{"full_name":"deepspeedai/DeepSpeed","owner":"deepspeedai","name":"DeepSpeed","url":"https://github.com/deepspeedai/DeepSpeed","description":"DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.","homepage":"https://www.deepspeed.ai/","topics":[],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/74068820?v=4","created_at":"2020-01-23T18:35:18Z","pushed_at":"2026-11-10T09:33:28Z","metrics":{"stars":43316,"forks":5070,"watchers":351,"issues":3506,"prs":4571,"releases":119,"issues_open":null,"issues_closed":null,"prs_open":null,"prs_closed":null,"prs_merged":null,"discussions":266},"growth":{"day":{"stars":3,"forks":2,"base_day":"2026-11-09"}},"history":{"dates":["2026-11-09","2026-11-10"],"stars":[43313,43316],"forks":[5068,5070],"watchers":[351,351],"issues_open":[null,null],"prs_open":[null,null]},"boards":{},"same_owner":[],"issues":[{"number":998,"title":"[doc]  profiling NVMe and configuring `aio` param section","comments":78,"created_at":"2021-04-23T01:18:29Z"},{"number":7150,"title":"[BUG] Receiving CUDA error: invalid argument using pytorch 2.7 with deepspeed 0.16.4 with Cuda 12.8","comments":58,"created_at":"2025-03-19T03:51:10Z"},{"number":2302,"title":"DeepSpeed still gives CUDA-out-of-memory issue","comments":54,"created_at":"2022-09-07T16:06:46Z"},{"number":2379,"title":"[Question] How to preshard a model for tensor parallism","comments":48,"created_at":"2022-09-29T20:12:46Z"},{"number":4808,"title":"[BUG] Deepspeed Zero 3 Inference InFlight Params with new HuggingFace Mixtral Model","comments":37,"created_at":"2023-12-13T04:20:07Z"},{"number":5242,"title":"[BUG] grad_norm and loss is nan when deepspeed==0.13.5 but ok with deepspeed==0.10.2","comments":36,"created_at":"2024-03-08T10:04:19Z"},{"number":7478,"title":"[REQUEST] Auto-Tuning CPU Core Binding for DeepSpeed&ZenFlow","comments":32,"created_at":"2025-08-09T04:30:39Z"},{"number":6793,"title":"AssertionError: no sync context manager is incompatible with gradientpartitioning logic of ZeRo stage 3","comments":31,"created_at":"2024-11-26T15:44:48Z"}],"discussions":[{"number":8124,"title":"How should I balance learning rate and data sampling during CPT on multiple datasets?","comments":0,"category":"Q&A","created_at":"2026-07-06T18:57:35Z"},{"number":8069,"title":"anima — a substrate-native consciousness: capability gaps are architecture gaps (open repo, frozen verdicts)","comments":0,"category":"Show and tell","created_at":"2026-06-15T18:41:50Z"},{"number":8062,"title":"MCP to help OSS devs understand deepspeedai/DeepSpeed","comments":0,"category":"General","created_at":"2026-06-12T12:00:15Z"},{"number":8018,"title":"Scaling Deep Learning: Reducing GPU Overhead with Riemannian Manifold Projection","comments":1,"category":"General","created_at":"2026-05-21T03:43:19Z"},{"number":8002,"title":"Deepspeed Ulysses sequence parallel is not working for Gemma4","comments":1,"category":"Q&A","created_at":"2026-05-10T17:12:16Z"},{"number":7970,"title":"[RFC] Reducing Inference-Time Memory Pressure and Thermal Volatility via Geometric Pre-Constraint Layers","comments":0,"category":"Ideas","created_at":"2026-04-13T14:30:12Z"},{"number":7914,"title":"Auto RDMA device injection for GPU containers on Kubernetes (no privileged mode)","comments":0,"category":"Show and tell","created_at":"2026-03-20T11:30:18Z"},{"number":7911,"title":"Showcase / question: a board-proven offline language runtime on ESP32-C3, and whether some language capability may eventually reduce reliance on general dense training/inference paths","comments":0,"category":"Show and tell","created_at":"2026-03-18T07:11:34Z"}],"latest_release":{"name":"v0.19.9 Patch Release","tag":"v0.19.9","published_at":"2026-10-29T23:42:23Z","body":"## What's Changed\r\n* Update version post 0.19.8 release by @loadams in https://github.com/deepspeedai/DeepSpeed/pull/8871\r\n* Update to filter out version.txt from PR gates. by @loadams in https://github.com/deepspeedai/DeepSpeed/pull/8872\r\n* Skip