`nemo_automodel.components.models.mistral3.model`#

Module Contents#

Classes#

`Ministral3Config`	Configuration for Ministral3 text decoder.
`Ministral3PreTrainedModel`
`Ministral3ModelOutputWithPast`
`Ministral3CausalLMOutputWithPast`
`Ministral3RotaryEmbedding`
`Ministral3Attention`
`Ministral3MLP`
`Ministral3RMSNorm`
`GradientCheckpointingLayer`
`Ministral3DecoderLayer`
`Ministral3Model`
`Ministral3ForCausalLM`

Functions#

`rotate_half`
`apply_rotary_pos_emb`
`repeat_kv`
`eager_attention_forward`
`_get_llama_4_attn_scale`
`_patched_from_config`

Data#

`logger`
`mod_pkg`
`config_mod`
`modeling_mod`
`_orig_auto_from_config`
`ModelClass`

API#

nemo_automodel.components.models.mistral3.model.logger#: ‘get_logger(…)’

class nemo_automodel.components.models.mistral3.model.Ministral3Config(

vocab_size: Optional[int] = 131072,

hidden_size: Optional[int] = 4096,

intermediate_size: Optional[int] = 14336,

num_hidden_layers: Optional[int] = 34,

num_attention_heads: Optional[int] = 32,

num_key_value_heads: Optional[int] = 8,

head_dim: Optional[int] = 128,

hidden_act: Optional[str] = 'silu',

max_position_embeddings: Optional[int] = 262144,

initializer_range: Optional[float] = 0.02,

rms_norm_eps: Optional[float] = 1e-05,

use_cache: Optional[bool] = True,

pad_token_id: Optional[int] = 11,

bos_token_id: Optional[int] = 1,

eos_token_id: Optional[int] = 2,

tie_word_embeddings: Optional[bool] = False,

rope_parameters: Optional[dict] = None,

sliding_window: Optional[int] = None,

attention_dropout: Optional[float] = 0.0,

**kwargs,

)#

Bases: transformers.configuration_utils.PretrainedConfig

Configuration for Ministral3 text decoder.

Initialization

model_type#: ‘ministral3’

keys_to_ignore_at_inference#: [‘past_key_values’]

base_model_tp_plan#: None

base_model_pp_plan#: None

nemo_automodel.components.models.mistral3.model.rotate_half(x)#

class nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel#

Bases: transformers.modeling_utils.PreTrainedModel

config: nemo_automodel.components.models.mistral3.model.Ministral3Config#: None

base_model_prefix#: ‘model’

supports_gradient_checkpointing#: True

_no_split_modules#: [‘Ministral3DecoderLayer’]

_skip_keys_device_placement#: [‘past_key_values’]

_supports_flash_attn#: True

_supports_sdpa#: True

_supports_flex_attn#: True

_can_compile_fullgraph#: True

_supports_attention_backend#: True

_can_record_outputs#: None

class nemo_automodel.components.models.mistral3.model.Ministral3ModelOutputWithPast#

Bases: transformers.modeling_outputs.BaseModelOutputWithPast

image_hidden_states: Optional[torch.FloatTensor]#: None

class nemo_automodel.components.models.mistral3.model.Ministral3CausalLMOutputWithPast#: Bases: transformers.modeling_outputs.CausalLMOutputWithPast

class nemo_automodel.components.models.mistral3.model.Ministral3RotaryEmbedding( config: nemo_automodel.components.models.mistral3.model.Ministral3Config, device=None, )#

Bases: torch.nn.Module

Initialization

inv_freq: torch.Tensor#: None

static compute_default_rope_parameters( config: Optional[nemo_automodel.components.models.mistral3.model.Ministral3Config] = None, device: Optional[torch.device] = None, seq_len: Optional[int] = None, ) → tuple[torch.Tensor, float]#

forward(x, position_ids)#

nemo_automodel.components.models.mistral3.model.apply_rotary_pos_emb( q, k, cos, sin, position_ids=None, unsqueeze_dim=1, )#

nemo_automodel.components.models.mistral3.model.repeat_kv(hidden_states: torch.Tensor, n_rep: int) → torch.Tensor#

nemo_automodel.components.models.mistral3.model.eager_attention_forward(

module: torch.nn.Module,

query: torch.Tensor,

key: torch.Tensor,

value: torch.Tensor,

attention_mask: Optional[torch.Tensor],

scaling: float,

dropout: float = 0.0,

**kwargs: transformers.processing_utils.Unpack[transformers.utils.TransformersKwargs],

)#

nemo_automodel.components.models.mistral3.model._get_llama_4_attn_scale( positions_ids: torch.Tensor, beta: float, max_position_embeddings: int, ) → torch.Tensor#

class nemo_automodel.components.models.mistral3.model.Ministral3Attention( config: nemo_automodel.components.models.mistral3.model.Ministral3Config, layer_idx: int, )#

