from trl to fine-tune our model. The SFTTrainer makes it straightfoward to supervise fine-tune open LLMs. The SFTTrainer is a subclass of the Trainer from the transformers library and supports all the same features, including logging, evaluation, and checkpointing, but adds additiional quality of life features, including:

• Dataset formatting, including conversational and instruction format
• Training on completions only, ignoring prompts
• Packing datasets for more efficient training
• PEFT (parameter-efficient fine-tuning) support including Q-LoRA
• Preparing the model and tokenizer for conversational fine-tuning (e.g. adding special tokens)

We will use the dataset formatting, packing and PEFT features in our example. As peft method we will use QLoRA a technique to reduce the memory footprint of large language models during finetuning, without sacrificing performance by using quantization. If you want to learn more about QLoRA and how it works, check out Making LLMs even more accessible with bitsandbytes, 4-bit quantization and QLoRA blog post.

Now, lets get started! 🚀

First, we need to load our dataset from disk.

In [1]:

from datasets import load_dataset

# Load jsonl data from disk
dataset = load_dataset("json", data_files="train_dataset.json", split="train")

from datasets import load_dataset # Load jsonl data from disk dataset = load_dataset("json", data_files="train_dataset.json", split="train")

Next, we will load our LLM. For our use case we are going to use Llama 3.1 8B. But we can easily swap out the model for another model, e.g. Mistral or Mixtral models, TII Falcon, or any other LLMs by changing our model_id variable. We will use bitsandbytes to quantize our model to 4-bit.

Note: Be aware the bigger the model the more memory it will require. In our example we will use the 8B version, which can be tuned on 24GB GPUs. If you have a smaller GPU.

Correctly, preparing the LLM and Tokenizer for training chat/conversational models is crucial. We need to add new special tokens to the tokenizer and model and teach to understand the different roles in a conversation. In trl we have a convinient method called setup_chat_format, which:

• 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.

In [2]:

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
from trl import setup_chat_format

# Hugging Face model id
model_id = "meta-llama/Meta-Llama-3.1-8B" # or `mistralai/Mistral-7B-v0.1`

# BitsAndBytesConfig int-4 config
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16
)

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    attn_implementation="flash_attention_2",
    torch_dtype=torch.bfloat16,
    quantization_config=bnb_config
)
tokenizer = AutoTokenizer.from_pretrained(model_id)
tokenizer.padding_side = 'right' # to prevent warnings

# # set chat template to OAI chatML, remove if you start from a fine-tuned model
model, tokenizer = setup_chat_format(model, tokenizer)

import torch from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig from trl import setup_chat_format # Hugging Face model id model_id = "meta-llama/Meta-Llama-3.1-8B" # or `mistralai/Mistral-7B-v0.1` # BitsAndBytesConfig int-4 config bnb_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16 ) # Load model and tokenizer model = AutoModelForCausalLM.from_pretrained( model_id, device_map="auto", attn_implementation="flash_attention_2", torch_dtype=torch.bfloat16, quantization_config=bnb_config ) tokenizer = AutoTokenizer.from_pretrained(model_id) tokenizer.padding_side = 'right' # to prevent warnings # # set chat template to OAI chatML, remove if you start from a fine-tuned model model, tokenizer = setup_chat_format(model, tokenizer)

Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]

The SFTTrainer  supports a native integration with peft, which makes it super easy to efficiently tune LLMs using, e.g. QLoRA. We only need to create our LoraConfig and provide it to the trainer. Our LoraConfig parameters are defined based on the qlora paper and sebastian's blog post.

In [3]:

from peft import LoraConfig

# LoRA config based on QLoRA paper & Sebastian Raschka experiment
peft_config = LoraConfig(
        lora_alpha=128,
        lora_dropout=0.05,
        r=256,
        bias="none",
        target_modules="all-linear",
        task_type="CAUSAL_LM", 
)

from peft import LoraConfig # LoRA config based on QLoRA paper & Sebastian Raschka experiment peft_config = LoraConfig( lora_alpha=128, lora_dropout=0.05, r=256, bias="none", target_modules="all-linear", task_type="CAUSAL_LM", )

Before we can start our training we need to define the hyperparameters (TrainingArguments) we want to use.

