Datasets:

---

renwei2024

/

nemo-github-issues

like 0

[]

[]

[]

[]

[]

[ "In FilterbankFeatures class we extend time length features to multiples of `pad_to` value here 16(default) for faster processing. Try changing that to -1 and try again. " ]

[]

[]

[ "This pull request **introduces 2 alerts** and **fixes 2** when merging 072d52065e9037808be763b0750644bfd8af655b into cc4bfc06993326fcd8bb978460754b89026341ff - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-f732130f599d696fda78aa55922c86c8d40e6c8a)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 4e3b807f8c121a71fb279e8b4502d749bd169d1a into cc4bfc06993326fcd8bb978460754b89026341ff - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-db6028c8702367473549e3fb741cb7c1d2b64aca)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 231a1382d1bb297b6ae07059296be821b955e906 into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-662e258b2346b26bacff9996a324c051934957f0)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 22db673e461619030f66935b86bacc6fa509689e into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-d58b56ae97f994ad0477474b742f9a8b6c2fe000)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 223d341bdcb2a27eba566ca50a6d07a2c2642da2 into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-3a192f36101dadffacd6c56f4a209cb1a52013c4)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 3b452eb2b69de885fb1ee7e07bd370124533c3ac into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-408dbd2c4cb053671ed03d06527ddb82b1f99730)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 7005c8da934ea0f468da93c86e62c74e62a6a685 into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-2d040971cf6149163bfdc19c74da5533224f3fa9)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 7ee9da912d46fce1a4115c5f6ec65fd86a5a4e19 into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-224a06a1a6f50988202283354429ccf4ee59c9cb)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging c17c56c0bf6cd8d41521283eac05fe484dc7a771 into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-080f8478a274fb0d0d6ddb7588d53740dcad8fcf)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 7f45d809f112c81b6c602fa133bae461e8418259 into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-9f8aeb6559c160e8f60d361c720ea6343f2c84d2)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 4030eea1fab14407f4c9d91ffeea83f708fa9a4c into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-560f86f8dd634a83f1b74a43ceb1035ee16e3bb9)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 32825ead739c2b65f8567b874168e7c2db0c51a6 into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-8f9d38b08b22e21efed33e8bf906a7820a9b4ca4)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** and **fixes 2** when merging 51fb1b4200be65e3cb3ac5aae769a936101f811d into 1666139b6b3334dc8f699f69d5a97aec4fe53419 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-9d71946af76dda9f3124239d4097559fbcc3514f)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import" ]

[]

[]

[]

[]

[ { "login": "borisfom", "id": 14189615, "node_id": "MDQ6VXNlcjE0MTg5NjE1", "avatar_url": "https://avatars.githubusercontent.com/u/14189615?v=4", "gravatar_id": "", "url": "https://api.github.com/users/borisfom", "html_url": "https://github.com/borisfom", "followers_url": "https://api.github.com/users/borisfom/followers", "following_url": "https://api.github.com/users/borisfom/following{/other_user}", "gists_url": "https://api.github.com/users/borisfom/gists{/gist_id}", "starred_url": "https://api.github.com/users/borisfom/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/borisfom/subscriptions", "organizations_url": "https://api.github.com/users/borisfom/orgs", "repos_url": "https://api.github.com/users/borisfom/repos", "events_url": "https://api.github.com/users/borisfom/events{/privacy}", "received_events_url": "https://api.github.com/users/borisfom/received_events", "type": "User", "user_view_type": "public", "site_admin": false } ]

[]

[]

[]

[]

[]

[ { "login": "MaximumEntropy", "id": 9114321, "node_id": "MDQ6VXNlcjkxMTQzMjE=", "avatar_url": "https://avatars.githubusercontent.com/u/9114321?v=4", "gravatar_id": "", "url": "https://api.github.com/users/MaximumEntropy", "html_url": "https://github.com/MaximumEntropy", "followers_url": "https://api.github.com/users/MaximumEntropy/followers", "following_url": "https://api.github.com/users/MaximumEntropy/following{/other_user}", "gists_url": "https://api.github.com/users/MaximumEntropy/gists{/gist_id}", "starred_url": "https://api.github.com/users/MaximumEntropy/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/MaximumEntropy/subscriptions", "organizations_url": "https://api.github.com/users/MaximumEntropy/orgs", "repos_url": "https://api.github.com/users/MaximumEntropy/repos", "events_url": "https://api.github.com/users/MaximumEntropy/events{/privacy}", "received_events_url": "https://api.github.com/users/MaximumEntropy/received_events", "type": "User", "user_view_type": "public", "site_admin": false }, { "login": "okuchaiev", "id": 22035961, "node_id": "MDQ6VXNlcjIyMDM1OTYx", "avatar_url": "https://avatars.githubusercontent.com/u/22035961?v=4", "gravatar_id": "", "url": "https://api.github.com/users/okuchaiev", "html_url": "https://github.com/okuchaiev", "followers_url": "https://api.github.com/users/okuchaiev/followers", "following_url": "https://api.github.com/users/okuchaiev/following{/other_user}", "gists_url": "https://api.github.com/users/okuchaiev/gists{/gist_id}", "starred_url": "https://api.github.com/users/okuchaiev/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/okuchaiev/subscriptions", "organizations_url": "https://api.github.com/users/okuchaiev/orgs", "repos_url": "https://api.github.com/users/okuchaiev/repos", "events_url": "https://api.github.com/users/okuchaiev/events{/privacy}", "received_events_url": "https://api.github.com/users/okuchaiev/received_events", "type": "User", "user_view_type": "public", "site_admin": false } ]