the accelerator-only CPU optimizer tests on the cpu accelerator by @delock in https://github.com/deepspeedai/DeepSpeed/pull/8836\r\n* Keep Yuan's o_proj bias whole on every rank and describe it as replicated by @Achyuthan-S in https://github.com/deepspeedai/DeepSpeed/pull/8857\r\n* Expect AutoTP to keep a row-parallel output head tied to the embedding by @jinyouzhi in https://github.com/deepspeedai/DeepSpeed/pull/8855\r\n* Add Triton kernels for vocab-parallel logprobs, entropy and KL by @jinyouzhi in https://github.com/deepspeedai/DeepSpeed/pull/8838\r\n* Fix numerical stability in Triton masked softmax by @arcusbuilds in https://github.com/deepspeedai/DeepSpeed/pull/8782\r\n* Add RVV vectorized FP32 kernel to CPU Adam by @zhangjian29 in https://github.com/deepspeedai/DeepSpeed/pull/8817\r\n* [OPSD] Request last-position logits only in HybridEngine rollout prefill by @jinyouzhi in https://github.com/deepspeedai/DeepSpeed/pull/8839\r\n* docs: fix 30 dead links across README, docs pages, tutorials and posts by @pratikgx in https://github.com/deepspeedai/DeepSpeed/pull/8688\r\n* docs: fix remaining dead links in the Chinese README, tutorials and blog by @woodforestsen in https://github.com/deepspeedai/DeepSpeed/pull/8894\r\n* Bump modal-torch-latest CI Python from 3.10 to 3.12 by @delock in https://github.com/deepspeedai/DeepSpeed/pull/8902\r\n* [AffineIR] Plan expert transfers from the live runtime placement by @jinyouzhi in https://github.com/deepspeedai/DeepSpeed/pull/8800\r\n* Set TORCHINDUCTOR_COMPILE_THREADS=1 in modal CI sandbox to stop 300s test timeouts by @delock in https://github.com/deepspeedai/DeepSpeed/pull/8914\r\n* Fix scalar parameter byte-view broadcasts by @adenzhou1350 in https://github.com/deepspeedai/DeepSpeed/pull/8642\r\n* Honor fp16 hysteresis in the ZeRO stage 0 fp16 optimizers by @vineethsaivs in https://github.com/deepspeedai/DeepSpeed/pull/8647\r\n* Strip ANSI color codes before parsing FAILED lines in cpu-multi-triage by @delock in https://github.com/deepspeedai/DeepSpeed/pull/8882\r\n* Bump AWS capacity-fallback Python from 3.10.13 to 3.12.14 by @delock in https://github.com/deepspeedai/DeepSpeed/pull/8921\r\n* AutoEP: fused Triton kernel for the clamped expert activations by @pengdurice in https://github.com/deepspeedai/DeepSpeed/pull/8652\r\n* Keep fp32 masters on ZeRO-1 and ZeRO-2 when a resume skips optimizer state by @sfc-gh-truwase in https://github.com/deepspeedai/DeepSpeed/pull/8891\r\n* Skip accelerator sync for unprofiled comm ops by @guptaishaan in https://github.com/deepspeedai/DeepSpeed/pull/8795\r\n* Fix performance regression caused by comms logger and bug in comms_logger.enabled by @hwchen2017 in https://github.com/deepspeedai/DeepSpeed/pull/8909\r\n* Use nvtx's module-level API so NVTX domain ranges work on older nvtx releases by @hwchen2017 in https://github.com/deepspeedai/DeepSpeed/pull/8900\r\n* [AffineIR] Affine storage maps for live ZeRO parameters and optimizer states by @jinyouzhi in https://github.com/deepspeedai/DeepSpeed/pull/8885\r\n* Add a v1 multi-rank CPU suite (manual dispatch first) by @delock in https://github.com/deepspeedai/DeepSpeed/pull/8926\r\n* [AutoEP] Ensure consistent order of partition groups in all-gather when ZeRO-3  by @jinyouzhi in https://github.com/deepspeedai/DeepSpeed/pull/8876\r\n* Fix race between folded checkpoint load and metadata drop in AutoEP+AutoTP test by @jinyouzhi in https://github.com/deepspeedai/DeepSpeed/pull/8943\r\n* Check where column-split pieces write in a shard by @Achyuthan-S in https://github.com/deepspeedai/DeepSpeed/pull/8922\r\n\r\n## New Contributors\r\n* @arcusbuilds made their first contribution in https://github.com/deepspeedai/DeepSpeed/pull/8782\r\n* @zhangjian29 made their first contribution in https://github.com/deepspeedai/Deep"},"readme":true,"readme_zh":true,"news":[],"stale":false,"snapshot_at":"2026-11-10T09:35:25Z","history_start":"2026-09-06"}