In [4]:

from transformers import TrainingArguments

args = TrainingArguments(
    output_dir="code-llama-3-1-8b-text-to-sql", # directory to save and repository id
    num_train_epochs=3,                     # number of training epochs
    per_device_train_batch_size=1,          # batch size per device during training
    gradient_accumulation_steps=8,          # number of steps before performing a backward/update pass
    gradient_checkpointing=True,            # use gradient checkpointing to save memory
    optim="adamw_torch_fused",              # use fused adamw optimizer
    logging_steps=10,                       # log every 10 steps
    save_strategy="epoch",                  # save checkpoint every epoch
    learning_rate=2e-4,                     # learning rate, based on QLoRA paper
    bf16=True,                              # use bfloat16 precision
    tf32=True,                              # use tf32 precision
    max_grad_norm=0.3,                      # max gradient norm based on QLoRA paper
    warmup_ratio=0.03,                      # warmup ratio based on QLoRA paper
    lr_scheduler_type="constant",           # use constant learning rate scheduler
    push_to_hub=True,                       # push model to hub
    report_to="tensorboard",                # report metrics to tensorboard
)

from transformers import TrainingArguments args = TrainingArguments( output_dir="code-llama-3-1-8b-text-to-sql", # directory to save and repository id num_train_epochs=3, # number of training epochs per_device_train_batch_size=1, # batch size per device during training gradient_accumulation_steps=8, # number of steps before performing a backward/update pass gradient_checkpointing=True, # use gradient checkpointing to save memory optim="adamw_torch_fused", # use fused adamw optimizer logging_steps=10, # log every 10 steps save_strategy="epoch", # save checkpoint every epoch learning_rate=2e-4, # learning rate, based on QLoRA paper bf16=True, # use bfloat16 precision tf32=True, # use tf32 precision max_grad_norm=0.3, # max gradient norm based on QLoRA paper warmup_ratio=0.03, # warmup ratio based on QLoRA paper lr_scheduler_type="constant", # use constant learning rate scheduler push_to_hub=True, # push model to hub report_to="tensorboard", # report metrics to tensorboard )

We now have every building block we need to create our SFTTrainer to start then training our model.

In [5]:

from trl import SFTTrainer

max_seq_length = 2048 # max sequence length for model and packing of the dataset

trainer = SFTTrainer(
    model=model,
    args=args,
    train_dataset=dataset,
    peft_config=peft_config,
    max_seq_length=max_seq_length,
    tokenizer=tokenizer,
    packing=True,
    dataset_kwargs={
        "add_special_tokens": False,  # We template with special tokens
        "append_concat_token": False, # No need to add additional separator token
    }
)

from trl import SFTTrainer max_seq_length = 2048 # max sequence length for model and packing of the dataset trainer = SFTTrainer( model=model, args=args, train_dataset=dataset, peft_config=peft_config, max_seq_length=max_seq_length, tokenizer=tokenizer, packing=True, dataset_kwargs={ "add_special_tokens": False, # We template with special tokens "append_concat_token": False, # No need to add additional separator token } )

/opt/conda/envs/pytorch/lib/python3.10/site-packages/huggingface_hub/utils/_deprecation.py:100: FutureWarning: Deprecated argument(s) used in '__init__': max_seq_length, packing, dataset_kwargs. Will not be supported from version '1.0.0'.

Deprecated positional argument(s) used in SFTTrainer, please use the SFTConfig to set these arguments instead.
  warnings.warn(message, FutureWarning)
/opt/conda/envs/pytorch/lib/python3.10/site-packages/trl/trainer/sft_trainer.py:192: UserWarning: You passed a `packing` argument to the SFTTrainer, the value you passed will override the one in the `SFTConfig`.
  warnings.warn(
/opt/conda/envs/pytorch/lib/python3.10/site-packages/trl/trainer/sft_trainer.py:280: UserWarning: You passed a `max_seq_length` argument to the SFTTrainer, the value you passed will override the one in the `SFTConfig`.
  warnings.warn(
/opt/conda/envs/pytorch/lib/python3.10/site-packages/trl/trainer/sft_trainer.py:366: UserWarning: You passed a `dataset_kwargs` argument to the SFTTrainer, the value you passed will override the one in the `SFTConfig`.
  warnings.warn(

Start training our model by calling the train() method on our Trainer instance. This will start the training loop and train our model for 3 epochs. Since we are using a PEFT method, we will only save the adapted model weights and not the full model.