[ "This pull request **introduces 1 alert** when merging c5907d9e737d91da1be9f0dd3a1bde351feb6325 into bb84595e2811f374c93866ee50d65518fc33faed - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-d35a3ab6e82c4b3a3cba216f1d3e26e53f98351e)\n\n**new alerts:**\n\n* 1 for Wrong name for an argument in a class instantiation", "This pull request **introduces 1 alert** when merging 09b8ef0bbc2e3af89faa8e6e7aad02e67b5f7c21 into bb84595e2811f374c93866ee50d65518fc33faed - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-1bd20499f4ea178421f0b93b73dd15db5d80181d)\n\n**new alerts:**\n\n* 1 for Wrong name for an argument in a class instantiation", "This pull request **introduces 1 alert** when merging 5b8958dabb847c937c10cd3a8ecdc9a014f005cc into bb84595e2811f374c93866ee50d65518fc33faed - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-67a157eeb9c41fbd5329479d54ac1a2e1c27325e)\n\n**new alerts:**\n\n* 1 for Wrong name for an argument in a class instantiation" ]

[]

[]

[ "This pull request **introduces 2 alerts** when merging eeeebf02faa79e72f4921a91517cba26173669a1 into 16b87d1f7d22126b8eb471ca0538f75ade9620cf - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-aa22cf3bd77bdd09b203bef1f11fdcc603a8ffeb)\n\n**new alerts:**\n\n* 1 for Unused local variable\n* 1 for Unused import" ]

[ { "id": 1485815557, "node_id": "MDU6TGFiZWwxNDg1ODE1NTU3", "url": "https://api.github.com/repos/NVIDIA/NeMo/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": "Something isn't working" } ]

[]

