Project NVIDIA/NeMo
On this site you can view the documentation of the Protocol Buffers (Protobuf) files / API in the NeMo project.
Last checked for updates at:
(we're working on daily update checks)
This website is new and is still being built. You're welcome to talk with us about possible improvements we can make.
Source
Generated from commit: 2be5853
This documentation applies to the following branches and tags, which have the same .proto files as commit 2be5853:
- Branches: 1d_stride, 24.09-alpha, 24.09-beta, 25.04-alpha, 2d-bucketing-and-tps-improvements, aanoosheh/fix-25.03-distill-test, aanoosheh/intermediates-distill, aanoosheh/mcore-pp-refactor, aanoosheh/modelopt-specdec-converter, abhi/llava_next_sp, add_frame_vad, add-infer-requirements, add-pydantic-settings, add_rope_dynamic_scale, add_rope_dynamic_scale_zero_sc_eval, add_t5_fa, add_torchfix, adil/mlm-custom-loop-inference-loading, adithayre/peft_eval, adithyare/check_O2_preprocess, adithyare/clip_max_len_fix, adithyare/convert_adapter_ckpt_to_nemo, adithyare/dump_encoded_input, adithyare/inverse_peft, adithyare/leftonly_prompt_tokens, adithyare/linear_combination, adithyare/llm_embeddings, adithyare/mamaba_updates, adithyare/mamba_dist_ckpt_updates, adithyare/mamba_refac, adithyare/max_len_gen_fix, adithyare/memmap_dataset, adithyare/merge_fix, adithyare/mixed_task_lora_batch, adithyare/nl_lora, adithyare/peft_eval_setting, adithyare/peft_training_fldr, adithyare/prec_fix_in_conversion, adithyare/p_tuning_refactor_4, adithyare/refac_ptuning_part3, adithyare/refac_sft_peft, adithyare/residual_ptuning, adithyare/sft_layer_freeze, adithyare/t5_val_acc, adithyare/te_type_error, adithyare/tied_lora, adithyare/tiny_attn, adithyare/vera, adithyare/zero_shot_eval_prompt_learning, aficek/adapter-retro, aficek/retro_adapters_impl, aficek/retro_adapters_impl_improved, aficek/retro_adapters_v1.6, aficek/retro_ptuning_adapter, ahmadki/fabric, ahmadki/megatron_bw_hang_temp_fix, ahmadki/megatron_llm, ahmadki/nemov2_nemotron, akoumparouli/2412_moe_fix, akoumparouli/api_without_nemo_run, akoumparouli/athitten/packed_sequence_automodel, akoumparouli/automodel_cpu_only_examples, akoumparouli/automodel_ddp_fix, akoumparouli/automodel_hsdp, akoumparouli/automodel_move_jittransform, akoumparouli/automodel_reraise_exception, akoumparouli/automodel_update_fp8_autocast, akoumparouli/bump_mcore_commit, akoumparouli/bump_transformers_version, akoumparouli/canonical_lora_for_non_mcore, akoumparouli/chat_template_train_llama3, akoumparouli/cherry-pik-10780-r2.0.0, akoumparouli/copy_parallel_size, akoumparouli/default_param_fix_tp_comm_overlap_disable_qkv, akoumparouli/deprecate_app_state, akoumparouli/destructor_singleton_in_megatronstrategy_dtor, akoumparouli/disable_dynamo_for_tests, akoumparouli/fa3, akoumparouli/fix-bump-ci-container--NVIDIA-Megatron-LM-2025-01-23, akoumparouli/fix_ci_ddp_test, akoumparouli/fix_fp8_tests, akoumparouli/fix_init_model_parallel, akoumparouli/fix_progress_printer, akoumparouli/fsdp_peft, akoumparouli/generate_sequence_parallelism_bugfix, akoumparouli/gpt_sft_model_minor_fix, akoumparouli/guarded_imports, akoumparouli/helpers_sanity_check, akoumparouli/make_HFDatasetDataModule_arg_accept_path_or_dataset, akoumparouli/mcore_dist_opt, akoumparouli/mcore_tokenizer_signature_pairing_dense_log_utils, akoumparouli/minor_fix_autotokenizer, akoumparouli/moe_fix_bitexact_test, akoumparouli/move_automodel_examples, akoumparouli/move_call, akoumparouli/nemo_automodel_peft_te_linear_fix, akoumparouli/nemotron_49b_hf_export_bugfix, akoumparouli/nemo_ux_auto_model, akoumparouli/nemo_ux_cudagraph_plugin, akoumparouli/nemo_ux_fix_hf_auto_model_param_names, akoumparouli/nemo_ux_hangman, akoumparouli/nemo_ux_llama_sft_and_peft_examples, akoumparouli/nemo_ux_make_llm_api_ingest_nn_module, akoumparouli/nemo_ux_memprofiler_log_on_batch, akoumparouli/nemo_ux_mixtral, akoumparouli/nemo_ux_module_import_proxy, akoumparouli/nemo_ux_nemo_save_meta_check, akoumparouli/nemo_ux_optim_states