FSDP (PyTorch)
https://pytorch.org/docs/stable/fsdp.html
Native PyTorch fully-sharded data parallel implementation — Meta's response to DeepSpeed ZeRO; FSDP2 in 2024 with improved API.
Active · Verified 2026-10-27 · fresh as of 2026-10-27 · 15 cited facts, 1 with a third-party source · 1 recorded change · Suggest a correction
At a glance
- Type
- Fully Sharded Data Parallel vendor fix
- Tier
- T1
- Section
- Training infrastructure
- Created
- 2022 (FSDP), 2024 (FSDP2) vendor fix
- Latest release
- v2.14.1 (2026-09-30) — torch registry fix
- License
- BSD-3-Clause (PyTorch) vendor fix
- Pricing
- searched not found fix
- Funding
- searched not found fix
- Freshness
- 2026-10-27 fix
- Layer
- model-infra repo fix
- Licensing
- open-source vendor fix
- Developer activity
- OpenRank 479.59 (12-mo avg 626.17) · 1033 participants — 2026-09 third-party fix
When to use
Optimised for: searched not found
Anti-fit: searched not found
Pros & cons
Pros
searched not found
Cons
searched not found
Claims & capabilities
searched not found
Technical surface
- API surface
- Python (torch.distributed.fsdp: fully_shard / FSDP2, FullyShardedDataParallel / FSDP1)
- Backend storage
- searched not found
- Deployment
- Self-hosted only — library (part of PyTorch)
- Embedding model
- not applicable — not a memory product
- Multi-tenancy
- searched not found
- MCP
- searched not found
- A2A
- searched not found
- OpenTelemetry
- searched not found
Similar systems
Other training infrastructure in the catalog, ranked by inbound references.
- MLflow T1
Databricks-incubated open platform for ML lifecycle — tracking, projects, models, model registry. Standard ML-ops baseline. MLflow 3.0 (2024) added LLM-eval features.
- AWS SageMaker T1
AWS's managed ML platform — Studio, Training, HyperPod (large-scale clusters), Endpoints. Industry-standard for enterprise AWS-hosted training.
- Weights & Biases T1
Experiment-tracking and ML-ops platform (metrics, artifacts, sweeps, Weave tracing/evals). Acquired by CoreWeave (completed 2025-05-05, ~$1.7B); now part of CoreWeave Forge.
- Comet ML T2
ML experiment tracking + LLM evaluation platform (Opik for LLM eval). Comet competes head-to-head with Weights & Biases.
- Azure Machine Learning T1
Microsoft Azure's managed ML platform — pipelines, AutoML, Designer, prompt flow (LLM). Heavy integration with Azure OpenAI and Azure AI Studio.
- LoRA T3
Microsoft Research's Low-Rank Adapters method — trains rank-r matrices added to attention weights; 10000x parameter reduction. Foundation of all adapter-style fine-tuning.
Sources (14 cited facts, 1 third-party)
Each fact links to the page it came from; the badge says who published it (how to read sources).
- What it is Fully Sharded Data Parallel vendor
- What it does & what's distinct Native PyTorch fully-sharded data parallel implementation — Meta's response to DeepSpeed… vendor
- Created 2022 (FSDP), 2024 (FSDP2) vendor
- Latest release v2.14.1 (2026-09-30) — torch registry
- License BSD-3-Clause (PyTorch) vendor
- GitHub pytorch/pytorch — ~105k stars, last commit 2026-10 repo
- Deployment Self-hosted only — library (part of PyTorch) vendor
- Code/weights release Active — last commit 2026-05 · 99,857★ repo
- API surface Python (torch.distributed.fsdp: fully_shard / FSDP2, FullyShardedDataParallel / FSDP1) vendor
- architecture-deployment-locus local vendor
- licensing-bucket open-source vendor
- positioning-platform-vs-product platform vendor
- taxonomy-three-layer-rollup model-infra repo
- developer-activity-opendigger {"activity":3289.34,"as_of":"2026-09","openrank_12mo":626.17,"openrank_current":479.59,"… third-party
What changed
Changes to this record's lifecycle, licence, release, pricing and hosting facts, from the catalog's git history. Subscribe (Atom)