[ "``.restore_from`` only works with ``.nemo`` files. You can try loading regular Pytorch weights via regular Pytorch load_state_dict methods because any NeMo model is a Pytorch module", "Thank you very much ! ", "this error occurs when ```restore_from``` with ```.nemo ``` files.\r\n```sh\r\nasr_model = nemo_asr.models.EncDecCTCModelBPE.restore_from(\r\n restore_path=os.path.join(MODELS_URL, \"stt_en_conformer_ctc_xlarge.nemo\"),\r\n map_location=device,\r\n)\r\n```\r\nI tried restore_path with this code. But this occurs this error:\r\n ```sh\r\nnot a gzip file\r\n```\r\nand\r\n```sh\r\nCould not load English CTC model\r\n[NeMo I 2024-01-03 15:43:46 tokenizer_utils:125] Getting HuggingFace AutoTokenizer with pretrained_model_name: bert-base-uncased, vocab_file: /tmp/tmpy2yvwun8/tokenizer.vocab_file, special_tokens_dict: {}, and use_fast: False\r\n[NeMo W 2024-01-03 15:43:47 modelPT:135] If you intend to do training or fine-tuning, please call the ModelPT.setup_training_data() method and provide a valid configuration file to setup the train data loader.\r\n Train config : \r\n use_tarred_dataset: false\r\n label_info_save_dir: null\r\n text_file: text_train.txt\r\n labels_file: labels_train.txt\r\n tokens_in_batch: null\r\n max_seq_length: 128\r\n num_samples: -1\r\n use_cache: true\r\n cache_dir: null\r\n get_label_frequences: false\r\n verbose: true\r\n n_jobs: 0\r\n tar_metadata_file: null\r\n tar_shuffle_n: 1\r\n shuffle: true\r\n drop_last: false\r\n pin_memory: true\r\n num_workers: 8\r\n persistent_workers: true\r\n ds_item: punct_dataset_complete\r\n \r\n[NeMo W 2024-01-03 15:43:47 modelPT:142] If you intend to do validation, please call the ModelPT.setup_validation_data() or ModelPT.setup_multiple_validation_data() method and provide a valid configuration file to setup the validation data loader(s). \r\n Validation config : \r\n use_tarred_dataset: false\r\n label_info_save_dir: null\r\n text_file: text_dev.txt\r\n labels_file: labels_dev.txt\r\n tokens_in_batch: null\r\n max_seq_length: 128\r\n num_samples: -1\r\n use_cache: true\r\n cache_dir: null\r\n get_label_frequences: false\r\n verbose: true\r\n n_jobs: 0\r\n tar_metadata_file: null\r\n tar_shuffle_n: 1\r\n shuffle: true\r\n drop_last: false\r\n pin_memory: true\r\n num_workers: 8\r\n persistent_workers: true\r\n ds_item: punct_dataset_complete\r\n \r\n[NeMo W 2024-01-03 15:43:47 modelPT:148] Please call the ModelPT.setup_test_data() or ModelPT.setup_multiple_test_data() method and provide a valid configuration file to setup the test data loader(s).\r\n Test config : \r\n use_tarred_dataset: false\r\n label_info_save_dir: null\r\n text_file: text_dev.txt\r\n labels_file: labels_dev.txt\r\n tokens_in_batch: null\r\n max_seq_length: 128\r\n num_samples: -1\r\n use_cache: true\r\n cache_dir: null\r\n get_label_frequences: false\r\n verbose: true\r\n n_jobs: 0\r\n tar_metadata_file: null\r\n tar_shuffle_n: 1\r\n shuffle: true\r\n drop_last: false\r\n pin_memory: true\r\n num_workers: 8\r\n persistent_workers: true\r\n ds_item: punct_dataset_complete\r\n \r\n[NeMo W 2024-01-03 15:43:47 nlp_overrides:208] Apex was not found. Please see the NeMo README for installation instructions: https://github.com/NVIDIA/NeMo#megatron-gpt.\r\n[NeMo W 2024-01-03 15:43:47 modelPT:1088] World size can only be set by PyTorch Lightning Trainer.\r\n[NeMo W 2024-01-03 15:43:47 nlp_model:63] Apex was not found. Please see the NeMo README for installation instructions: https://github.com/NVIDIA/NeMo#megatron-gpt.\r\n[NeMo W 2024-01-03 15:43:47 punctuation_capitalization_model:681] The artifact `class_labels.punct_labels_file` was not found in checkpoint. Will rely on `punct_label_ids` parameter\r\n[NeMo W 2024-01-03 15:43:47 punctuation_capitalization_model:703] The artifact `class_labels.capit_labels_file` was not found in checkpoint. Will rely on `capit_label_ids` parameter\r\n[NeMo W 2024-01-03 15:43:47 modelPT:202] You tried to register an artifact under config key=tokenizer.vocab_file but an artifact for it has already been registered.\r\nError(s) in loading state_dict for PunctuationCapitalizationModel:\r\n Unexpected key(s) in state_dict: \"bert_model.embeddings.position_ids\". \r\nCould not load English NLP model\r\nnot a gzip file\r\nCould not load Russian CTC model\r\nnot a gzip file\r\nCould not load Russian NLP model\r\nSystem check identified no issues (0 silenced).\r\n```\r\nI re-download .nemo models from official website [Nemo Models](https://catalog.ngc.nvidia.com/models)\r\nBut it still throw that error,\r\n \r\nhow to fix this error ???", "Please own a new thread instead of responding to a closed one. \r\n\r\n---\r\n\r\nThis is an error not in asr but in the Punctuation capitalization model - \r\n\r\nError(s) in loading state_dict for PunctuationCapitalizationModel:\r\n Unexpected key(s) in state_dict: \"bert_model.embeddings.position_ids\". \r\n\r\n```\r\n[NeMo W 2024-01-03 15:43:47 nlp_model:63] Apex was not found. Please see the NeMo README for installation instructions: https://github.com/NVIDIA/NeMo#megatron-gpt.\r\n[NeMo W 2024-01-03 15:43:47 punctuation_capitalization_model:681] The artifact `class_labels.punct_labels_file` was not found in checkpoint. Will rely on `punct_label_ids` parameter\r\n[NeMo W 2024-01-03 15:43:47 punctuation_capitalization_model:703] The artifact `class_labels.capit_labels_file` was not found in checkpoint. Will rely on `capit_label_ids` parameter\r\n```" ]