, akoumparouli/nemo_ux_remove_cleanup, akoumparouli/nemo_ux_te_import_fix, akoumparouli/peft_match_by_type, akoumparouli/reference_cycle_detector, akoumparouli/remove_pipeline_dtype_from_megatronstrategy_ctor, akoumparouli/set_model_parallel_attributes_io, akoumparouli/skip_test_data_download, akoumparouli/split_jit_configs, akoumparouli/te_lora_gemm_fork, akoumparouli/undo_t5_test_changes, akoumparouli/update_megatron_gpt_cont_training, aligner/nemotron5, aligner/NIM-24.07, aligner_trtllm, alit/add_sbert, alit/bert_embedding, alit/draffpp, alit/draft_plus, alit/embeddings, alit/fix_bert_convertor, alit/fw_eval_nm5_ux, alit/griffin, alit/griffin_sft_bug, alit/griffin_sft_debug, alit/hyena_ux, alit/hyena_ux_dist_ckpt, alit/jamba, alit/mamba_embedding, alit/n5, alit/nm, alit/nm5_recipe_fix, alit/nm5_ux, alit/nmh_4b, alit/nmh4b, alit/optim_4k, alit/patch_ssm_num_groups, alit/r2.0, alit/r2_fix_mamba_convert, alit/roberta_convert, alit/rope_scale, alit/rope_scale_main-terryk-squash-onto-main, alit/rope_scale_main-terryk-squash-onto-r2.0.0rc1, alit/sbert, alit/sbert_generate, alit/torch_conv1d, alit/trt_byte_tokenizer_fix, amp_restore_fix, aot/async-fix, aot/bert_660m, aot/bert-embedding, aot/bert-recipes, aot/ckpt-load-fix, aot/fix-llama-embedding, aot/gemma-fix, aot/llama31-dapt-nemo2, aot/llama3-conversion, aot/llama4-GHA, aot/llama_nemotron_resume_path, aot/megatron-dist-ckpt-conversion, aot/mistral-variants, aot/mm_apex_gurad, aot/modelopt-export-te-attention, aot/modelopt_spec, aot/nemo2-sc, aot/nemotron-ux, aot/nemo-ux-nemotron-rc2, aot/sc1-hf-lora, ashors/aligner_nemotron5, ashors/chat-dataset-seq-packing, ashors/fix_ckpt_connector_async, ashors/fix-validation-printing, ashors/fsdp_io, ashors/nemo-skills-packing-scripts, ashors/nemo-ux-dist-ckpt-tests, ashors/nemo-ux-documentation, ashors/nemo-ux-drop-optim-states, ashors/nemo-ux-drop-optim-states-async, ashors/nemo-ux-llm-design, ashors/simplify-data-sampler, ashors/te-apex-optional-docs, ashors/te-layernorm, ashors/update-torch-norm-refs, asr-aed-ngpt, asr_normalize, asr_run, asr_tts_models, async-checkpoint-appstate, athitten/add_logprobs, athitten/delay_prefetch, athitten/deploy_nemo2.0, athitten/eval_mnode_runs, athitten/eval_test_bugfix, athitten/eval_test_r2.2.0, athitten/fix_log, athitten/fix_log_ptl2.0, athitten/fix_no_validation, athitten/in-fw-eval-OAI-API-bak, athitten/precision_fix_2.0, athitten/prefix_check_float_limit_val_batches, athitten/reconfigure_post_dataset_build, athitten/resume_pretraining_fix, athitten/save_context_true_r2.0.0, athitten/thunder_examine, athitten/thunder.jit, audio_lang_id, auto_model, averaging_torch_dist, bert1-dep, bert_cp, bert_fix_mcore, bm25, bobchen/deepseek, bobchen/gemma2_trtllm, bobchen/glm, bobchen/gpt_trtllm, bobchen/mamba_hybrid, bobchen/mcore_trtllm, bobchen/mt5, bobchen/qwen, bobchen/safetensor, bobchen/tp, bobchen/trt10, bobchen/trtllm_pytorch, bot/chore/update-changelog-into-ko3n1g/ci/fix-changelog-generator, bot/chore/update-changelog-into-r2.3.0, boxiangw/add-fsdp2-sharding-strategy, boxiangw/automodel-cp-check, boxiangw/automodel-cp-support, boxiangw/automodel-multinode-tut, boxiangw/change-pr-template, boxiangw/deepseek-automodel, boxiangw/mcore-fsdp2, bucketed_sharded, bugfix_for_bufferd_transducers_notebooks, bugfix-noise-perturb, build-bitsandbytes, bump-ci-container--NVIDIA-Megatron-LM-2025-02-10, bump-ci-container--NVIDIA-Megatron-LM-2025-02-13, bump-ci-container--NVIDIA-Megatron-LM-2025-02-14, bump-ci-container--NVIDIA-Megatron-LM-2025-02-15, bump-ci-container--NVIDIA-Megatron-LM-2025-02-16, bump-ci-container--NVIDIA-Megatron-LM-2025-02-17, bump-ci-container--NVIDIA-Megatron-LM-2025-02-18, bump-ci-container--NVIDIA-Megatron-LM-2025-02-21, bump-ci-container--NVIDIA-Megatron-LM-2025-02-22, bump-ci-container--NVIDIA-Megatron-LM-2025-02-27, bump-ci-container--NVIDIA-Megatron-LM-2025-02-28, bump-ci-container--NVIDIA-Megatron-LM-2025-03-10, bump-ci-container--NVIDIA-Megatron-LM-2025-03-11, bump-ci-container--NVIDIA-Megatron-LM-2025-03-12, bump-ci-container--NVIDIA-Megatron-LM-2025-03-15, bump-ci-container--NVIDIA-Megatron-LM-2025-03-16, bump-ci-container--NVIDIA-Megatron-LM-2025-03-17, bump-ci-container--NVIDIA-Megatron-LM-2025-03-18, bump-ci-container--NVIDIA-Megatron-LM-2025-03-19, bump-ci-container--NVIDIA-Megatron-LM-2025-03-20, bump-ci-container--NVIDIA-Megatron-LM-2025-03-21, bump-ci-container--NVIDIA-Megatron-LM-2025-03-22, bump-ci-container--NVIDIA-Megatron-LM-2025-03-23, bump-ci-container--NVIDIA-Megatron-LM-2025-03-24, bump-ci-container--NVIDIA-Megatron-LM-2025-03-25, bump-ci-container--NVIDIA-Megatron-LM-2025-03-26, bump-ci-container--NVIDIA-Megatron-LM-2025-04-04, bump-ci-container--NVIDIA-Megatron-LM-2025-04-08, bump-ci-container--NVIDIA-Megatron-LM-2025-04-09, bump-ci-container--NVIDIA-Megatron-LM-2025-04-10, bump-ci-container--NVIDIA-Megatron-LM-2025-04-11, bump-ci-container--NVIDIA-Megatron-LM-2025-04-12, bump-ci-container--NVIDIA-Megatron-LM-2025-04-13, bump-ci-container--NVIDIA-Megatron-LM-2025-04-14, bump-ci-container--NVIDIA-Megatron-LM-2025-04-15, bump-ci-container--NVIDIA-Megatron-LM-2025-04-16, bump-ci-container--NVIDIA-Megatron-LM-2025-04-17, bump-ci-container--NVIDIA-Megatron-LM-2025-04-18, bump-ci-container--NVIDIA-Megatron-LM-2025-04-19, bump-ci-container--NVIDIA-Megatron-LM-2025-04-20, bump-ci-container--NVIDIA-Megatron-LM-2025-04-22-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-23-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-24-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-04-24-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-25-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-04-25-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-26-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-04-26-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-27-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-04-27-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-28-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-04-28-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-29-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-04-29-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-04-30-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-04-30-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-05-01-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-05-01-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-05-02-weekly-bump-r2.3.0, bump-ci-container--NVIDIA-Megatron-LM-2025-05-03-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-05-04-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-05-05-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-05-06-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-05-07-weekly-bump-main, bump-ci-container--NVIDIA-Megatron-LM-2025-05-08-weekly-bump-main, bump-resil, cache_aware_ssl, canary, canary-2, canary-2-before-squash, canary-bpe-tokenizer, canary_buffer_infer, canary-fsdp2, canary-natural-prompt, canary-natural-prompt-eval, canary-temp-sampling, changelog-r2.2.0, change_rope_fusion_default, charlie-test, chatusd, chatusd-hcs, chatusd_kl-div, chatusd-tiktoken_240729, chatusd-tiktoken_240729_kl-div, chatusd-tiktoken_240729_kl-div_david, chatusd_tiktoken_kldiv_mirror, chcui/alt_packed_seq_format, chcui/averaged_metric_bug_fix, chcui/conversion_precision_on_24.03, chcui/deepseek_debug, chcui/deepseek_perf, chcui/finetune_recipe, chcui/fix_ptuning_O2, chcui/gemma2, chcui/gemma2_fix, chcui/gemma-pp, chcui/llama_fast_tok, chcui/lookahead_attention, chcui/mcore_peft_hooks, chcui/mcore_t5, chcui/nemotron5_support, chcui/num_tokens_loss, chcui/starcoder_conversion, chcui/t5_peft_refactor, chcui/te_free_peft, chcui/update_mm_dataprep_tut, check-lint, cherrypick-11587, cherry-pick-12209-r2.2.0, cherry-pick-12238-r2.2.0, cherry-pick-12242-r2.2.0, cherry-pick-12276-r2.2.0, cherry-pick-12377-r2.2.0, cherry-pick-12396-r2.2.0, cherry-pick-12415-r2.2.0, cherry-pick-12417-r2.2.0, cherry-pick-12424-r2.2.0, cherry-pick-12