> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo/automodel/_mcp/server.

# nemo_automodel.components.models.deepseek_v4

## Subpackages

* **[`nemo_automodel.components.models.deepseek_v4.kernels`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels)**

## Submodules

* **[`nemo_automodel.components.models.deepseek_v4.config`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/config)**
* **[`nemo_automodel.components.models.deepseek_v4.cp`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/cp)**
* **[`nemo_automodel.components.models.deepseek_v4.fsdp`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/fsdp)**
* **[`nemo_automodel.components.models.deepseek_v4.layers`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/layers)**
* **[`nemo_automodel.components.models.deepseek_v4.model`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/model)**
* **[`nemo_automodel.components.models.deepseek_v4.mtp`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/mtp)**
* **[`nemo_automodel.components.models.deepseek_v4.optimized_kernels`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/optimized_kernels)**
* **[`nemo_automodel.components.models.deepseek_v4.parallelization`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/parallelization)**
* **[`nemo_automodel.components.models.deepseek_v4.processing`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/processing)**
* **[`nemo_automodel.components.models.deepseek_v4.state_dict_adapter`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/state_dict_adapter)**
* **[`nemo_automodel.components.models.deepseek_v4.vision`](/nemo-automodel/nemo_automodel/components/models/deepseek_v4/vision)**

## Docs

- [nemo_automodel.components.models.deepseek_v4.config](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/config.md)
- [nemo_automodel.components.models.deepseek_v4.cp](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/cp.md)
- [nemo_automodel.components.models.deepseek_v4.fsdp](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/fsdp.md)
- [nemo_automodel.components.models.deepseek_v4.kernels](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels.md)
- [nemo_automodel.components.models.deepseek_v4.kernels.sparse_attention](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels/sparse_attention.md)
- [nemo_automodel.components.models.deepseek_v4.kernels.tilelang_indexer](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels/tilelang_indexer.md)
- [nemo_automodel.components.models.deepseek_v4.kernels.tilelang_indexer_bwd](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels/tilelang_indexer_bwd.md)
- [nemo_automodel.components.models.deepseek_v4.kernels.tilelang_indexer_fwd](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels/tilelang_indexer_fwd.md)
- [nemo_automodel.components.models.deepseek_v4.kernels.tilelang_sparse_mla_bwd](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels/tilelang_sparse_mla_bwd.md)
- [nemo_automodel.components.models.deepseek_v4.kernels.tilelang_sparse_mla_fwd](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/kernels/tilelang_sparse_mla_fwd.md)
- [nemo_automodel.components.models.deepseek_v4.layers](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/layers.md)
- [nemo_automodel.components.models.deepseek_v4.model](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/model.md)
- [nemo_automodel.components.models.deepseek_v4.mtp](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/mtp.md)
- [nemo_automodel.components.models.deepseek_v4.optimized_kernels](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/optimized_kernels.md)
- [nemo_automodel.components.models.deepseek_v4.parallelization](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/parallelization.md)
- [nemo_automodel.components.models.deepseek_v4.processing](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/processing.md)
- [nemo_automodel.components.models.deepseek_v4.state_dict_adapter](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/state_dict_adapter.md)
- [nemo_automodel.components.models.deepseek_v4.vision](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/models/deepseek_v4/vision.md)