[ { "id": 1485815557, "node_id": "MDU6TGFiZWwxNDg1ODE1NTU3", "url": "https://api.github.com/repos/NVIDIA/NeMo/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": "Something isn't working" } ]

[]

[ "You can use from_pretrained(..., map_location=\"cpu\")", "closing, please re-open is @titu1994 's suggestion did not help." ]

[]

[]

[ "just to confirm that this is not a bugfix for rc1? ", "@okuchaiev Part of this (Point 2) should be fixed in RC but this might raise conflict and not very essential (just logging), so we decide to merge it to main. I could split it and merge to RC if it's better. ", "Split this PR into two please. Bugfix to RC, logging to main. You can revert the RC patch from this PR and do it in another or vice-versa.", "@titu1994 \r\nMoved the following to RC in https://github.com/NVIDIA/NeMo/pull/1878. Could you please review it? \r\n1. Change self.log(tensorboard_logs) and remove the 'log' part in the return requested in PR **before regresion**\r\n2. Fix for logging for top-k acc metrics\r\n\r\nLet wait for RC to merge to main and I will update self.log and bug fix for **regression** to avoid any conflict.\r\n", "Close this since first part has been merged in https://github.com/NVIDIA/NeMo/pull/1878 and second part https://github.com/NVIDIA/NeMo/pull/1907 is ready for review" ]

[]

[]

[]

[]

[]

[]

[]

[]

[ "Overall it's good to go from me", "this change broke training on draco" ]

[]

[]

[]

[]

[]

[ "> ```python\r\n> change_decoding_strategy\r\n> ```\r\n\r\n+ Can we call the change_decoding_strategy automatically in the transcribe method if necessary so that user does not need to do that?\r\n\r\n+ Why do we have setting preserve_alignments for for RNNT but not CTC?\r\n\r\n+ Why do we need to use asr_model.decoding.decode_tokens_to_str for RNNT but asr_model._wer.decode_tokens_to_str for CTC models? is there a way to unify how these stuff work for both models?\r\n\r\n+ Do we have preserve_alignments in our configs now? Do we need to add it to all configs?", "1) Calling it automatically is wrong - user might not want to change config or might not want to silently update config. That's why there needs to be explicit call. \r\n\r\n2) CTC is cheap to decode, just argmax. RNNT is Autoregressive and expensive to compute alignments and preserve them (each alignment gets pushed onto CPU). So it needs a special flag to prevent degradation of performance in general case when used doesn't care about alignments.\r\n\r\n3) This is unified already per model type. CTC doesn't need an additional decoding class. RNNT does. The two models are too different to share more attributes. \r\n\r\n4) we don't need to add preserve_alignments to config, it's optional. Eventually we will have a tutorial for RNNT models which will demonstrate this. ", "1- One solution can be that we store the preserve_alignments somewhere in the transcribe(), change it true, then restore it.\r\n3-How about adding an extra method decode_tokens_to_str to the model and it calls the main corresponding decode_tokens_to_str based on the model. In this case, we will have a unfied API for all models. It is nice to cover differences from user's perspective as much as we can." ]

[]

[]

[ "Yes you can either add it to config and call the script or while calling the script you may give test_ds options as command line arguments prepended `+`. For reference look at hydra config https://hydra.cc/docs/tutorials/basic/your_first_app/config_groups ", "Thanks for your answer, Nithin! I did an experiment with speaker verification without setting up test_ds in the model training stage and set up `test_ds` the way it was done in the tutorial\r\n```python \r\ntest_config = OmegaConf.create(dict(\r\n manifest_filepath='/embeddings_manifest.json',\r\n sample_rate=16000,\r\n labels=None,\r\n batch_size=1,\r\n shuffle=False,\r\n time_length=8,\r\n embedding_dir='./'\r\n))\r\nverification_model.setup_test_data(test_config)\r\n```\r\nShould this be one way to do it? Thanks again for your help!", "yes, pass this to setup_test_data with model instantiated and extract embeddings. ", "Many thanks again for your help! " ]

[]

[]

[ "Can you please sign your committed commits? See DCO's bot output how to sign already committed commits https://github.com/NVIDIA/NeMo/pull/1767/checks?check_run_id=1925053254 ", "also, please re-base on the latest main, because currently CI does not pass due to a bug (which was fixed) in older version of main ", "Can you allow us write access to this PR? We can't merge unless your branch is up to date with main and we update main many times a day" ]

[]

[]

