optimum VS FasterTransformer

Shout out to Huggingface's Optimum – which made it easier to quantize models.

Optimum to accelerate inference of transformers with hardware optimization

Yes Optimum lib's documentation is unfortunately not yet in best shape. I would be really thankful if you fill an issue detailing where the doc can be improved: https://github.com/huggingface/optimum/issues . Also, if you have features request, such as having a more flexible API, we are eager for community contributions or suggestions!

In order to support BetterTransformer with the canonical Transformer models from Transformers library, an integration was done with the open-source library Optimum as a one-liner:

It’s Lewis here from the open-source team at Hugging Face 🤗. I'm excited to share the latest release of our Optimum library, which provides a suite of performance optimization tools to make Transformers run fast on accelerated hardware!

Create libraries to optimize ML models during training and inference for specific hardware https://github.com/huggingface/optimum

Have you seen this article from HF https://huggingface.co/blog/bert-cpu-scaling-part-2 , there is also a lib https://github.com/huggingface/optimum? is the gain worth the tweaking? is OneDNN stuff easy to deploy on Triton?

https://docs.nvidia.com/ai-enterprise/overview/0.1.0/platfor...

RIVA: NVIDIA® Riva, a premium edition of NVIDIA AI Enterprise software, is a GPU-accelerated speech and translation AI SDK

FasterTransformer: https://github.com/NVIDIA/FasterTransformer an

There could be some spectrum of this expertise. For instance, https://github.com/NVIDIA/FasterTransformer, https://github.com/microsoft/DeepSpeed

Have anyone tried to optimize the forward and backward using custom Cuda code or fused kernel to speed up the training time of current LLMs? I only have seen FasterTransformer ( NVIDIA/FasterTransformer) and other similar tools but they're only focusing on inference.

Replit built Ghostwriter on the open source scene based on Salesforce’s Codegen, using Nvidia’s FasterTransformer and Triton server for highly optimized decoders, and the knowledge distillation process of the CodeGen model from two billion parameters to a faster model of one billion parameters.

Nvidia FasterTransformer is a mix of Pytorch and CUDA/C++ dedicated code. The performance boost is huge on T5, they report a 10X speedup like TensorRT. However, the speedup is computed on a translation task where sequences are 25 tokens long on average. In our experience, improvement on very short sequences tend to decrease by large margins on longer ones. Still we plan to dig deeper into this project as it implements very interesting ideas.

On the other side of the spectrum, there is Nvidia demos (here or there) showing us how to build manually a full Transformer graph (operator by operator) in TensorRT to get best performance from their hardware. It’s out of reach for many NLP practitioners and it’s time consuming to debug/maintain/adapt to a slightly different architecture (I tried). Plus, there is a secret: the very optimized model only works for specific sequence lengths and batch sizes. Truth is that, so far (and it will improve soon), it’s mainly for MLPerf benchmark (the one used to compare DL hardware), marketing content, and very specialized engineers.