> 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/moe/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.moe

## Subpackages

* **[`nemo_automodel.components.moe.megatron`](/nemo-automodel/nemo_automodel/components/moe/megatron)**
* **[`nemo_automodel.components.moe.uccl_ep`](/nemo-automodel/nemo_automodel/components/moe/uccl_ep)**

## Submodules

* **[`nemo_automodel.components.moe.config`](/nemo-automodel/nemo_automodel/components/moe/config)**
* **[`nemo_automodel.components.moe.experts`](/nemo-automodel/nemo_automodel/components/moe/experts)**
* **[`nemo_automodel.components.moe.fsdp_mixin`](/nemo-automodel/nemo_automodel/components/moe/fsdp_mixin)**
* **[`nemo_automodel.components.moe.layers`](/nemo-automodel/nemo_automodel/components/moe/layers)**
* **[`nemo_automodel.components.moe.load_balance_metrics`](/nemo-automodel/nemo_automodel/components/moe/load_balance_metrics)**
* **[`nemo_automodel.components.moe.mok_experts`](/nemo-automodel/nemo_automodel/components/moe/mok_experts)**
* **[`nemo_automodel.components.moe.mxfp8`](/nemo-automodel/nemo_automodel/components/moe/mxfp8)**
* **[`nemo_automodel.components.moe.optimized_ops`](/nemo-automodel/nemo_automodel/components/moe/optimized_ops)**
* **[`nemo_automodel.components.moe.parallelizer`](/nemo-automodel/nemo_automodel/components/moe/parallelizer)**
* **[`nemo_automodel.components.moe.quantized_experts`](/nemo-automodel/nemo_automodel/components/moe/quantized_experts)**
* **[`nemo_automodel.components.moe.router_replay`](/nemo-automodel/nemo_automodel/components/moe/router_replay)**
* **[`nemo_automodel.components.moe.state_dict_mixin`](/nemo-automodel/nemo_automodel/components/moe/state_dict_mixin)**
* **[`nemo_automodel.components.moe.state_dict_utils`](/nemo-automodel/nemo_automodel/components/moe/state_dict_utils)**
* **[`nemo_automodel.components.moe.swiglu_triton`](/nemo-automodel/nemo_automodel/components/moe/swiglu_triton)**
* **[`nemo_automodel.components.moe.tp_plan_validation`](/nemo-automodel/nemo_automodel/components/moe/tp_plan_validation)**

## Docs

- [nemo_automodel.components.moe.config](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/config.md)
- [nemo_automodel.components.moe.experts](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/experts.md)
- [nemo_automodel.components.moe.fsdp_mixin](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/fsdp_mixin.md)
- [nemo_automodel.components.moe.layers](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/layers.md)
- [nemo_automodel.components.moe.load_balance_metrics](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/load_balance_metrics.md)
- [nemo_automodel.components.moe.megatron](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/megatron.md)
- [nemo_automodel.components.moe.megatron.fused_a2a](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/megatron/fused_a2a.md)
- [nemo_automodel.components.moe.megatron.fused_indices_converter](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/megatron/fused_indices_converter.md)
- [nemo_automodel.components.moe.megatron.moe_utils](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/megatron/moe_utils.md)
- [nemo_automodel.components.moe.megatron.token_dispatcher](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/megatron/token_dispatcher.md)
- [nemo_automodel.components.moe.mok_experts](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/mok_experts.md)
- [nemo_automodel.components.moe.mxfp8](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/mxfp8.md)
- [nemo_automodel.components.moe.optimized_ops](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/optimized_ops.md)
- [nemo_automodel.components.moe.parallelizer](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/parallelizer.md)
- [nemo_automodel.components.moe.quantized_experts](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/quantized_experts.md)
- [nemo_automodel.components.moe.router_replay](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/router_replay.md)
- [nemo_automodel.components.moe.state_dict_mixin](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/state_dict_mixin.md)
- [nemo_automodel.components.moe.state_dict_utils](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/state_dict_utils.md)
- [nemo_automodel.components.moe.swiglu_triton](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/swiglu_triton.md)
- [nemo_automodel.components.moe.tp_plan_validation](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/tp_plan_validation.md)
- [nemo_automodel.components.moe.uccl_ep](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/uccl_ep.md)
- [nemo_automodel.components.moe.uccl_ep.buffer](https://docs.nvidia.com/nemo/automodel/v0.4/nemo-automodel/nemo_automodel/components/moe/uccl_ep/buffer.md)