nemo_automodel.components.models.deepseek_v3.model

Module Contents

Classes

Name	Description
`Block`	-
`DeepseekV3ForCausalLM`	-
`DeepseekV3Model`	-

Data

ModelClass

API

class nemo_automodel.components.models.deepseek_v3.model.Block(
    layer_idx: int,
    config: transformers.models.deepseek_v3.configuration_deepseek_v3.DeepseekV3Config,
    moe_config: nemo_automodel.components.moe.config.MoEConfig,
    backend: nemo_automodel.components.models.common.BackendConfig
)

Bases: Module

input_layernorm

is_moe_layer

= layer_idx >= config.first_k_dense_replace

mlp

post_attention_layernorm

self_attn

= MLA(config, backend)

nemo_automodel.components.models.deepseek_v3.model.Block._mlp(
    x: torch.Tensor,
    padding_mask: torch.Tensor
) -> torch.Tensor

nemo_automodel.components.models.deepseek_v3.model.Block.forward(
    x: torch.Tensor,
    freqs_cis: torch.Tensor,
    attention_mask: torch.Tensor | None = None,
    padding_mask: torch.Tensor | None = None,
    attn_kwargs: typing.Any = {}
) -> tuple[torch.Tensor, torch.Tensor | None]

Forward pass for the Transformer block.

Parameters:

torch.Tensor

Input tensor.

freqs_cis

torch.Tensor

Precomputed complex exponential values for rotary embeddings.

padding_mask

torch.TensorDefaults to None

Boolean tensor indicating padding positions.

Returns: torch.Tensor

torch.Tensor: Output tensor after block computation.

nemo_automodel.components.models.deepseek_v3.model.Block.init_weights(
    buffer_device: torch.device
)

class nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM(
    config: transformers.models.deepseek_v3.configuration_deepseek_v3.DeepseekV3Config,
    moe_config: nemo_automodel.components.moe.config.MoEConfig | None = None,
    backend: nemo_automodel.components.models.common.BackendConfig | None = None,
    kwargs = {}
)

Bases: HFCheckpointingMixin, Module, MoEFSDPSyncMixin

_keep_in_fp32_modules_strict

= ['e_score_correction_bias']

backend

= backend or BackendConfig()

lm_head

model

state_dict_adapter

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.forward(
    input_ids: torch.Tensor,
    position_ids: torch.Tensor | None = None,
    attention_mask: torch.Tensor | None = None,
    padding_mask: torch.Tensor | None = None,
    logits_to_keep: typing.Union[int, torch.Tensor] = 0,
    output_hidden_states: typing.Optional[bool] = None,
    attn_kwargs: typing.Any = {}
) -> transformers.modeling_outputs.CausalLMOutputWithPast

Forward pass returning :class:~transformers.modeling_outputs.CausalLMOutputWithPast.

Parameters:

input_ids

torch.Tensor

Input token IDs. BSHD: [B, S]; THD: [1, T] (squeezed internally).

position_ids

torch.Tensor | NoneDefaults to None

Optional position indices.

attention_mask

torch.Tensor | NoneDefaults to None

Optional attention mask.

padding_mask

torch.Tensor | NoneDefaults to None

Optional padding mask.

logits_to_keep

Union[int, torch.Tensor]Defaults to 0

If 0 (default), compute logits for all positions; otherwise only compute logits for the last logits_to_keep positions (avoids materialising the full logit matrix during generation / fused CE training).

output_hidden_states

Optional[bool]Defaults to None

Whether to carry the final hidden states on the output.

**attn_kwargs

AnyDefaults to {}

Additional arguments forwarded to the base model (e.g. qkv_format, cu_seqlens, CP kwargs).

Returns: CausalLMOutputWithPast

class:~transformers.modeling_outputs.CausalLMOutputWithPast with logits and,

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.from_config(
    config: transformers.models.deepseek_v3.configuration_deepseek_v3.DeepseekV3Config,
    moe_config: nemo_automodel.components.moe.config.MoEConfig | None = None,
    backend: nemo_automodel.components.models.common.BackendConfig | None = None,
    kwargs = {}
)

classmethod

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.from_pretrained(
    pretrained_model_name_or_path: str,
    model_args = (),
    kwargs = {}
)

classmethod

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.get_input_embeddings()

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.get_output_embeddings()

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.initialize_weights(
    buffer_device: torch.device | None = None,
    dtype: torch.dtype = torch.bfloat16
) -> None

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.set_input_embeddings(
    value
)

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.set_output_embeddings(
    new_embeddings
)

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM.update_moe_gate_bias() -> None

class nemo_automodel.components.models.deepseek_v3.model.DeepseekV3Model(
    config: transformers.models.deepseek_v3.configuration_deepseek_v3.DeepseekV3Config,
    backend: nemo_automodel.components.models.common.BackendConfig,
    moe_config: nemo_automodel.components.moe.config.MoEConfig | None = None,
    moe_overrides: dict | None = None
)

Bases: Module

embed_tokens

layers

= torch.nn.ModuleDict()

max_seq_len

= config.max_position_embeddings

moe_config

= moe_config or MoEConfig(**moe_defaults)

norm

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3Model.forward(
    input_ids: torch.Tensor | None = None,
    inputs_embeds: torch.Tensor | None = None,
    position_ids: torch.Tensor | None = None,
    attention_mask: torch.Tensor | None = None,
    padding_mask: torch.Tensor | None = None,
    attn_kwargs: typing.Any = {}
) -> tuple[torch.Tensor, torch.Tensor | None]

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3Model.init_weights(
    buffer_device: torch.device | None = None
) -> None

nemo_automodel.components.models.deepseek_v3.model.DeepseekV3Model.update_moe_gate_bias() -> None

nemo_automodel.components.models.deepseek_v3.model.ModelClass = DeepseekV3ForCausalLM