[ "This pull request **introduces 1 alert** and **fixes 2** when merging d285219e23c779109783e6db98f856ccf20730c6 into d7ad10435e97024c7c9fdbdd4228e0011dc9e432 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-b278125132b6fe7b7e4bf372a40fa8d12c683374)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import", "This pull request **introduces 1 alert** and **fixes 2** when merging 331c9d4e2b693d48c8ec024576ee34aab65672a4 into eff003635f6d7c4fbb1be4f46b747db63b58cd34 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-d3c05e91e5cd8e8107cc690ddd36c5c7a8aac545)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for Unused import" ]

[]

[]

[]

[]

[]

[ "This pull request **introduces 2 alerts** when merging 900c52a2fc5e01c04dddb29d51b3894c504e837b into 45aaaf4aad712d53b58e6b14db6343c4e9940f25 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-3758cb3dc44cd572ab179a17194a1cafa9166522)\n\n**new alerts:**\n\n* 2 for Unused import", "This pull request **introduces 2 alerts** when merging ed9d82b7a6032b7304e436434c0c6e78eff6e983 into 2b0711778f3463188ea87bdfd1b0b4ecc48ec11a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-f13ab895beee0599ab6d003ce727ebb54e1dc13c)\n\n**new alerts:**\n\n* 2 for Unused import", "This pull request **introduces 6 alerts** and **fixes 3** when merging fd12af9df4dacabd9125780096e4894df72d3fd3 into 2b0711778f3463188ea87bdfd1b0b4ecc48ec11a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-a88f3b11a7402e4cda247156a26cc184b2d2cd9f)\n\n**new alerts:**\n\n* 6 for Unused import\n\n**fixed alerts:**\n\n* 3 for Unused import", "This pull request **fixes 5 alerts** when merging c0c4b099004111aad1369cb2d8dc0279262f223a into 2b0711778f3463188ea87bdfd1b0b4ecc48ec11a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-53f33ff6e93e839ad537f3d533f1298f69698046)\n\n**fixed alerts:**\n\n* 5 for Unused import", "This pull request **fixes 5 alerts** when merging 6d082c04ce7cf0265b4e3d6f50f5581b1967e8b0 into 2b0711778f3463188ea87bdfd1b0b4ecc48ec11a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-5bc4526daea285700462f767ea9bb14afb250381)\n\n**fixed alerts:**\n\n* 5 for Unused import" ]

[]

[]