469-r2.2.0, cherry-pick-12506-r2.2.0, cherry-pick-12532-r2.2.0, cherry-pick-12559-r2.2.0, cherry-pick-12819-r2.3.0, cherry-pick-12839-r2.3.0, cherry-pick-12898-r2.3.0, cherry-pick-12905-r2.3.0, cherry-pick-12947-r2.3.0, cherry-pick-13012-r2.3.0, cherry-pick-13028-r2.3.0, cherry-pick-13029-r2.3.0, cherry-pick-13040-r2.3.0, cherry-pick-13073-r2.3.0, cherry-pick-13113-r2.3.0, cherry-pick-13124-r2.3.0, cherry-pick-13217-r2.3.0, cherry-pick-13306-r2.3.0, cherry-pick-13339-r2.3.0, cherrypick_mlperf_prs, chery-pick-main-c2daa916b6, chime_sde, chsieh/aquarium-v24.03.01, chsieh/aquarium-v24.07, chsieh/fix_gpt_sft_checkpt, chsieh/gpt-scrolls, chtruong/2.3.0-cherry-pick-pt3, chtruong/add-link-check-false-positives, chtruong/cherry-pick-13350-r2.3.0, chtruong/cherry-pick-ci-r2.3.0-pt1, chtruong/cherry-pick-ci-r2.3.0-pt2, chtruong/cherry-pick-llama4-recipe-fix, chtruong/ctc_segmentation, chtruong/debug-flux-tests, chtruong/debug-my-vm, chtruong/debug-optimizer-test, chtruong/debug-runners, chtruong/docparam, chtruong/fix-trt-llm, chtruong/llama4-optional-test, chtruong/llama4_recipe_test, chtruong/main-copy, chtruong/r2.3.0-cherry-pick-pt4, chtruong/r2.3.0-cherry-pick-pt5, chtruong/r2.3.0-cherry-pick-pt6, chtruong/r2.3.0-cherry-pick-pt7, chtruong/r2.3.0-cherry-pick-pt8, chtruong/r2.3.0-cherry-picks-pt10, chtruong/r2.3.0-cherry-picks-pt9, chtruong/test-approval-2, chtruong/test-lfs, chtruong/vllm_update_0_8_5_ci, ci/bump-2.1.0rc1, ci/bump-2.2.0rc3.dev0, ci/bump-2.3.0rc0.dev0, ckpt_convert_script_refactor, ckpt_convert_script_refactor_r1.23.0, ckpt-safe-deserialize, cleanup_vad_tutorial, codec_inference, colt_ul2, concat_samples, Conf_no_attention, conformer_rms_norm, contrastive_sft, contrastive_sft2, conv1d2lin, convert_llama_to_hf_fixes, core_safetensors, cosmos_tokenizer, cye/auto-mcore-fsdp2-dtensor, cye/auto-mcore-fsdp2-exp, cye/hyena-gpt-infer-context, dataloader-inspector, dataloader-prefetcher, davidm/23.08, davidm/addingPRto115, davidm/beam_search, davidm/cherrypick_r1.15.0, davidm/cherrypick_r1.16.0, davidm/cherrypick_r1.19.0, davidm/docker-fix-test, davidm/gpt_cuda_error_1.14.0, davidm/gpt_sft_dataset_fix, davidm/lora_ct, davidm/megatron_export_float16module_update, davidm/megatron_export_triton_update, davidm/megatron_nmt_finetune, davidm/nmt_beam_search, davidm/peft_server, davidm/peft_server_1.19, davidm/peft_server_fix, davidm/PTL-1.8, davidm/r1.13.0_cherrypick, davidm/t5_encoder_type, davidm/t5_encoder_type_fix, davidm/t5_peft, debug-zarr, degert/fix-sft-clamp, degert/sampler-limit-fix, detached, detached2, detached3, dev/simiao_zhang/knob_onlinesampling, dgalvez/speedup-batched-hyps-to-hyps, didow/sd_fp8, diffusion, dist-ckpt-bug-debugging, distributed_checkpoint, docs_tn, Doc_update_win, donghyukc/ci_base_upgrade, donghyukc/export_test, donghyukc/nemo_run_req, donghyukc/pip_install, donghyukc/pip_install_cleanup, donghyukc/resil_requirement, donghyukc/setuptools_pin, donghyukc/te_patch_update, donghyukc/te_ub_update, donghyukc/update_requirements_nlp, dont_name_branch_using_release, dp_narrative_eval, dpykhtar/23.11_memory_fix, dpykhtar/24.09_drop_layers, dpykhtar/change_dcp, dpykhtar/convert_mlm, dpykhtar/deprecate_non_mcore, dpykhtar/drop_layers, dpykhtar/drop_layers_main, dpykhtar/fused_adam_fix, dpykhtar/llama3_8b_24.09_train, dpykhtar/mcore_dataset_changes, dpykhtar/memory_fix, dpykhtar/mistral_12b_24.09, dpykhtar/r1.20-cherrypicked, dpykhtar/remove_deprecated_dialogue, dpykhtar/sft_eval_fix, dpykhtar/sp_tokenizer_fix, dpykhtar/te_import_guards, dpykhtar/tokenizers, duplex-s2s-new, duplex-s2s-new-debugdata, duplex-s2s-new-hfhub, duplex-s2s-new-salm, enc_dec_logprob, end_of_utterance, erastorgueva/nfa_rnnt, export_cherry_pick_eval, FA, fabric/fixes, farhadr/fabric, fastpitch_e2e, fast-specaug, fayejf-patch-1, feat/aws-asr-multi, feat/conformer_triton, feat/diar-eda-pit-vox, feat/diar-eda-pit-vox-bce, feat/diar-eda-pit-vox-bce-curriculum, feat/diar-eda-pit-vox-bce-eda-feat, feat/diar-eda-pit-vox-bce-mem, feat/diar-eda-pit