transformers
fairseq
Our great sponsors
transformers | fairseq | |
---|---|---|
175 | 89 | |
124,557 | 29,205 | |
2.7% | 1.6% | |
10.0 | 6.6 | |
7 days ago | 6 days ago | |
Python | Python | |
Apache License 2.0 | MIT License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
transformers
-
Maxtext: A simple, performant and scalable Jax LLM
Is t5x an encoder/decoder architecture?
Some more general options.
The Flax ecosystem
https://github.com/google/flax?tab=readme-ov-file
or dm-haiku
https://github.com/google-deepmind/dm-haiku
were some of the best developed communities in the Jax AI field
Perhaps the “trax” repo? https://github.com/google/trax
Some HF examples https://github.com/huggingface/transformers/tree/main/exampl...
Sadly it seems much of the work is proprietary these days, but one example could be Grok-1, if you customize the details. https://github.com/xai-org/grok-1/blob/main/run.py
-
Lossless Acceleration of LLM via Adaptive N-Gram Parallel Decoding
The HuggingFace transformers library already has support for a similar method called prompt lookup decoding that uses the existing context to generate an ngram model: https://github.com/huggingface/transformers/issues/27722
I don't think it would be that hard to switch it out for a pretrained ngram model.
-
AI enthusiasm #6 - Finetune any LLM you want💡
Most of this tutorial is based on Hugging Face course about Transformers and on Niels Rogge's Transformers tutorials: make sure to check their work and give them a star on GitHub, if you please ❤️
-
Schedule-Free Learning – A New Way to Train
* Superconvergence + LR range finder + Fast AI's Ranger21 optimizer was the goto optimizer for CNNs, and worked fabulously well, but on transformers, the learning rate range finder sadi 1e-3 was the best, whilst 1e-5 was better. However, the 1 cycle learning rate stuck. https://github.com/huggingface/transformers/issues/16013
-
Gemma doesn't suck anymore – 8 bug fixes
Thanks! :) I'm pushing them into transformers, pytorch-gemma and collabing with the Gemma team to resolve all the issues :)
The RoPE fix should already be in transformers 4.38.2: https://github.com/huggingface/transformers/pull/29285
My main PR for transformers which fixes most of the issues (some still left): https://github.com/huggingface/transformers/pull/29402
- HuggingFace Transformers: Qwen2
- HuggingFace Transformers Release v4.36: Mixtral, Llava/BakLlava, SeamlessM4T v2
- HuggingFace: Support for the Mixtral Moe
-
Paris-Based Startup and OpenAI Competitor Mistral AI Valued at $2B
If you want to tinker with the architecture Hugging Face has a FOSS implementation in transformers: https://github.com/huggingface/transformers/blob/main/src/tr...
If you want to reproduce the training pipeline, you couldn't do that even if you wanted to because you don't have access to thousands of A100s.
-
Fail to reproduce the same evaluation metrics score during inference.
I am aware that using mixed precision reduces the stability of weight and there will be little consistency but don't expect it to be this much. I have attached the graph of evaluation metrics. If someone can give me some insight into this issue, that would be great.
fairseq
- Sequence-to-Sequence Toolkit Written in Python
- Unsupervised (Semi-Supervised) ASR/STT training recipes
-
Nvidia's 900 tons of GPU muscle bulks up server market, slims down wallets
> Is there really no way to partition the workload to run with 16gb memory per card?
It really depends and this can get really complicated really fast. I'll give a tldr and then a longer explanation.
TLDR:
Yes, you can easily split networks up. If your main bottleneck is batch size (i.e. training) then there aren't huge differences in spreading across multiple GPUs assuming you have good interconnects (GPU direct is supported). If you're running inference and the model fits on the card you're probably fine too unless you need to do things like fancy inference batching (i.e. you have LOTS of users)
Longer version:
You can always split things up. If we think about networks we recognize some nice properties about how they operate as mathematical groups. Non-residual networks are compositional, meaning each layer can be treated as a sub network (every residual block can be treated this way too). Additionally, we may have associative and distributive properties depending on the architecture (some even have commutative!). So we can use these same rules to break apart networks in many different ways. There are often performance hits for doing this though, as it practically requires you touching the disk more often but in some more rare cases (at least to me, let me know if you know more) they can help.