[ "I am unable to generate vectors of audio files using trained wav2vec model checkpoints. What is the specific method to call to extract features of any input audio files. @titu1994 @okuchaiev ", "@SeanNaren and @VahidooX ", "I have created a class named ExtractWav2VecEmbeddingsModel using BaseClass Wav2VecEncoderModel.\r\n\r\n`from pytorch_lightning import Trainer\r\n\r\nfrom typing import Dict, Optional, Union\r\n\r\nclass ExtractWav2VecEmbeddingsModel(Wav2VecEncoderModel):\r\n \r\n \"\"\"\r\n This Model class facilitates extraction of vector from a pretrained model.\r\n Respective embedding file is saved in self.vector dir passed through cfg\r\n \"\"\"\r\n \r\n def __init__(self, cfg: DictConfig, trainer: Trainer = None):\r\n super().__init__(cfg=cfg, trainer=trainer)\r\n\r\n def setup_test_data(self, test_data_config: Optional[Union[DictConfig, Dict]]):\r\n if 'shuffle' not in test_data_config:\r\n test_data_config['shuffle'] = False\r\n # preserve config\r\n self._update_dataset_config(dataset_name='test', config=test_data_config)\r\n\r\n self._test_dl = self._setup_dataloader_from_config(config=test_data_config)\r\n self.test_manifest = test_data_config.get('manifest_filepath', None)\r\n\r\n def test_step(self, batch, batch_idx, dataloader_idx=0):\r\n loss, feature_loss, prob_ppl_loss, target = self._step(batch)\r\n self.log('test_loss', loss, prog_bar=True, on_epoch=True, sync_dist=True)\r\n return {'vec': target, 'idx': batch_idx}\r\n\r\n def _step(self, batch):\r\n audio_signal, audio_lengths, _, _ = batch\r\n\r\n padding_mask = self._create_padding_mask(audio_lengths)\r\n\r\n self._update_quantizer_temp()\r\n logits, targets, sampled_negatives, _, features_penalty, prob_ppl_loss, _ = self(\r\n source=audio_signal, padding_mask=padding_mask\r\n )\r\n loss, feature_loss, prob_ppl_loss = self.loss(\r\n logits=logits,\r\n targets=targets,\r\n negatives=sampled_negatives,\r\n prob_ppl_loss=prob_ppl_loss,\r\n feature_loss=features_penalty,\r\n )\r\n return loss, feature_loss, prob_ppl_loss, targets\r\n\r\n def test_epoch_end(self, outputs):\r\n print ( type(outputs), len(outputs), self.test_manifest)\r\n for emb in outputs:\r\n print (emb['vec'].shape, emb['idx'])\r\n \r\nAfter that i called pytorch lightning Trainer.\r\n \r\n`trainer = pl.Trainer(**params.trainer)\r\n\r\nextract_vec_model = ExtractWav2VecEmbeddingsModel(cfg=params.model, trainer=trainer)\r\n\r\nckpt_PATH = \"/content/drive/MyDrive/nemo_experiments/Wav2vec_Pretrain/2021-02-17_09-21-19/checkpoints/Wav2vec_Pretrain---val_loss=34.72-epoch=97.ckpt\"\r\n\r\nextract_vec_model.load_state_dict(torch.load(ckpt_PATH), strict=False)`\r\n\r\n`test_config = OmegaConf.create(dict(\r\n manifest_filepath = '/content/hi_commonvoice_test_manifest.json',\r\n sample_rate = 16000,\r\n labels = None,\r\n batch_size = 1,\r\n shuffle = False,\r\n embedding_dir = './'\r\n))\r\nextract_vec_model.setup_test_data(test_config)`\r\n\r\n At the calling trainer test to evaluate the model\r\n`trainer.test(extract_vec_model)`\r\n\r\nNow i am getting the tensor of dimension described below for 4 test files\r\n\r\ntorch.Size([1, 103, 768]) 0\r\ntorch.Size([1, 78, 768]) 1\r\ntorch.Size([1, 88, 768]) 2\r\ntorch.Size([1, 73, 768]) 3\r\n\r\nI guess 768 is the output dimension and 2nd pos of tensor shape is timestamp factor of audio file. But for same duration audio file i am getting different value.\r\n\r\nIs my approach correct ? ", "i calling trainer model to inferrence ,the modes' weights is diffrent ,why ?here is my aligner code,training in gpu,inferring in cpu\r\n```\r\n@hydra_runner(config_path='conf', config_name='aligner')\r\ndef infer(cfg):\r\n from nemo.collections.asr.modules.audio_preprocessing import AudioToMelSpectrogramPreprocessor\r\n from nemo.collections.tts.models import AlignerModel\r\n from nemo.collections.asr.data.audio_to_text import AudioToCharWithPriorDataset,AudioToCharWithDursF0Dataset\r\n import torch\r\n from nemo.collections.asr.data.vocabs import Phonemes\r\n pho = Phonemes()\r\n AudioToCharWithPriorDataset_EXP = AudioToCharWithPriorDataset(\r\n attn_prior_folder=\"/storage2t/storage2/NeMo/examples/tts/ali1\",\r\n manifest_filepath=\"/storage2t/storage2/NeMo/scripts/dataset_processing/ljspeech/ljspeech_same.json\",\r\n vocab={\"notation\":\"phonemes\", \"punct\":True, \"spaces\":True, \"stresses\":False, \"add_blank_at\":None, \"pad_with_space\":True},\r\n sample_rate=22050\r\n )\r\n \r\n device = torch.device('cpu')\r\n model.load_from_checkpoint(\"/storage2t/storage2/NeMo/nemo_experiments/Aligner/2021-06-04_11-17-45/checkpoints/Aligner--val_forward_sum_loss=1.18-epoch=2349.ckpt\",strict=True,map_location=device)\r\n model.freeze()\r\n for parameters in model.parameters():\r\n print(parameters)\r\n\r\n for name, parameters in model.named_parameters():\r\n print(name, parameters)\r\n print(\"==================================================\")\r\n model.eval()\r\n with torch.no_grad():\r\n for item in range(AudioToCharWithPriorDataset_EXP.__len__()):\r\n audio, audio_len, text, text_len, attn_prior,tag = AudioToCharWithPriorDataset_EXP[item]\r\n a2p = AudioToMelSpectrogramPreprocessor(features=80,sample_rate=22050,pad_to=1,normalize=None,n_window_stride=256,n_fft=1024,n_window_size=1024,window_size=None,window_stride=None)\r\n audio = audio.unsqueeze(0)\r\n audio_len = audio_len.unsqueeze(0)\r\n text = text.unsqueeze(0)\r\n text_len = text_len.unsqueeze(0)\r\n spec, spec_len = a2p.get_features(audio, audio_len)\r\n # print(audio)\r\n for name, parameters in model.named_parameters():\r\n print(name, parameters)\r\n ", "Closing. @sam1373 re-open if issue is still relevant." ]

[]

[]