-vox-bce-mem-bf16-logits, feat/expose_rotary, feat/inference_diar_add, feat/llm-reasoning-llama, feat/multi_speaker, feature/support-relative-paths-lhotse, fim, fine_tune_qa, fix_asr-chunk_inference, fix_bert, fix-bitsandbytes, fix_bucketing_bug, fix_c10_Error, fix-checkpoint-deadlocks, fix_cuda_graphs_toggle, fix_finetune, fix_fp8_args_table, fix_fp8_scale_export_in_megatron_to_nemo_converter, fix_fp8_scales_in_convert_ckpt, fix_hf_login_notebook, fix/issues7101, fix_maxutts, fix-megatron-t5-eval, fixModelLoad, fix_nfa_buffered_streaming, fix_nlp_megatron_api, fix-packed-attention-2201, fix-reinstall, fix-sft-dataset-tokenization, fix_special_token, fix/system_vad_diar, fix-t5-trainer-test, fix-t5-trainer-test-m, fix-t5-trainer-test-m1.9, fix/update-audio-to-text-dataset, flan-t5-p-tuning, flan_t5_zero_shot, frame_vad, gemma3_automodel, geshen/8b_exp, geshen/checkpoint, geshen/fix_file_move_on_large_runs, geshen/fix_race_condition, geshen/llama3.1_fix, geshen/llama3_nemo_trt, geshen/nemo_cherrypick, geshen/nemotron5_save_fix, geshennvm/te_update, geshen/rlhf_2308, geshen/save_top_k_hack, geshen/top_k_save, get-checkpoint, gpt-alibi, gpt-alibi-FA-infer-e, gpt-alibi-y, gpt_core_transformer, gpt_core_transformer_dist_ckpt, gpt_spawn, guyueh1/cpu_offload_debug, haifengqian/batch_inf_checkin, hcs-23.11fc_David-inference, hcs-nemofw-training-2311_David, hcs-v1220, heh/add_safetensor, heh/modular_speechlm, heh/modular_speechlm_nightly, heh/speech_conv_dev, heh/speechlm_clap, hemil/always-reload-params, hemil/automodel-custom-loop, hemil/automodel-mvp-v2, hemil/chpk_llava_next, hemil/ckpt-converter, hemil/comm-overlap, hemil/custom-converters-v2, hemil/custom-loop-updates, hemil/custom-sft, hemil/export-tron, hemil/fix-autoresume, hemil/mlm-cfg-refactor, hemil/nemo-run-reqs, hemil/overlap-param-gather, hemil/remove-factory, hemil/update-long-context-recipes, hemil/use-model-parallel-config, hiddens-sampling-fix, huiyingl/expose_flash_decode, huvu/avlm_model, huvu/cherrypick_11584, huvu/mcore_retro, huvu/mcore_retro_deprecation_warning, huvu/mcore_retro_docs, huvu/mcore_retro_eval, huvu/mcore_t5_checklist, huvu/mcore_t5_inference, huvunvidia-patch-1, huvu/oncall_NFS87, huvu/openvla_dataloader, huvu/rag_llamaindex_fix, huvu/rag_pipeline_nemo2.0, huvu/remove_retro, huvu/t5_15T_experiments, huvu/t5_biasmoe_fix, huvu/t5_glue_finetuning, huvu/t5_nemo2.0_3b11b, huvu/t5_nemo2.0_nemorun_example, huvu/t5_nemo2.0_pp, huvu/t5_nemo2.0_pretrain_recipe, huvu/t5_nemo2.0_recipes_update, huvu/t5_test_coverage, huvu/t5_test_mlmcommit, huvu/test_nemoci_t5_alibi, intent_slot_bio_type, io-hf-tokenizer, IPL_mixin, IPL_script, jaeminc/full_te_layer_20240128_autocast_ckpt_fix_mcore_tot, jasonwan/fsdp2, jasonwan/mcore-fsdp2, jbaczek/DataFetcherWrapper_WAR, jbaczek/dummy_branch, jbaczek/llm/ci-debug, jbaczek/llm/fix_war_for_fp8_load, jbaczek/llm/fp8_load_war_update, jbaczek/mcore_parallel_state_api_change, jbaczek/mcore_parallel_state_api_change_2, jbaczek/revert_8847b2e9, jbaczek/te_extra_state_war, jbaczek/test_mcore_dataset, jennifchen/sft_chat_template, jiaqiz/titoken_fix, jiemingz/almost_full_iter_cg, jiemingz/cudagraphs, jiemingz/lora_fixes, jinja_templates, jlasek/add_setup_export_script, jlasek/black_examples, jlasek/build_trtllm_for_qnemo, jlasek/cleanup_unused_imports, jlasek/infer_in_framework_bugfix, jlasek/nemo_autotokenizer_extensions, jlasek/nemo_export_in_framework_test, jlasek/ptq_tests, jlasek/quant_cfg_overrides, jlasek/quantization_docs_update, jlasek/quantization_int4_awq_test, jlasek/r220_relax_modelopt, jlasek/super_register_artifact, jlasek/test_torch_onnx_quant, jlasek/update_quant_docs, jlasek/vllm_github_ci_test, jlasek/vllm_tokenizer_bugfix, jlasek/vllm_v1_update, jm/save-checkpoint-n-steps, Jorjeous-patch-1, jstjohn/alternative_hyena_medium_init, jstjohn/data_sampler_constant_len, jstjohn/iomixin_mutation, jstjohn/wip_convert_vortex_ckpt_to_nemo2, jupinderp/22b_nemo_conversion, jupinderp/mcore-mlm-to-nemo-conversion, karpnv/beam, karpnv/beamsearch, karpnv/beamsearch1, karpnv/ipl, karpnv/nemo_run_ipl, kmorabia/modelopt-refactoring, ko3n1g/build/bump-mcore-2, ko3n1g/build/bump-mcore-46d4069, ko3n1g/build/bump-pyt-25.01, ko3n1g/build/bump-pyt-25.01-2, ko3n1g/build/bump-pyt-25.01-ak, ko3n1g/build/nvrx-arm, ko3n1g/build/pin-setuptools, ko3n1g/chore/bump-25.03, ko3n1g/chore/bump-version, ko3n1g/chore/test-infra, ko3n1g/ci/add-torchrun, ko3n1g/ci/allow-bumpto-bypass, ko3n1g/ci/automodels-examples, ko3n1g/ci/auto-selective-triggering, ko3n1g/ci/bump-modelopt, ko3n1g/ci/cherry-pick-73f5eb5, ko3n1g/ci/cuda-driver-12.8, ko3n1g/ci/enable-codecov-checks, ko3n1g/ci/fix-mcore, ko3n1g/ci/improve-vm-maintenance, ko3n1g/ci/manual-selective-triggering, ko3n1g/ci/nemo-install-light, ko3n1g/ci/restructure-tests, ko3n1g/ci/test-maintenance, ko3n1g/feat/pip-installable, ko3n1g-patch-1, ko3n1g/pranav_deploy_unit_tests, ko3n1g/revert/build, ko3n1g/trt-llm-stage, koel_onlinepo, kpuvvada/speech_lm, lgrigoryan/asr_tutor_bugfix, lgrigoryan/batched_beam_search_pr, lgrigoryan/ctc_beam_decoding, lgrigoryan/ctc_beam_search_pr, lgrigoryan/fix-eval_beamsearch_ngram_ctc, lgrigoryan/pr-batch-beam-search, lgrigoryan/r2.2.0, lgrigoryan/r2.3.0_asr_context_biasing_bugfix, lgrigoryan/rm-redundant-calculations, lgrigoryan/rnnt_batched_beam_search, lgrigoryan/signed_batched_beam_search_for_pr, lhotse-dataloading-tutorial, lhotse-nemo-aligner-jsonl, llama2_tp_overlap_cfg, llava-next, llava_next_fix, llmb-nemo, llmb-nemo-r2.3.0, llm/nemo-1-25, llm_v2, maanug/add-cleanup-util, maanug/bump-pyt-fixes, maanug/cherrypick-missing-r2.0.0, maanug/collect-r200-cherrypicks, maanug/default-eps, maanug/fix-metric-logging-main, maanug/latest-ckpt-api, maanug/modified-forward-step, maanug/move-precision-tests, maanug/neva-example, maanug/new-r200, maanug/perf-recipe-default, maanug/rank0-log-first, maanug/resume-checkpoint-priority-rebased, maanug/resume-checkpoint-prio-test, maanug/resume-tests-validation, maanug/sched-max-steps, maanug/use-builder, maanug/use-builder-new, magpietts_2503_release, main, malay/flops_no_tb, malay/flux_12b, malay/null_tokenizer, malay/peft_pp_debug, malay/perf_scripts, malay/perf_scripts_updates, maxpool_subsampling, mblaz/async-dist-ckpt-minimal-base, mblaz/debug-mlperf-parallel-load, mblaz/expose-pyt-dist-multiproc, mblaz/expose-pyt-dist-multiproc-24.05, mblaz/fix-load-strictness-load-model, mblaz/fix-neva-finetune, mblaz/simplify-save-checkpoint, mcore_convert, MegatronBaseModel-tppp-properties, megatron-core_sentencepiece_quick-fix, megatron_multi_query_attention, megatron_parallel_docs, megatron-step-cleanup, megatron_vit_remove_jit, meister/nfa, mfutrega/cp_comm, mfutrega/cp_comm_type, mfutrega/cp_loss, mfutrega/cp_scaling, mfutrega/lora_2.0, mfutrega/loss, mfutrega/nemo2, mfutrega/sd_fix, mfutrega/sd_mcore_optim, mfutrega/sharp, mingyuanm/add_fp8_support_for_sd, mingyuanm/flux_controlnet_sharded_dict, mingyuanm/flux_fp8, mingyuanm/github_ci_flux_2, mingyuanm/remove_links, mingyuanm/remove_sdxl_tutorial, mingyuanm/sd_fp8, mingyuanm/update_data_module, mlm-pretrain-loop, modular_speechllm, modular_speechllm_asrset, modular_speechllm_heh, modular_speechllm_r1.21, modular_speechllm_r1.21_run, ms_diar_decoder, msekoyan/aistore_data_fix_2, msekoyan/oomptimizer_with_max_tps, msekoyan/punctuation_in_timestamps, msekoyan/_skipme_fix, msekyan/for_oci, multimodal-prompt-formatter, multimodal-prompt-formatter-draft, mwawrzos/fix-torch-inductor-pytorch-ToT, neftune, nemo20-e2elargerun-090824, nemo20-e2elargerun-091124, nemo20-e2elargerun-160824, nemo_ckpt, nemo/collections/nlp/models/language_modeling/megatron_gpt_sft_model.py, nemo_evaluator_run-uni_tok, nemo_experiments, nemo_exp_gptevalchanges, nemo_run_ipl, nemo_speech_codec, nemotron_ddp_fix, nemo-ux/parallelism-bug, nfa_conf_metrics, ngpt_decoder, ngpt_encoder_update_rope, nim_griffin, nkant/gpt_sft, nkant/gpt_sft_stable, noise_emb, no_length_clip, ntajbakhsh/landingpages, nvte-attn, online_diar_core_p2, online_streaming_vad_asr, online