Bases: torch.nn.Module

Initialization

forward(

hidden_states: torch.Tensor,

position_embeddings: tuple[torch.Tensor, torch.Tensor],

attention_mask: Optional[torch.Tensor],

past_key_values: Optional[transformers.cache_utils.Cache] = None,

cache_position: Optional[torch.LongTensor] = None,

**kwargs: transformers.processing_utils.Unpack[transformers.modeling_flash_attention_utils.FlashAttentionKwargs],

) → tuple[torch.Tensor, Optional[torch.Tensor]]#

class nemo_automodel.components.models.mistral3.model.Ministral3MLP( config: nemo_automodel.components.models.mistral3.model.Ministral3Config, )#

Bases: torch.nn.Module

Initialization

forward(x)#

class nemo_automodel.components.models.mistral3.model.Ministral3RMSNorm(hidden_size, eps=1e-06)#

Bases: torch.nn.Module

Initialization

forward(hidden_states)#

class nemo_automodel.components.models.mistral3.model.GradientCheckpointingLayer#

Bases: torch.nn.Module

abstractmethod forward(*args, **kwargs)#

class nemo_automodel.components.models.mistral3.model.Ministral3DecoderLayer( config: nemo_automodel.components.models.mistral3.model.Ministral3Config, layer_idx: int, )#

Bases: nemo_automodel.components.models.mistral3.model.GradientCheckpointingLayer

Initialization

forward(

hidden_states: torch.Tensor,

attention_mask: Optional[torch.Tensor] = None,

position_ids: Optional[torch.LongTensor] = None,

past_key_values: Optional[transformers.cache_utils.Cache] = None,

use_cache: Optional[bool] = False,

cache_position: Optional[torch.LongTensor] = None,

position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,

**kwargs: transformers.processing_utils.Unpack[transformers.utils.TransformersKwargs],

) → torch.Tensor#

class nemo_automodel.components.models.mistral3.model.Ministral3Model( config: nemo_automodel.components.models.mistral3.model.Ministral3Config, )#

Bases: nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel

Initialization

forward(

input_ids: Optional[torch.LongTensor] = None,

attention_mask: Optional[torch.Tensor] = None,

position_ids: Optional[torch.LongTensor] = None,

past_key_values: Optional[transformers.cache_utils.Cache] = None,

inputs_embeds: Optional[torch.FloatTensor] = None,

use_cache: Optional[bool] = None,

cache_position: Optional[torch.LongTensor] = None,

**kwargs: transformers.processing_utils.Unpack[transformers.utils.TransformersKwargs],

) → transformers.modeling_outputs.BaseModelOutputWithPast#

class nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM( config: nemo_automodel.components.models.mistral3.model.Ministral3Config, )#

Bases: nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel, transformers.generation.GenerationMixin

Initialization

_tied_weights_keys#: None

_tp_plan#: None

_pp_plan#: None

forward(

input_ids: Optional[torch.LongTensor] = None,

attention_mask: Optional[torch.Tensor] = None,

position_ids: Optional[torch.LongTensor] = None,

past_key_values: Optional[transformers.cache_utils.Cache] = None,

inputs_embeds: Optional[torch.FloatTensor] = None,

labels: Optional[torch.LongTensor] = None,

use_cache: Optional[bool] = None,

cache_position: Optional[torch.LongTensor] = None,

logits_to_keep: Union[int, torch.Tensor] = 0,

**kwargs: transformers.processing_utils.Unpack[transformers.utils.TransformersKwargs],

) → transformers.modeling_outputs.CausalLMOutputWithPast#

nemo_automodel.components.models.mistral3.model.mod_pkg#: ‘setdefault(…)’

nemo_automodel.components.models.mistral3.model.config_mod#: ‘ModuleType(…)’

nemo_automodel.components.models.mistral3.model.modeling_mod#: ‘ModuleType(…)’

nemo_automodel.components.models.mistral3.model._orig_auto_from_config#: None

nemo_automodel.components.models.mistral3.model._patched_from_config(cls, config, *model_args, **kwargs)#

nemo_automodel.components.models.mistral3.model.ModelClass#: None

nemo_automodel.components.models.mistral3.model#

Module Contents#

Classes#

Functions#

Data#

API#

`nemo_automodel.components.models.mistral3.model`#