[ "This pull request **fixes 3 alerts** when merging 0bc9d88cc28dbdb604348502888931556acde702 into 7d8b7d679a2a4c9d3a6b2d8def8888e53c804ded - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-1f5d9f48dd6cb74ec4d5390dc1d8290bf990ed9d)\n\n**fixed alerts:**\n\n* 2 for Unused local variable\n* 1 for Unused import", "This pull request **fixes 2 alerts** when merging 4715c44a944ff344f6fb967393df078aea321216 into c5cd85f4202a94a44aba3103628fc39d73c48379 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-c0f6315b050920f0072cc5321f3a29a7b3a0b65d)\n\n**fixed alerts:**\n\n* 2 for Unused local variable", "Can we still make changes to b4 branch? @tkornuta-nvidia @okuchaiev ?", "> \r\n> \r\n> Can we still make changes to b4 branch? @tkornuta-nvidia @okuchaiev ?\r\n\r\nWhich branch should we use for bug fixes needed for TLT?", "This pull request **fixes 2 alerts** when merging af60927630e6ae145748eb9ef73ba7e82172a714 into c5cd85f4202a94a44aba3103628fc39d73c48379 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-ba6fbbbe145ae3615601b252ce75d1d24357b625)\n\n**fixed alerts:**\n\n* 2 for Unused local variable" ]

[]

[]

[]

[]

[]

[ "@ericharper Can you pull this branch and verify NMT models have no issue running their WebDataset training on some toy / small dataset?\r\n\r\n@fayejf Same for ASR Classification models\r\n\r\nI've verified that ASR CTC/RNNT work 1:1 as before." ]

[]

[]

[ "This is a great question ! Surprisingly, the QuartzNet config does not show two very nice parameters of experiment manager that are used precisely for resuming training over multiple runs - https://github.com/NVIDIA/NeMo/blob/main/examples/asr/conf/citrinet/citrinet_384.yaml#L381-L382\r\n\r\nYou can add them to the quartznet config manually, or use hydra to add them with the `+exp_manager.resume_if_exists=true` and `+exp_manager.resume_ignore_no_checkpoint=true` flags. Note, when doing multiple such runs, give the experiment directory a unique name with `exp_manager.name=<some unique experiment name>`. \r\n\r\nAfter the first run, simply rerun the training script with the same config / overrides and same experiment manager flags - and it will load up the model + its checkpoint + the optimizer state and scheduler state automatically and continue training.", "A little explanation of what those two flags do\r\n\r\n`resume_if_exists` = Resume the experiment if an earlier checkpoint exists. [Docstring](https://github.com/NVIDIA/NeMo/blob/main/nemo/utils/exp_manager.py#L122-L124)\r\n\r\n`resume_ignore_no_checkpoint` = If you use the above flag alone for the first run, (when no checkpoint exists yet to resume from), experiment manager will throw an error (since it cant resume from a checkpoint which doesnt exist). This second flag tells exp_manager to ignore that the checkpoint doesnt exist for the first run - so the first run will \"train from scratch\". When it finishes, subsequent calls will see that a checkpoint exists so experiment manager will use it. [Docstring](https://github.com/NVIDIA/NeMo/blob/main/nemo/utils/exp_manager.py#L128-L130)", "Thanks for your detailed response! \r\nI tried a training following the steps I listed before in my question and it seems not to improve: learning rate stays at 0, probably because Cosine Annealing took it to 0 at the end of the first training, but someway, if the old trainer is re-used increasing its `max_epochs` flag, the lr value is not increased anymore. So I don't think mine was the correct way. \r\n\r\n\r\nNow I'm trying adding the parameters you suggested to the quartznet config. If I understood correctly, after the first run I can resume from checkpoint with `quartznet = nemo_asr.models.EncDecCTCModel.restore_from(\"./out/my_model.nemo\")`, then re-assign experiment manager flags and overrides (i.e. `quartznet.setup_training_data(train_data_config=params['model']['train_ds'])` and so on, as in point 4 of my post) and, in the end, run `trainer.fit(quartznet)` without necessarily having to restore my old trainer as I did until now (point 3 of my post) but being able to create a new trainer such as `trainer = pl.Trainer(gpus=1, max_epochs=20, precision=16, amp_level='O1')`. Is it correct?", "Another question is, when I setup my trainer, should I give it a `max_epochs` flag equal to the total number of epochs I intend to play (i.e. 256) or a smaller number which is the number of epochs to be played during the current session (around 20)? Because I think this would influence the learning rate decay", "Please see the examples/asr/speech_to_text.py script in NeMo. You can replace the constructor EncDecCTCModel(...) With model = EncDecCTCModel.restore_from(<nemo file>).\r\n\r\nYou will also need to use model.set_trainer(trainer), model.setup_training_data(cfg.model.train_ds), model.setup_validation_data(cfg.model.validation_ds) since restored models don't automatically setup their datasets and trainer. \r\n\r\nAfter that, simply calling the script multiple times with the suggested experimental manager settings should be sufficient. \r\n\r\nNote - after the first run, exp manager will create .ckpt Pytorch Lightning checkpoints - and will restore those for the next run. .nemo files are \"final packaged\" models - not intermediate training checkpoints. They don't contain the optimizer state etc, just the model parameters and the artifacts necessary to build the model. \r\n\r\nOnce your training finishes, exp manager will create a final . nemo file for you as well to show that training is over. ", "Max epochs is the number of epochs it will train - and will affect the scheduler + learning rate. ", "Thanks again for your answers. I followed the steps you suggested and now it seems to correctly resume from last checkpoint. The only thing I'm missing regards the number of epochs to be passed to the trainer. In my last experiment I set it to 256 (which is my target epochs value) and I manually stopped the trainer around the 20th epoch, then resumed after. Would it be different if I instead set the `max_epochs` value to 20 then increase it whenever I want to resume the training? I bet on setting it to 256 since you told me that `max_epochs` affects the scheduler + learning rate, but I'm not sure if manually stopping the training is the correct way or not.", "Once you begin a run, the history of what parameters you used at the beginning is preserved inside the .ckpt intermediate checkpoints. So you cannot increase number of epochs as such after the first run - it will complete it according to whatever was the original config only. \r\n\r\nWhen restarting training, experiement manager will use whichever was the last checkpoint saved. If you stop training between two save steps, the subsrqu run will have to again train on the steps that were not saved prior to being stopped. ", "Many thanks, you have clarified all my doubts. I think we can close the issue.", "> You will also need to use model.set_trainer(trainer), model.setup_training_data(cfg.model.train_ds), model.setup_validation_data(cfg.model.validation_ds) since restored models don't automatically setup their datasets and trainer.\r\n\r\n@titu1994 Hello!\r\nWhen learning from *-lost.ckpt, where does Nemo get the rest of the training parameters from, am I interested in optimal.sched.min_lr?\r\nI trained the model with min_lr = 1.0e-05, then continued training from .ckpt with min_lr = 1.0e-09, but lr did not fall below 1.0e-05. \r\n\r\n```\r\nasr_model = nemo_asr.models.EncDecCTCModelBPE.restore_from(\r\n cfg.model.init_from_nemo_model.citrinet.path, map_location=\"cpu\"\r\n ) \r\n...\r\n...\r\nasr_model.cfg.optim = cfg.model.optim # in config set new value for lr \r\ntrainer = pl.Trainer(**cfg.trainer) \r\nasr_model.set_trainer(trainer) \r\ntrainer.fit(asr_model) \r\n```" ]