In [6]:

# start training, the model will be automatically saved to the hub and the output directory
trainer.train()

# save model 
trainer.save_model()

# start training, the model will be automatically saved to the hub and the output directory trainer.train() # save model trainer.save_model()

`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
/opt/conda/envs/pytorch/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py:600: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. In version 2.4 we will raise an exception if use_reentrant is not passed. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.
  return fn(*args, **kwargs)
The input hidden states seems to be silently casted in float32, this might be related to the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in torch.bfloat16.
/opt/conda/envs/pytorch/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]

[186/186 2:19:28, Epoch 2/3]

Step	Training Loss
10	0.912000
20	0.635400
30	0.605700
40	0.583800
50	0.568800
60	0.554400
70	0.508300
80	0.484600
90	0.475800
100	0.484700
110	0.481500
120	0.478700
130	0.436100
140	0.391300
150	0.396100
160	0.396200
170	0.404600
180	0.408300

/opt/conda/envs/pytorch/lib/python3.10/site-packages/peft/utils/save_and_load.py:232: UserWarning: Setting `save_embedding_layers` to `True` as the embedding layer has been resized during finetuning.
  warnings.warn(
/opt/conda/envs/pytorch/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py:600: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. In version 2.4 we will raise an exception if use_reentrant is not passed. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.
  return fn(*args, **kwargs)
/opt/conda/envs/pytorch/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/opt/conda/envs/pytorch/lib/python3.10/site-packages/peft/utils/save_and_load.py:232: UserWarning: Setting `save_embedding_layers` to `True` as the embedding layer has been resized during finetuning.
  warnings.warn(
/opt/conda/envs/pytorch/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py:600: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. In version 2.4 we will raise an exception if use_reentrant is not passed. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.
  return fn(*args, **kwargs)
/opt/conda/envs/pytorch/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/opt/conda/envs/pytorch/lib/python3.10/site-packages/peft/utils/save_and_load.py:232: UserWarning: Setting `save_embedding_layers` to `True` as the embedding layer has been resized during finetuning.
  warnings.warn(
/opt/conda/envs/pytorch/lib/python3.10/site-packages/peft/utils/save_and_load.py:232: UserWarning: Setting `save_embedding_layers` to `True` as the embedding layer has been resized during finetuning.
  warnings.warn(
/opt/conda/envs/pytorch/lib/python3.10/site-packages/peft/utils/save_and_load.py:232: UserWarning: Setting `save_embedding_layers` to `True` as the embedding layer has been resized during finetuning.
  warnings.warn(

events.out.tfevents.1724931649.ip-172-31-27-157.2377.0:   0%|          | 0.00/10.1k [00:00<?, ?B/s]

The training with Flash Attention for 3 epochs with a dataset of 10k samples took 02:05:58 on a g6.2xlarge. The instance costs 1,212$/h which brings us to a total cost of only 1.8$.

In [7]:

# free the memory again
del model
del trainer
torch.cuda.empty_cache()

# free the memory again del model del trainer torch.cuda.empty_cache()

Merge LoRA adapter in to the original model¶

When using QLoRA, we only train adapters and not the full model. This means when saving the model during training we only save the adapter weights and not the full model. If you want to save the full model, which makes it easier to use with Text Generation Inference you can merge the adapter weights into the model weights using the merge_and_unload method and then save the model with the save_pretrained method. This will save a default model, which can be used for inference.

Note: This requires > 30GB CPU Memory.

In [8]:

#### COMMENT IN TO MERGE PEFT AND BASE MODEL ####
from peft import AutoPeftModelForCausalLM

# Load PEFT model on CPU
model = AutoPeftModelForCausalLM.from_pretrained(
    args.output_dir,
    torch_dtype=torch.float16,
    low_cpu_mem_usage=True,
)  
# Merge LoRA and base model and save
merged_model = model.merge_and_unload()
merged_model.save_pretrained(args.output_dir,safe_serialization=True, max_shard_size="2GB")

#### COMMENT IN TO MERGE PEFT AND BASE MODEL #### from peft import AutoPeftModelForCausalLM # Load PEFT model on CPU model = AutoPeftModelForCausalLM.from_pretrained( args.output_dir, torch_dtype=torch.float16, low_cpu_mem_usage=True, ) # Merge LoRA and base model and save merged_model = model.merge_and_unload() merged_model.save_pretrained(args.output_dir,safe_serialization=True, max_shard_size="2GB")

Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]

4. Test Model and run Inference¶

After the training is done we want to evaluate and test our model. We will load different samples from the original dataset and evaluate the model on those samples, using a simple loop and accuracy as our metric.

Note: Evaluating Generative AI models is not a trivial task since 1 input can have multiple correct outputs. If you want to learn more about evaluating generative models, check out Evaluate LLMs and RAG a practical example using Langchain and Hugging Face blog post.

In [1]:

import torch
from transformers import AutoTokenizer, pipeline, AutoModelForCausalLM

model_id = "./code-llama-3-1-8b-text-to-sql"

# Load Model with PEFT adapter
model = AutoModelForCausalLM.from_pretrained(
  model_id,
  device_map="auto",
  torch_dtype=torch.float16
)
tokenizer = AutoTokenizer.from_pretrained(model_id)
# load into pipeline
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

import torch from transformers import AutoTokenizer, pipeline, AutoModelForCausalLM model_id = "./code-llama-3-1-8b-text-to-sql" # Load Model with PEFT adapter model = AutoModelForCausalLM.from_pretrained( model_id, device_map="auto", torch_dtype=torch.float16 ) tokenizer = AutoTokenizer.from_pretrained(model_id) # load into pipeline pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

/opt/conda/envs/pytorch/lib/python3.10/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: '/opt/conda/envs/pytorch/lib/python3.10/site-packages/torchvision/image.so: undefined symbol: _ZN3c1017RegisterOperatorsD1Ev'If you don't plan on using image functionality from `torchvision.io`, you can ignore this warning. Otherwise, there might be something wrong with your environment. Did you have `libjpeg` or `libpng` installed before building `torchvision` from source?
  warn(

Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]

Let’s load our test dataset try to generate an instruction.

In [2]:

from datasets import load_dataset 
from random import randint


# Load our test dataset
eval_dataset = load_dataset("json", data_files="test_dataset.json", split="train")
rand_idx = randint(0, len(eval_dataset))

# Test on sample 
prompt = pipe.tokenizer.apply_chat_template(eval_dataset[rand_idx]["messages"][:2], tokenize=False, add_generation_prompt=True)
outputs = pipe(prompt, max_new_tokens=256, do_sample=False, temperature=0.1, top_k=50, top_p=0.1, eos_token_id=pipe.tokenizer.eos_token_id, pad_token_id=pipe.tokenizer.pad_token_id)

print(f"Query:\n{eval_dataset[rand_idx]['messages'][1]['content']}")
print(f"Original Answer:\n{eval_dataset[rand_idx]['messages'][2]['content']}")
print(f"Generated Answer:\n{outputs[0]['generated_text'][len(prompt):].strip()}")

from datasets import load_dataset from random import randint # Load our test dataset eval_dataset = load_dataset("json", data_files="test_dataset.json", split="train") rand_idx = randint(0, len(eval_dataset)) # Test on sample prompt = pipe.tokenizer.apply_chat_template(eval_dataset[rand_idx]["messages"][:2], tokenize=False, add_generation_prompt=True) outputs = pipe(prompt, max_new_tokens=256, do_sample=False, temperature=0.1, top_k=50, top_p=0.1, eos_token_id=pipe.tokenizer.eos_token_id, pad_token_id=pipe.tokenizer.pad_token_id) print(f"Query:\n{eval_dataset[rand_idx]['messages'][1]['content']}") print(f"Original Answer:\n{eval_dataset[rand_idx]['messages'][2]['content']}") print(f"Generated Answer:\n{outputs[0]['generated_text'][len(prompt):].strip()}")

Generating train split: 0 examples [00:00, ? examples/s]

/opt/conda/envs/pytorch/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:567: UserWarning: `do_sample` is set to `False`. However, `temperature` is set to `0.1` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `temperature`.
  warnings.warn(
/opt/conda/envs/pytorch/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:572: UserWarning: `do_sample` is set to `False`. However, `top_p` is set to `0.1` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_p`.
  warnings.warn(

Query:
What is the Highest first elected year that has a district of 06.0 6, and a committee of economic matters?
Original Answer:
SELECT MAX(first_elected) FROM table_name_99 WHERE district = "06.0 6" AND committee = "economic matters"
Generated Answer:
SELECT MAX(first_elected) FROM table_name_99 WHERE district = "06.0 6" AND committee = "economic matters"

Nice! Our model was able to generate a SQL query based on the natural language instruction. Lets evaluate our model on the full 2,500 samples of our test dataset. Note: As mentioned above, evaluating generative models is not a trivial task. In our example we used the accuracy of the generated SQL based on the ground truth SQL query as our metric. An alternative way could be to automatically execute the generated SQL query and compare the results with the ground truth. This would be a more accurate metric but requires more work to setup.

In [3]:

from tqdm import tqdm


def evaluate(sample):
    prompt = pipe.tokenizer.apply_chat_template(sample["messages"][:2], tokenize=False, add_generation_prompt=True)
    outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95, eos_token_id=pipe.tokenizer.eos_token_id, pad_token_id=pipe.tokenizer.pad_token_id)
    predicted_answer = outputs[0]['generated_text'][len(prompt):].strip()
    if predicted_answer == sample["messages"][2]["content"]:
        return 1 
    else:
        return 0

success_rate = []
number_of_eval_samples = 1000
# iterate over eval dataset and predict
for s in tqdm(eval_dataset.shuffle().select(range(number_of_eval_samples))):
    success_rate.append(evaluate(s))

# compute accuracy
accuracy = sum(success_rate)/len(success_rate)

print(f"Accuracy: {accuracy*100:.2f}%")  
        

from tqdm import tqdm def evaluate(sample): prompt = pipe.tokenizer.apply_chat_template(sample["messages"][:2], tokenize=False, add_generation_prompt=True) outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95, eos_token_id=pipe.tokenizer.eos_token_id, pad_token_id=pipe.tokenizer.pad_token_id) predicted_answer = outputs[0]['generated_text'][len(prompt):].strip() if predicted_answer == sample["messages"][2]["content"]: return 1 else: return 0 success_rate = [] number_of_eval_samples = 1000 # iterate over eval dataset and predict for s in tqdm(eval_dataset.shuffle().select(range(number_of_eval_samples))): success_rate.append(evaluate(s)) # compute accuracy accuracy = sum(success_rate)/len(success_rate) print(f"Accuracy: {accuracy*100:.2f}%")

  1%|          | 9/1000 [00:12<21:30,  1.30s/it]You seem to be using the pipelines sequentially on GPU. In order to maximize efficiency please use a dataset
100%|██████████| 1000/1000 [24:27<00:00,  1.47s/it]

Accuracy: 80.00%

We evaluated our model on 1000 samples from the evaluation dataset and got an accuracy of 80.00%, which took ~25 minutes. This is quite good, but as mentioned you need to take this metric with a grain of salt. It would be better if we could evaluate our model by running the qureies against a real database and compare the results. Since there might be different "correct" SQL queries for the same instruction. There are also several ways on how we could improve the performance by using few-shot learning, using RAG, Self-healing to generate the SQL query.

6. Deploy the LLM for Production¶

You can now deploy your model to production. For deploying open LLMs into production we recommend using Text Generation Inference (TGI). TGI is a purpose-built solution for deploying and serving Large Language Models (LLMs). TGI enables high-performance text generation using Tensor Parallelism and continous batching for the most popular open LLMs, including Llama, Mistral, Mixtral, StarCoder, T5 and more. Text Generation Inference is used by companies as IBM, Grammarly, Uber, Deutsche Telekom, and many more. There are several ways to deploy your model, including:

• Deploy LLMs with Hugging Face Inference Endpoints
• Hugging Face LLM Inference Container for Amazon SageMaker

If you have docker installed you can use the following command to start the inference server.

Note: Make sure that you have enough GPU memory to run the container. Restart kernel to remove all allocated GPU memory from the notebook.

In [1]:

%%bash 
# model=$PWD/{args.output_dir} # path to model
model=$(pwd)/code-llama-3-1-8b-text-to-sql # path to model
num_shard=1             # number of shards
max_input_length=1024   # max input length
max_total_tokens=2048   # max total tokens

docker run -d --name tgi --gpus all -ti -p 8080:80 \
  -e MODEL_ID=/workspace \
  -e NUM_SHARD=$num_shard \
  -e MAX_INPUT_LENGTH=$max_input_length \
  -e MAX_TOTAL_TOKENS=$max_total_tokens \
  -v $model:/workspace \
  ghcr.io/huggingface/text-generation-inference:2.2.0

%%bash # model=$PWD/{args.output_dir} # path to model model=$(pwd)/code-llama-3-1-8b-text-to-sql # path to model num_shard=1 # number of shards max_input_length=1024 # max input length max_total_tokens=2048 # max total tokens docker run -d --name tgi --gpus all -ti -p 8080:80 \ -e MODEL_ID=/workspace \ -e NUM_SHARD=$num_shard \ -e MAX_INPUT_LENGTH=$max_input_length \ -e MAX_TOTAL_TOKENS=$max_total_tokens \ -v $model:/workspace \ ghcr.io/huggingface/text-generation-inference:2.2.0

Unable to find image 'ghcr.io/huggingface/text-generation-inference:2.2.0' locally
2.2.0: Pulling from huggingface/text-generation-inference
aece8493d397: Already exists
45f7ea5367fe: Already exists
3d97a47c3c73: Already exists
12cd4d19752f: Already exists
da5a484f9d74: Already exists
4f4fb700ef54: Already exists
43566b48e5d6: Already exists
f165933352a8: Already exists
f166ffc7c7b4: Already exists
58165ae83a0e: Already exists
074d930e1b90: Already exists
1033b2636622: Already exists
e0aa534acffe: Already exists
130989d28b48: Already exists
a65ea9ebfaba: Already exists
7225b2c46f88: Already exists
43154e73908f: Already exists
8f400e318724: Already exists
f694acf6c40f: Already exists
44fc79164bc4: Already exists
8bc7c142e917: Already exists
021f7d48bdcb: Already exists
c9d01f7d10cc: Already exists
400740bc31be: Already exists
bd4b49ea4512: Already exists
141228b9bdde: Already exists
4f4fb700ef54: Already exists
34d4a7457184: Already exists
66e724dff43a: Already exists
25c75c242d08: Already exists
6a4be63c7e70: Already exists
b2d83f4bca52: Already exists
373c47aa4b50: Already exists
4f4fb700ef54: Already exists
Digest: sha256:d39d513f13727ffa9b6a4d0e949f36413b944aabc9a236c0aa2986c929906769
Status: Downloaded newer image for ghcr.io/huggingface/text-generation-inference:2.2.0

42be7f00ddeb0a3214920a09a5ea303d8eb034942d7020155b6a6761fca87193

Once your container is running you can send requests using the openai or huggingface_hub sdk. Here we ll use the openai sdk to send a request to our inference server. If you don't have the openai sdk installed you can install it using pip install openai.

In [7]:

from openai import OpenAI
from datasets import load_dataset
from random import randint

# create client 
client = OpenAI(base_url="http://localhost:8080/v1",api_key="-")

# Load our test dataset
eval_dataset = load_dataset("json", data_files="test_dataset.json", split="train")
rand_idx = randint(0, len(eval_dataset))

# Take a random sample from the dataset and remove the last message and send it to the model
response = client.chat.completions.create(
	model="code-llama-3-1-8b-text-to-sql",
	messages=eval_dataset[rand_idx]["messages"][:2],
	stream=False, # no streaming
	max_tokens=1024,
)
response = response.choices[0].message.content

# Print results
print(f"Query:\n{eval_dataset[rand_idx]['messages'][1]['content']}")
print(f"Original Answer:\n{eval_dataset[rand_idx]['messages'][2]['content']}")
print(f"Generated Answer:\n{response}")

from openai import OpenAI from datasets import load_dataset from random import randint # create client client = OpenAI(base_url="http://localhost:8080/v1",api_key="-") # Load our test dataset eval_dataset = load_dataset("json", data_files="test_dataset.json", split="train") rand_idx = randint(0, len(eval_dataset)) # Take a random sample from the dataset and remove the last message and send it to the model response = client.chat.completions.create( model="code-llama-3-1-8b-text-to-sql", messages=eval_dataset[rand_idx]["messages"][:2], stream=False, # no streaming max_tokens=1024, ) response = response.choices[0].message.content # Print results print(f"Query:\n{eval_dataset[rand_idx]['messages'][1]['content']}") print(f"Original Answer:\n{eval_dataset[rand_idx]['messages'][2]['content']}") print(f"Generated Answer:\n{response}")

Query:
Name the first elected for kentucky 1
Original Answer:
SELECT first_elected FROM table_2668378_5 WHERE district = "Kentucky 1"
Generated Answer:
SELECT first_elected FROM table_2668378_5 WHERE district = "Kentucky 1"

Awesome, Don't forget to stop your container once you are done.

In [8]:

!docker stop tgi

!docker stop tgi

tgi