DeepSpeed
Finetune_LLMs
Our great sponsors
- ONLYOFFICE ONLYOFFICE Docs — document collaboration in your environment
- CodiumAI - TestGPT | Generating meaningful tests for busy devs
- Sonar - Write Clean Python Code. Always.
- InfluxDB - Access the most powerful time series database as a service
| DeepSpeed | Finetune_LLMs | |
|---|---|---|
| 41 | 2 | |
| 25,088 | 304 | |
| 61.0% | - | |
| 9.6 | 0.0 | |
| 2 days ago | 26 days ago | |
| Python | Python | |
| Apache License 2.0 | GNU Affero General Public License v3.0 |
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.
DeepSpeed
-
Using --deepspeed requires lots of manual tweaking
Filed a discussion item on the deepspeed project: https://github.com/microsoft/DeepSpeed/discussions/3531
Solution: I don't know; this is where I am stuck. https://github.com/microsoft/DeepSpeed/issues/1037 suggests that I just need to 'apt install libaio-dev', but I've done that and it doesn't help.
-
Whether the ML computation engineering expertise will be valuable, is the question.
There could be some spectrum of this expertise. For instance, https://github.com/NVIDIA/FasterTransformer, https://github.com/microsoft/DeepSpeed
- FLiPN-FLaNK Stack Weekly for 17 April 2023
- DeepSpeed Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-Like Models
- DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-Like Models
-
12-Apr-2023 AI Summary
DeepSpeed Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales (https://github.com/microsoft/DeepSpeed/tree/master/blogs/deepspeed-chat)
- Microsoft DeepSpeed
-
Apple: Transformer architecture optimized for Apple Silicon
I'm following this closely, together with other efforts like GPTQ Quantization and Microsoft's DeepSpeed, all of which are bringing down the hardware requirements of these advanced AI models.
-
Facebook LLAMA is being openly distributed via torrents
- https://github.com/microsoft/DeepSpeed
Anything that could bring this to a 10GB 3080 or 24GB 3090 without 60s/it per token?
Finetune_LLMs
-
[D] Fine-tuning GPT-J: lessons learned
And this: https://github.com/mallorbc/Finetune_GPTNEO_GPTJ6B
What are some alternatives?
ColossalAI - Making large AI models cheaper, faster and more accessible
fairscale - PyTorch extensions for high performance and large scale training.
TensorRT - NVIDIA® TensorRT™, an SDK for high-performance deep learning inference, includes a deep learning inference optimizer and runtime that delivers low latency and high throughput for inference applications.
Megatron-LM - Ongoing research training transformer models at scale
fairseq - Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
mesh-transformer-jax - Model parallel transformers in JAX and Haiku
llama - Inference code for LLaMA models
gpt-neox - An implementation of model parallel autoregressive transformers on GPUs, based on the DeepSpeed library.
server - The Triton Inference Server provides an optimized cloud and edge inferencing solution.
text-generation-webui - A gradio web UI for running Large Language Models like LLaMA, llama.cpp, GPT-J, Pythia, OPT, and GALACTICA.
Pytorch - Tensors and Dynamic neural networks in Python with strong GPU acceleration
flash-attention - Fast and memory-efficient exact attention