I mentioned the batching above and this can get kinda complicated. There are actually performance differences when you batch in groups of data (i.e. across GPUs) compared to batching on a single accelerator. This difference isn't talked about a lot. But it is going to come down to how often your algorithm depends on batching and what operations are used, such as batch norm. The batch norm is calculated across the GPU's batch, not the distributed batch (unless you introduce blocking). This is because your gradients AND inference are going to be computed differently. In DDP your whole network is cloned across cards so you basically run inference on multiple networks and then do an all reduce on the loss then calculate the gradient and then recopy the weights to all cards. There is even a bigger difference when you use lazy regularization (don't compute gradients for n-minibatches). GANs are notorious for using this and personally I've seen large benefits to distributed training for these. GANs usually have small batch sizes and aren't getting anywhere near the memory of the card anyways (GANs are typically unstable so large batch sizes can harm them), but also pay attention to this when evaluating papers (of course as well as how much hyper-parameter tuning has been done. This is always tricky when comparing works, especially between academia and big labs. You can easily be fooled by which is a better model. Evaluating models is way tougher than people give credit to and especially in the modern era of LLMs. I could rant a lot about just this alone). Basically in short, we can think of this as an ensembling method, except our models are actually identical (you could parallel reduce lazily too and that will create some periodic divergence between your models but that's not important for conceptually understanding, just worth noting).
There is are also techniques to split a single model up called model sharding and checkpointing. Model sharding is where you split a single model across multiple GPUs. You're taking advantage of the compositional property of networks, meaning that as long as there isn't a residual layer between your split location you can actually treat one network as a series of smaller networks. This has obvious drawbacks as you need to feed one into another and so the operations have to be synchronous, but sometimes this isn't too bad. Checkpointing is very similar but you're just doing the same thing on the same GPU. Your hit here is in I/O, but may or may not be too bad with GPU Direct and highly depends on your model size (were you splitting because batch size or because model size?).
This is all still pretty high level but if you want to dig into it more META developed a toolkit called fairseq that will do a lot of this for you and they optimized it
https://engineering.fb.com/2021/07/15/open-source/fsdp/
https://github.com/facebookresearch/fairseq
TLDR: really depends on your use case, but it is a good question.
-
Talk back and forth with AI like you would with a person
How do they do the text to voice conversion so fast? https://github.com/facebookresearch/fairseq/tree/main (open source takes sub-minute to do text to voice.
-
Voice generation AI (TTS)
It might be worth checking out Meta's TTS tho, I haven't gotten the chance to fiddle around with it but it looks somewhat promising https://github.com/facebookresearch/fairseq/tree/main/examples/mms
-
Translation app with TTS (text-to-speech) for Persian?
They have instructions on how to use it in command line and a notebook on how to use it as a python library.
- Why no work on open source TTS (Text to speech) models
-
Meta's Massively Multilingual Speech project supports 1k languages using self supervised learning
Github - https://github.com/facebookresearch/fairseq/tree/main/examples/mms Paper - https://research.facebook.com/publications/scaling-speech-technology-to-1000-languages/
-
AI — weekly megathread!
Meta released a new open-source model, Massively Multilingual Speech (MMS) that can do both speech-to-text and text-to-speech in 1,107 languages and can also recognize 4,000+ spoken languages. Existing speech recognition models only cover approximately 100 languages out of the 7,000+ known spoken languages. [Details | Research Paper | GitHub].
- Meta's MMS: Scaling Speech Technology to 1000+ languages (How to Run colab)
What are some alternatives?
sentence-transformers - Multilingual Sentence & Image Embeddings with BERT
gpt-neox - An implementation of model parallel autoregressive transformers on GPUs, based on the DeepSpeed library.
llama - Inference code for Llama models
DeepSpeed - DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.
transformer-pytorch - Transformer: PyTorch Implementation of "Attention Is All You Need"
text-to-text-transfer-transformer - Code for the paper "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer"
text-generation-webui - A Gradio web UI for Large Language Models. Supports transformers, GPTQ, AWQ, EXL2, llama.cpp (GGUF), Llama models.
espnet - End-to-End Speech Processing Toolkit
huggingface_hub - The official Python client for the Huggingface Hub.
Pytorch - Tensors and Dynamic neural networks in Python with strong GPU acceleration
OpenNMT-py - Open Source Neural Machine Translation and (Large) Language Models in PyTorch
taro - 开放式跨端跨框架解决方案,支持使用 React/Vue/Nerv 等框架来开发微信/京东/百度/支付宝/字节跳动/ QQ 小程序/H5/React Native 等应用。 https://taro.zone/