12 results found Sort:

hpcaitech
ColossalAI hpcaitech
4.3k
38.1k
apache-2.0
379
Making large AI models cheaper, faster and more accessible
ai hpc big-model inference large-scale deep-learning data-parallelism foundation-models model-parallelism pipeline-parallelism distributed-computing heterogeneous-training
Created 2021-10-28
3,357 commits to main branch, last one 21 hours ago
microsoft
DeepSpeed microsoft
3.9k
33.2k
apache-2.0
335
DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.
gpu zero pytorch inference compression deep-learning data-parallelism machine-learning model-parallelism billion-parameters mixture-of-experts trillion-parameters pipeline-parallelism
Created 2020-01-23
2,316 commits to master branch, last one 11 hours ago
bigscience-workshop
petals bigscience-workshop
474
8.8k
mit
88
🌸 Run LLMs at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading
gpt nlp bloom llama falcon llama2 chatbot guanaco pytorch transformer deep-learning language-models neural-networks machine-learning pretrained-models tensor-parallelism distributed-systems volunteer-computing pipeline-parallelism large-language-models
Created 2022-06-12
508 commits to main branch, last one about a month ago
PaddlePaddle
PaddleFleetX PaddlePaddle
158
425
apache-2.0
23
飞桨大模型开发套件，提供大语言模型、跨模态大模型、生物计算大模型等领域的全流程开发工具链。
cloud elastic benchmark fleet-api lightning large-scale paddlecloud pretraining paddlepaddle data-parallelism model-parallelism pipeline-parallelism distributed-algorithm unsupervised-learning self-supervised-learning
Created 2018-12-12
440 commits to develop branch, last one 9 months ago
Oneflow-Inc
libai Oneflow-Inc
55
376
apache-2.0
43
LiBai(李白): A Toolbox for Large-Scale Distributed Parallel Training
nlp oneflow large-scale transformer deep-learning data-parallelism model-parallelism vision-transformer distributed-training pipeline-parallelism self-supervised-learning
Created 2021-10-25
348 commits to main branch, last one about a month ago
Coobiw
MiniGPT4Qwen Coobiw
14
260
unknown
4
Personal Project: MPP-Qwen14B(Multimodal Pipeline Parallel-Qwen14B). Don't let the poverty limit your imagination! Train your own 14B LLaVA-like MLLM on RTX3090/4090 24GB.
mllm qwen deepspeed fine-tuning pretraining model-parallel pipeline-parallelism multimodal-large-language-models
Created 2023-10-24
85 commits to master branch, last one a day ago
alibaba
EasyParallelLibrary alibaba
49
253
apache-2.0
13
Easy Parallel Library (EPL) is a general and efficient deep learning framework for distributed model training.
gpu deep-learning data-parallelism memory-efficient model-parallelism distributed-training pipeline-parallelism
Created 2022-02-23
21 commits to main branch, last one about a year ago
Shenggan
awesome-distributed-ml Shenggan
19
165
unknown
8
A curated list of awesome projects and papers for distributed training or inference
deep-learning machine-learning model-parallelism distributed-systems pipeline-parallelism high-performance-computing
Created 2022-04-22
29 commits to main branch, last one 2 months ago
torchpipe
torchpipe torchpipe
12
130
apache-2.0
6
An Alternative for Triton Inference Server. Boosting DL Service Throughput 1.5-4x by Ensemble Pipeline Serving with Concurrent CUDA Streams for PyTorch/LibTorch Frontend and TensorRT/CVCUDA, etc., Bac...
ray serve cvcuda pytorch serving tensorrt inference ray-serve torch2trt deployment pipeline-parallelism triton-inference-server
Created 2023-10-24
140 commits to main branch, last one 9 days ago
xrsrke
pipegoose xrsrke
17
74
mit
4
Large scale 4D parallelism pre-training for 🤗 transformers in Mixture of Experts *(still work in progress)*
moe zero-1 megatron megatron-lm transformers 3d-parallelism data-parallelism model-parallelism mixture-of-experts tensor-parallelism pipeline-parallelism sequence-parallelism distributed-optimizers huggingface-transformers large-scale-language-modeling
Created 2023-06-14
554 commits to main branch, last one 5 months ago
saareliad
FTPipe saareliad
7
41
unknown
4
FTPipe and related pipeline model parallelism research.
t5 nlp fine-tuning deep-neural-networks distributed-training pipeline-parallelism
Created 2020-06-29
2,372 commits to master branch, last one about a year ago
Shigangli
Chimera Shigangli
6
37
gpl-3.0
2
Chimera: Efficiently Training Large-Scale Neural Networks with Bidirectional Pipelines.
transformers pipeline-parallelism distributed-deep-learning
Created 2021-05-30
16 commits to main branch, last one 5 months ago