nemo_automodel.components.models#
Convenience model builders for NeMo Automodel.
Currently includes: • build_gpt2_model – returns a GPT-2 causal language model (Flash-Attention-2 by default).
Subpackages#
nemo_automodel.components.models.llamanemo_automodel.components.models.hy_mt2nemo_automodel.components.models.baichuannemo_automodel.components.models.gemma4_moenemo_automodel.components.models.ministral_bidirectionalnemo_automodel.components.models.glm_moe_dsanemo_automodel.components.models.mimo_v2_flashnemo_automodel.components.models.mistral3_vlmnemo_automodel.components.models.gemma4_drafternemo_automodel.components.models.qwen3_5_moenemo_automodel.components.models.hy_v3nemo_automodel.components.models.mistral3nemo_automodel.components.models.qwen3_5nemo_automodel.components.models.qwen2_5_omninemo_automodel.components.models.llama_bidirectionalnemo_automodel.components.models.ling_v2nemo_automodel.components.models.minimax_m2nemo_automodel.components.models.ernie4_5nemo_automodel.components.models.glm4_moe_litenemo_automodel.components.models.qwen2nemo_automodel.components.models.qwen3_nextnemo_automodel.components.models.deepseek_v32nemo_automodel.components.models.deepseek_v3nemo_automodel.components.models.qwen3_omni_moenemo_automodel.components.models.glm4_moenemo_automodel.components.models.mistral4nemo_automodel.components.models.nemotron_v3nemo_automodel.components.models.deepseek_v4nemo_automodel.components.models.deepseek_v4.kernelsnemo_automodel.components.models.deepseek_v4.kernels.tilelang_sparse_mla_bwdnemo_automodel.components.models.deepseek_v4.kernels.tilelang_indexer_fwdnemo_automodel.components.models.deepseek_v4.kernels.sparse_attentionnemo_automodel.components.models.deepseek_v4.kernels.tilelang_indexer_bwdnemo_automodel.components.models.deepseek_v4.kernels.tilelang_sparse_mla_fwdnemo_automodel.components.models.deepseek_v4.kernels.tilelang_indexer
nemo_automodel.components.models.deepseek_v4.layersnemo_automodel.components.models.deepseek_v4.modelnemo_automodel.components.models.deepseek_v4.optimized_kernelsnemo_automodel.components.models.deepseek_v4.fsdpnemo_automodel.components.models.deepseek_v4.confignemo_automodel.components.models.deepseek_v4.mtpnemo_automodel.components.models.deepseek_v4.state_dict_adapter
nemo_automodel.components.models.qwen3_moenemo_automodel.components.models.step3p5nemo_automodel.components.models.gpt_ossnemo_automodel.components.models.kimivlnemo_automodel.components.models.nemotron_parsenemo_automodel.components.models.llava_onevisionnemo_automodel.components.models.commonnemo_automodel.components.models.common.mtpnemo_automodel.components.models.common.hf_checkpointing_mixinnemo_automodel.components.models.common.utilsnemo_automodel.components.models.common.packingnemo_automodel.components.models.common.bidirectionalnemo_automodel.components.models.common.inbatch_neg_utils
nemo_automodel.components.models.kimi_k25_vlnemo_automodel.components.models.nemotron_omninemo_automodel.components.models.step3p7nemo_automodel.components.models.step3p7.vision_encodernemo_automodel.components.models.step3p7.modelnemo_automodel.components.models.step3p7.configuration_step3p7nemo_automodel.components.models.step3p7.mtpnemo_automodel.components.models.step3p7.processing_step3nemo_automodel.components.models.step3p7.state_dict_adapter
nemo_automodel.components.models.qwen3_vl_moe
Submodules#
Package Contents#
Classes#
Produces a helpful error when importing a non-existent model subpackage. |
Functions#
Return the set of model sub-package names shipped with this installation. |
|
Data#
API#
- nemo_automodel.components.models.__all__#
[‘build_gpt2_model’]
- nemo_automodel.components.models._MODELS_DIR#
None
- nemo_automodel.components.models._PACKAGE_PREFIX#
None
- nemo_automodel.components.models._available_model_submodules() set[str]#
Return the set of model sub-package names shipped with this installation.
- nemo_automodel.components.models._make_upgrade_message(name: str) str#
- nemo_automodel.components.models.__getattr__(name: str)#
- class nemo_automodel.components.models._MissingModelFinder#
Bases:
importlib.abc.MetaPathFinderProduces a helpful error when importing a non-existent model subpackage.
Installed at the end of
sys.meta_pathso it is only consulted after all real finders have already returnedNone. For any import of the formnemo_automodel.components.models.<name>(direct child only), it raises aModuleNotFoundErrorwith upgrade instructions instead of the default unhelpful message.- find_spec(fullname, path, target=None)#