core.models.multimodal.llava_spec

Module Contents

Functions

_te	The Transformer Engine provider, built on first use.
decoder_model_with_transformer_engine_default_spec	LLava decoder TE spec (uses Transformer Engine components).
decoder_model_with_local_default_spec	LLava decoder local spec.

Data

_local

API

core.models.multimodal.llava_spec._te()

The Transformer Engine provider, built on first use.

Deferred so that importing this module does not require Transformer Engine – only building a Transformer Engine spec does. That is what the HAVE_TE guard used to buy.

core.models.multimodal.llava_spec._local

‘get_backend(…)’

core.models.multimodal.llava_spec.decoder_model_with_transformer_engine_default_spec(

num_experts: Optional[int] = None,

moe_grouped_gemm: bool = False,

qk_layernorm: bool = False,

) → megatron.core.transformer.spec_utils.ModuleSpec

LLava decoder TE spec (uses Transformer Engine components).

core.models.multimodal.llava_spec.decoder_model_with_local_default_spec(

num_experts: Optional[int] = None,

moe_grouped_gemm: bool = False,

qk_layernorm: bool = False,

) → megatron.core.transformer.spec_utils.ModuleSpec

LLava decoder local spec.