deepspeed
1title: DeepSpeed
2email: deepspeed@microsoft.com
3description: >-
4DeepSpeed is a deep learning optimization library that makes distributed
5training easy, efficient, and effective.
6
7locale : "en-US"
8logo: /assets/images/deepspeed-logo-uppercase-bold-white-1.15.svg
9
10repository: microsoft/DeepSpeed
11baseurl: "/" # the subpath of your site, e.g. /blog
12url: "https://www.deepspeed.ai" # the base hostname & protocol for your site, e.g. http://example.com
13
14# Build settings
15remote_theme: "mmistakes/minimal-mistakes@4.19.0"
16minimal_mistakes_skin : "air"
17search: true
18
19plugins:
20- jekyll-feed
21- jekyll-include-cache
22- jekyll-paginate
23
24#paginate: 10
25#paginate_path: /blog/page:num
26
27include: ["_pages"]
28exclude: ["code-docs"]
29
30collections:
31tutorials:
32output: true
33permalink: /:collection/:path/
34order:
35- advanced-install.md
36- getting-started.md
37- azure.md
38- automatic-tensor-parallelism.md
39- bert-finetuning.md
40- bert-pretraining.md
41- cifar-10.md
42- curriculum-learning.md
43- data-efficiency.md
44- ds4sci_evoformerattention.md
45- flops-profiler.md
46- pytorch-profiler.md
47- autotuning.md
48- gan.md
49- lrrt.md
50- megatron.md
51- mixture-of-experts.md
52- mixture-of-experts-nlg.md
53- mixture-of-experts-inference.md
54- model-compression.md
55- monitor.md
56- comms-logging.md
57- one-cycle.md
58- onebit-adam.md
59- zero-one-adam.md
60- onebit-lamb.md
61- pipeline.md
62- progressive_layer_dropping.md
63- sparse-attention.md
64- transformer_kernel.md
65- zero-offload.md
66- zero.md
67
68defaults:
69- scope:
70path: ""
71values:
72layout: single
73author_profile: false
74read_time: false
75comments: false
76share: false
77related: false
78sneak_preview: false
79toc: true
80toc_label: "Contents"
81sidebar:
82nav: "lnav"
83- scope:
84path: "_pages"
85values:
86permalink: /docs/:basename/
87toc: true
88toc_label: "Contents"
89- scope:
90path: ""
91type: posts
92values:
93layout: single-full
94author_profile: false
95read_time: false
96comments: false
97share: true
98related: false
99toc: true
100toc_label: "Contents"
101toc_sticky: true
102show_date: true
103- scope:
104path: ""
105type: tutorials
106values:
107layout: single
108toc_sticky: true
109
110
111analytics:
112provider: "google-gtag"
113google:
114tracking_id: "UA-169781858-1"
115
116timezone: America/Los_Angeles
117breadcrumbs: true
118
119press_release_v3: https://www.microsoft.com/en-us/research/blog/deepspeed-extreme-scale-model-training-for-everyone/
120press_release_v5: https://www.microsoft.com/en-us/research/blog/deepspeed-powers-8x-larger-moe-model-training-with-high-performance/
121press_release_v6: https://www.microsoft.com/en-us/research/blog/deepspeed-advancing-moe-inference-and-training-to-power-next-generation-ai-scale/
122