_translation, onur/apply-te-transformer-layer, onur/export-deploy-unit-tests, onur/inframework_deploy_test, onur/nemo2_inframework_support, onur/nvembed-onnx-support, onur/onnx-trt-export-tests, onur/remove-export-deploy, onur/te-transformer-layer, onur/trt-llm-tests, oomptimizer-dirty, oomptimizer-fix-val, original_working_llama_to_hf_fixes, p1, pablo-garay-dev-container-bug-report, pablo-garay-patch-TE-update-11-27-24, pablo-garay/r2.0.0_cherrypick, pablo-garay-r2.1.0-update-ptl, pablo-garay-r2.1.0-update-ptl-1, pagaray/bugfix_make_radttsOptional, pagaray/ci_flaky_test_optional, pagaray/for_test_purpose_only_r2.0.0_mcoreUpdate, pagaray/nemo_cicd_2, pagaray/nemo_cicd_flaky_test_part3, pagaray/nemo_cicd_flaky_test_part4, pagaray/nemo_cicd_part17, pagaray/nemo_cicd_part18, pagaray/nemo_cicd_part24, pagaray/nemo_cicd_part4, pagaray/nemo_cicd_part6, pagaray/nemo_cicd_previous_old, pagaray/nemo_cicd_update_PT24.04, pagaray/nemo_cicd_v2, pagaray/optional_flaky_test, pagaray/reduce_storage_usage_in_favor_of_local_storage_part3, pagaray/update_PT_24.03, pagaray/update_PT_24.05, partial-pytorch-fixes, pc_iwslt, peft_inference, perplexity_eval, phonemes_from_file, physicalai, pikaminski/bugfix_inference_seqlen, pikaminski/enable_hf_export_in_nemo_export, pikaminski/hf_import_test, pikaminski/nemo_export_megatron_dependency_fix, pikaminski/ptq_parallel_config, pikaminski/vllm_export_via_hf, pikaminski/zarr_0_3_bump, pip-install, pmannan/neva2_seq_packing, pmannan/neva_cp_seq_packing, pmannan/neva_etp_epp_fix, pmannan/neva_flops_calculation, pmannan/neva_perf_scripts, pmannan/thd_cp_fix_neva, pmannan/vit10b_llama370b_benchmarking, pmannan/vlm_llama_test, pr-10913, pranav_deploy_inference_path, pranav-huggingface-export-trtllm, pranav_ray_serve, pranav/tensorrt_export_huggingface, pre-commit-ci-update-config, prompt_infer_fix, pr/test-cov-audio-losses, ptuning_intent_slot, pytest_coverage, QN20, quantize-fastpitch, r00000_test, r1.0.9, r1.10.0, r1.10.0-megamolbart, r1.10.0-megamolbart-ea2, r1.11.0, r1.11.0-temp, r1.11.1, r1.12.0, r1.13.0, r1.13.0docfixes, r1.13.1, r1.13.badval, r1.14.0, r1.15.0, r1.15.0_bf_tut, r1.15.0_bf_tut1, r1.15.0_docs_g2p, r1.16.0, r1.17.0, r1.17.0_pt_23.04, r1.17.0_pt_23.04_nan, r1.18.0, r1.18.1, r1.19.0, r1.19.1, r1.20.0, r1.20.0_pt_23.09, r1.20.0_pt_23.09_sharp, r1.21.0, r1.21.1, r1.22.0, r1.22.0-hiddnens-fix, r1.23.0, r1.23.0_mm_fix, r1.23.0_update_k2, r1.3.0, r1.3.1, r1.4.0, r1.5.0, r1.5.0-nltkfix, r1.5.1, r1.6.0, r1.6.1, r1.6.2, r1.7.0, r1.7.1, r1.7.2, r1.8.0, r1.8.1, r1.8.1_bugfix, r1.8.1_fastpitch_bugfix, r1.8.2, r1.9.0, r2.0.0, r2.0.0rc0, r2.0.0.rc0.beta, r2.0.0rc0_cuda_graphs_default_only_inference, r2.0.0rc1, r2.1.0, r2.1.0_fix_cicd, r2.1.1, r2.2.0, r2.3.0, refactor-speechllm-prompt-formatter, release/NIM-24.06, release/NIM-24.08, remove_tn, replace_text, retrieval, retro_lora, retro_lora_prompt_learning_megatron, retro-prospero-demo, retro_ptuning, revert-10643-yuya/update_megatron_parallel, revert-10693-partial_distopt, revert-10815-alit/r2_cherry_pick, revert-11739-huvu/t5_hf_convert, revert-11791-pablo-garay/revert_regression, revert-13485-aanoosheh/intermediates-distill, revert-8546-pablo-garay-update-mcore-version, revert-modelopt-nemo-2, rm_spawn, rnnt_greedy_torch_jit, rope-asr-aed-ngpt, rope_interpolate_e, rywolf/custom-order, rywolf/data-sampler, rywolf/multi-dc-docs, safe_bn, salm_t5_tts_2406, sandeepsub/gpt_continue_training, sandeepsub/gpt_sft, sandeepsub/gpt_sft_stable, sandeepsub/ul2_gpt, SDE_bugfix, sde_transcribe_29-09, sec_0.01, set_ptune_val_bs, set_use_pytorch_sdpa, sft_scripts, sft-scrolls-v2, shanmugamr1992-patch-1, sharatht/nmh_qad, shman_conf, shriya/update_mcore_overlap_params, sichu/bert_660m_nsys, slym/per_mb_loader, slym/r1.17.0_pt_23.04_nan, slym/sharp, slym/sharp_r1.20.0, soft-scrolls, solu/chat-dataset-changes, speaker_beam_general_overlap_wsj_conformermask_librimix_fix, spectrogram-enhancer, speechllm-develop-gen, speechllm-develop-gen-align, speechllm-develop-gen-duplex, speechllm-develop-merge-main-27nov24, speechllm_tts, speechllm_tts_2406_squashed, split-shards-lhotse, ssl_contrastive_time_merge, ssl_titanet, st2023, st2023_ev, st2023_ev_valid_flag, st2023_r1.16.0, st_attn_sampled_softmax, streaming-asr-aed-ngpt, streaming_mulspk_asr, st_tutorial, subhankarg/imagebind, subhankarg/speechllm, subhankarg/speechllm_main, support_older_config_msdd, support_partial_hypotheses, support_server_logprobs, support-sharded-nontarred-manifest-lhotse, swa, tango4j/llm_bsd, tdt, te_cuda_graph_support, te-patch, terryk/aligner-nemo2-export-changes, terryk/aligner-v12-changes, terryk/efficient-ckpt-fix, terryk/efficient-ckpt-fix-old, terryk/efficient-ckpt-fix-rebased, terryk/fix-lora-merge-by-loosening-ckpt-strictness, terryk/fix-nemo2-nemo1-converter, terryk/hemil/automodel-custom-loop-with-sahil-patch, terryk/llama-export-hf-missing-conf, terryk/mcore-converter-fix, terryk/r2.0.0rc1, terryk/reshard-extra-state-fp8-tensor-handling, terryk/v11-refit, test-alt-ci, test/asr-test-flakiness-2310, test_ci_action, test_jenkins, test_llama2_fixes, tkonuk/lora_mlp, tkonuk/nemotron5_tokenizer, tkonuk/nkb_adapter, tkonuk/pixtral, tkonuk/pixtral_mcore_optim, tkonuk/pixtral_trt, tkonuk/tiktoken, tk/static-sync-func, token_collapse_expr, toxicity_classifier, transducer_with_transformer, tsasr_12.11, tsasr_spec, tts_adapters, tts_adapters_cln, tts_annotation_before_merge, tts_clearml_logging, tts_inf, tts_rnnt, tts_tar_dataset, tts_tutorial_fix, ul2_debug, universal_gradient_boosting, universal_instruction, universal_old_sft, universal_sft, update_black_incremental_v2, update_black_test3, update-modelopt, upgrade_pytorch_container, val_packing_fix_cg, vfm, weekly-bump, weekly-bump-main, weekly-bump-r2.3.0, wip_ab/clip, wip_ab/openvla, wip_ab/vlm_automodel, xueyang-main-mix, xueyang/tts_new_transformer, yangzhang/automodel_bert, yangzhang/automodel-custom-bert, yash/chpk_llava_next, yash/dev_llava_next, yash/energon_datamodule_refactor, yash/llava_next_data, yash/llava_next_interleaved, yash/mimo_rebased, yash/mimo_wip, yash/ranking_fwd_pass_example, yash/test_sequence_packing, yifu/deepseek, yifu/deepseek_clean, yifu/deepseek_debug, yuya/2403_neva_patch, yuya/add_bert_hf_converter, yuya/add_neva_cp, yuya/add_neva_nemo1_to_nemo2_conversion, yuya/clip_to_mcore, yuya/fix_dreambooth_data, yuya/fix-neva-sp-seqpacking, yuya/fix_softmax_fallback, yuya/llama_31_dapt, yuya/llama4_debug1, yuya/llama4_long_context, yuya/llama4_tron, yuya/loss_per_token, yuya/mllama_config_fix, yuya/neva_2.0_update_examples, yuya/neva2_seq_packing, yuya/neva_epp_etp, yuya/neva_open_clip, yuya/neva_seq_pack, yuya/neva_update, yuya/r1.12.0_cherrypick, yuya/r1.13.0_cherrypick, yuya/r1.23.0_mm_patch, yuya/r2.0.0rc1_fix, yuya/rename_o2_to_O2, yuya/update_fusion_doc, yuya/update_neva_data_name, yuya/update_neva_doc, yuya/update-transfomers-version, yuya/vlm_api, yz/dev/49B, yz/dev/activation-ckpting, yz/dev/activation-ckpting-v2, yz/dev/akoumparouli/automodel_grad_ckpt, yz/dev/akoumparouli/fspd2_offload_policy_v2, yz/dev/automodel/bench, yz/dev/cut-loss-function, yz/dev/error_msg, yz/dev/expose-gradient-clip, yz/dev/hf/70b-hellaswag, yz/dev/hf-pretrain, yz/dev/linear-loss-v2, yz/dev/linear-loss-v3, yz/dev/peft-fsdp2, yz/dev/pretrain, yz/dev/sft-fsdp2-ckpting, yz/dev/test, zero_sc_eval, zero_sc_eval_ch, zero_sc_eval_e, zshao/add_callback_group
- Tags: 1.4.0, 24.09-alpha.rc0, 25.04-alpha.rc1, 25.04-alpha.rc2, nvidia-mlperf, r2.0.0rc1, stable, v1.10.0, v1.11.0, v1.12.0, v1.13.0, v1.14.0, v1.15.0, v1.16.0, v1.17.0, v1.17.0_pt_23.04, v1.18.0, v1.18.1, v1.19.0, v1.19.0_mm, v1.19.1, v1.20.0, v1.21.0, v1.22.0, v1.23.0, v1.3.0, v1.4.0, v1.5.0, v1.5.1, v1.6.0, v1.6.1, v1.6.2, v1.6.temp, v1.7.0, v1.7.1, v1.7.2, v1.8.0, v1.8.1, v1.8.2, v1.9.0, v2.0.0, v2.0.0rc0, v2.0.0.rc0.beta, v2.1.0, v2.1.0rc0, v2.1.0rc1, v2.1.0rc2, v2.2.0, v2.2.0rc0, v2.2.0rc1, v2.2.0rc2, v2.2.0rc3, v2.2.1, v2.3.0, v2.3.0rc2, v2.3.0rc3, v2.3.0rc4
Files
This documentation is generated from this file: