• Stars
    star
    8,181
  • Rank 4,359 (Top 0.09 %)
  • Language
    Python
  • License
    Apache License 2.0
  • Created about 4 years ago
  • Updated about 1 month ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Train transformer language models with reinforcement learning.

TRL - Transformer Reinforcement Learning

Train transformer language models with reinforcement learning.

What is it?

With trl you can train transformer language models with Proximal Policy Optimization (PPO). The library is built on top of the transformers library by πŸ€— Hugging Face. Therefore, pre-trained language models can be directly loaded via transformers. At this point most of decoder architectures and encoder-decoder architectures are supported.

Highlights:

  • PPOTrainer: A PPO trainer for language models that just needs (query, response, reward) triplets to optimise the language model.
  • AutoModelForCausalLMWithValueHead & AutoModelForSeq2SeqLMWithValueHead: A transformer model with an additional scalar output for each token which can be used as a value function in reinforcement learning.
  • Example: Train GPT2 to generate positive movie reviews with a BERT sentiment classifier.

How it works

Fine-tuning a language model via PPO consists of roughly three steps:

  1. Rollout: The language model generates a response or continuation based on query which could be the start of a sentence.
  2. Evaluation: The query and response are evaluated with a function, model, human feedback or some combination of them. The important thing is that this process should yield a scalar value for each query/response pair.
  3. Optimization: This is the most complex part. In the optimisation step the query/response pairs are used to calculate the log-probabilities of the tokens in the sequences. This is done with the model that is trained and and a reference model, which is usually the pre-trained model before fine-tuning. The KL-divergence between the two outputs is used as an additional reward signal to make sure the generated responses don't deviate to far from the reference language model. The active language model is then trained with PPO.

This process is illustrated in the sketch below:

Figure: Sketch of the workflow.

Installation

Python package

Install the library with pip:

pip install trl

From source

If you want to run the examples in the repository a few additional libraries are required. Clone the repository and install it with pip:

git clone https://github.com/lvwerra/trl.git
cd trl/
pip install .

If you wish to develop TRL, you should install in editable mode:

pip install -e .

How to use

Example

This is a basic example on how to use the library. Based on a query the language model creates a response which is then evaluated. The evaluation could be a human in the loop or another model's output.

# imports
import torch
from transformers import AutoTokenizer
from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead, create_reference_model
from trl.core import respond_to_batch

# get models
model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
model_ref = create_reference_model(model)

tokenizer = AutoTokenizer.from_pretrained('gpt2')

# initialize trainer
ppo_config = PPOConfig(
    batch_size=1,
)

# encode a query
query_txt = "This morning I went to the "
query_tensor = tokenizer.encode(query_txt, return_tensors="pt")

# get model response
response_tensor  = respond_to_batch(model, query_tensor)

# create a ppo trainer
ppo_trainer = PPOTrainer(ppo_config, model, model_ref, tokenizer)

# define a reward for response
# (this could be any reward such as human feedback or output from another model)
reward = [torch.tensor(1.0)]

# train model for one step with ppo
train_stats = ppo_trainer.step([query_tensor[0]], [response_tensor[0]], reward)

Advanced example: IMDB sentiment

For a detailed example check out the example python script examples/sentiment/scripts/gpt2-sentiment.py, where GPT2 is fine-tuned to generate positive movie reviews. An few examples from the language models before and after optimisation are given below:

Figure: A few review continuations before and after optimisation.

References

Proximal Policy Optimisation

The PPO implementation largely follows the structure introduced in the paper "Fine-Tuning Language Models from Human Preferences" by D. Ziegler et al. [paper, code].

Language models

The language models utilize the transformers library by πŸ€— Hugging Face.

Citation

@misc{vonwerra2022trl,
  author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert},
  title = {TRL: Transformer Reinforcement Learning},
  year = {2020},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/lvwerra/trl}}
}

More Repositories

1

transformers

πŸ€— Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Python
126,555
star
2

pytorch-image-models

PyTorch image models, scripts, pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (ViT), MobileNet-V3/V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more
Python
28,073
star
3

diffusers

πŸ€— Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch and FLAX.
Python
23,282
star
4

datasets

πŸ€— The largest hub of ready-to-use datasets for ML models with fast, easy-to-use and efficient data manipulation tools
Python
17,530
star
5

peft

πŸ€— PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.
Python
14,585
star
6

candle

Minimalist ML framework for Rust
Rust
14,045
star
7

tokenizers

πŸ’₯ Fast State-of-the-Art Tokenizers optimized for Research and Production
Rust
8,560
star
8

text-generation-inference

Large Language Model Text Generation Inference
Python
8,197
star
9

accelerate

πŸš€ A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support
Python
7,121
star
10

chat-ui

Open source codebase powering the HuggingChat app
TypeScript
6,584
star
11

alignment-handbook

Robust recipes to align language models with human and AI preferences
Python
4,079
star
12

deep-rl-class

This repo contains the syllabus of the Hugging Face Deep Reinforcement Learning Course.
MDX
3,680
star
13

lerobot

πŸ€— LeRobot: End-to-end Learning for Real-World Robotics in Pytorch
Python
3,558
star
14

notebooks

Notebooks using the Hugging Face libraries πŸ€—
Jupyter Notebook
3,329
star
15

distil-whisper

Distilled variant of Whisper for speech recognition. 6x faster, 50% smaller, within 1% word error rate.
Python
3,286
star
16

autotrain-advanced

πŸ€— AutoTrain Advanced
Python
3,283
star
17

diffusion-models-class

Materials for the Hugging Face Diffusion Models Course
Jupyter Notebook
3,280
star
18

neuralcoref

✨Fast Coreference Resolution in spaCy with Neural Networks
C
2,819
star
19

parler-tts

Inference and training library for high-quality TTS models.
Python
2,735
star
20

knockknock

πŸšͺ✊Knock Knock: Get notified when your training ends with only two additional lines of code
Python
2,682
star
21

safetensors

Simple, safe way to store and distribute tensors
Python
2,538
star
22

swift-coreml-diffusers

Swift app demonstrating Core ML Stable Diffusion
Swift
2,406
star
23

text-embeddings-inference

A blazing fast inference solution for text embeddings models
Rust
2,148
star
24

optimum

πŸš€ Accelerate training and inference of πŸ€— Transformers and πŸ€— Diffusers with easy to use hardware optimization tools
Python
2,086
star
25

setfit

Efficient few-shot learning with Sentence Transformers
Jupyter Notebook
2,044
star
26

course

The Hugging Face course on Transformers
MDX
2,005
star
27

awesome-papers

Papers & presentation materials from Hugging Face's internal science day
1,996
star
28

blog

Public repo for HF blog posts
Jupyter Notebook
1,962
star
29

evaluate

πŸ€— Evaluate: A library for easily evaluating machine learning models and datasets.
Python
1,825
star
30

transfer-learning-conv-ai

πŸ¦„ State-of-the-Art Conversational AI with Transfer Learning
Python
1,654
star
31

datatrove

Freeing data processing from scripting madness by providing a set of platform-agnostic customizable pipeline processing blocks.
Python
1,635
star
32

swift-coreml-transformers

Swift Core ML 3 implementations of GPT-2, DistilGPT-2, BERT, and DistilBERT for Question answering. Other Transformers coming soon!
Swift
1,543
star
33

pytorch-openai-transformer-lm

πŸ₯A PyTorch implementation of OpenAI's finetuned transformer language model with a script to import the weights pre-trained by OpenAI
Python
1,464
star
34

cookbook

Open-source AI cookbook
Jupyter Notebook
1,416
star
35

huggingface_hub

All the open source things related to the Hugging Face Hub.
Python
1,311
star
36

Mongoku

πŸ”₯The Web-scale GUI for MongoDB
TypeScript
1,289
star
37

huggingface.js

Utilities to use the Hugging Face Hub API
TypeScript
1,251
star
38

gsplat.js

JavaScript Gaussian Splatting library.
TypeScript
1,233
star
39

hmtl

🌊HMTL: Hierarchical Multi-Task Learning - A State-of-the-Art neural network model for several NLP tasks based on PyTorch and AllenNLP
Python
1,185
star
40

llm-vscode

LLM powered development for VSCode
TypeScript
1,148
star
41

pytorch-pretrained-BigGAN

πŸ¦‹A PyTorch implementation of BigGAN with pretrained weights and conversion scripts.
Python
986
star
42

nanotron

Minimalistic large language model 3D-parallelism training
Python
897
star
43

torchMoji

πŸ˜‡A pyTorch implementation of the DeepMoji model: state-of-the-art deep learning model for analyzing sentiment, emotion, sarcasm etc
Python
880
star
44

optimum-nvidia

Python
825
star
45

awesome-huggingface

πŸ€— A list of wonderful open-source projects & applications integrated with Hugging Face libraries.
821
star
46

naacl_transfer_learning_tutorial

Repository of code for the tutorial on Transfer Learning in NLP held at NAACL 2019 in Minneapolis, MN, USA
Python
718
star
47

dataset-viewer

Lightweight web API for visualizing and exploring any dataset - computer vision, speech, text, and tabular - stored on the Hugging Face Hub
Python
630
star
48

optimum-quanto

A pytorch quantization backend for optimum
Python
616
star
49

llm.nvim

LLM powered development for Neovim
Lua
590
star
50

exporters

Export Hugging Face models to Core ML and TensorFlow Lite
Python
559
star
51

transformers-bloom-inference

Fast Inference Solutions for BLOOM
Python
550
star
52

pytorch_block_sparse

Fast Block Sparse Matrices for Pytorch
C++
523
star
53

swift-transformers

Swift Package to implement a transformers-like API in Swift
Swift
504
star
54

llm-ls

LSP server leveraging LLMs for code completion (and more?)
Rust
498
star
55

node-question-answering

Fast and production-ready question answering in Node.js
TypeScript
459
star
56

large_language_model_training_playbook

An open collection of implementation tips, tricks and resources for training large language models
Python
441
star
57

lighteval

LightEval is a lightweight LLM evaluation suite that Hugging Face has been using internally with the recently released LLM data processing library datatrove and LLM training library nanotron.
Python
440
star
58

llm_training_handbook

An open collection of methodologies to help with successful training of large language models.
Python
416
star
59

swift-chat

Mac app to demonstrate swift-transformers
Swift
392
star
60

ratchet

A cross-platform browser ML framework.
Rust
390
star
61

tflite-android-transformers

DistilBERT / GPT-2 for on-device inference thanks to TensorFlow Lite with Android demo apps
Java
368
star
62

community-events

Place where folks can contribute to πŸ€— community events
Jupyter Notebook
368
star
63

text-clustering

Easily embed, cluster and semantically label text datasets
Python
367
star
64

nn_pruning

Prune a model while finetuning or training.
Jupyter Notebook
360
star
65

optimum-intel

πŸ€— Optimum Intel: Accelerate inference with Intel optimization tools
Jupyter Notebook
342
star
66

speechbox

Python
339
star
67

controlnet_aux

Python
326
star
68

100-times-faster-nlp

πŸš€100 Times Faster Natural Language Processing in Python - iPython notebook
HTML
325
star
69

education-toolkit

Educational materials for universities
Jupyter Notebook
320
star
70

unity-api

C#
302
star
71

datablations

Scaling Data-Constrained Language Models
Jupyter Notebook
296
star
72

open-muse

Open reproduction of MUSE for fast text2image generation.
Python
293
star
73

audio-transformers-course

The Hugging Face Course on Transformers for Audio
MDX
280
star
74

cosmopedia

Python
280
star
75

hf_transfer

Rust
238
star
76

hub-docs

Docs of the Hugging Face Hub
221
star
77

dataspeech

Python
207
star
78

optimum-benchmark

A unified multi-backend utility for benchmarking Transformers, Timm, PEFT, Diffusers and Sentence-Transformers with full support of Optimum's hardware optimizations & quantization schemes.
Python
206
star
79

diarizers

Python
206
star
80

simulate

🎒 Creating and sharing simulation environments for embodied and synthetic data research
Python
185
star
81

instruction-tuned-sd

Code for instruction-tuning Stable Diffusion.
Python
181
star
82

llm-swarm

Manage scalable open LLM inference endpoints in Slurm clusters
Python
176
star
83

optimum-neuron

Easy, fast and very cheap training and inference on AWS Trainium and Inferentia chips.
Jupyter Notebook
173
star
84

olm-datasets

Pipeline for pulling and processing online language model pretraining data from the web
Python
169
star
85

data-is-better-together

Let's build better datasets, together!
Jupyter Notebook
162
star
86

OBELICS

Code used for the creation of OBELICS, an open, massive and curated collection of interleaved image-text web documents, containing 141M documents, 115B text tokens and 353M images.
Python
159
star
87

diffusion-fast

Faster generation with text-to-image diffusion models.
Python
157
star
88

workshops

Materials for workshops on the Hugging Face ecosystem
Jupyter Notebook
146
star
89

api-inference-community

Python
145
star
90

jat

Distributed online training of a general multi-task Deep RL Agent
Python
136
star
91

chug

Minimal sharded dataset loaders, decoders, and utils for multi-modal document, image, and text datasets.
Python
136
star
92

sharp-transformers

A Unity plugin for using Transformers models in Unity.
C#
129
star
93

optimum-habana

Easy and lightning fast training of πŸ€— Transformers on Habana Gaudi processor (HPU)
Python
114
star
94

hf-hub

Rust client for the huggingface hub aiming for minimal subset of features over `huggingface-hub` python package
Rust
109
star
95

frp

FRP Fork
Go
102
star
96

competitions

Python
101
star
97

olm-training

Repo for training MLMs, CLMs, or T5-type models on the OLM pretraining data, but it should work with any hugging face text dataset.
Python
92
star
98

fuego

[WIP] A πŸ”₯ interface for running code in the cloud
Python
85
star
99

tune

Python
83
star
100

datasets-viewer

Viewer for the πŸ€— datasets library.
Python
83
star