[]

[]

[ "This pull request **introduces 6 alerts** when merging ace2826c3c6afa62827f773a1f427141640afdab into 562087498d68754aca6dcde8f4ac8813300e7b24 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-31be61d5dc5e90974461a3ee4fb9ee9ed9f5fd98)\n\n**new alerts:**\n\n* 6 for Unused import", "This pull request **introduces 3 alerts** when merging 76a9e6c008aa0b85cb04152a1f02e6c14d631768 into 562087498d68754aca6dcde8f4ac8813300e7b24 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-8f90296aecf16a1f1e165f812eb8da3f18077797)\n\n**new alerts:**\n\n* 3 for Unused import", "> @MaximumEntropy and @ericharper should we refactor `examples/nlp/machine_translation/create_tarred_parallel_dataset.py` to rely on logic in `nemo.collections.nlp.data.machine_translation.preproc_mt_data.MTDataPreproc` ?\r\n\r\nMaybe we can just use the MTDataPreproc since we have the do_training=False flag?", "> > @MaximumEntropy and @ericharper should we refactor `examples/nlp/machine_translation/create_tarred_parallel_dataset.py` to rely on logic in `nemo.collections.nlp.data.machine_translation.preproc_mt_data.MTDataPreproc` ?\r\n> \r\n> Maybe we can just use the MTDataPreproc since we have the do_training=False flag?\r\n\r\nI think it is fine and useful to have these two scripts, my biggest concern is that the actual logic needs to be implemented in one place (MTDataPreproc class probably)", "This pull request **introduces 3 alerts** when merging aca99cf52649fd13d52d23960a1d61d79a16ee95 into 16b87d1f7d22126b8eb471ca0538f75ade9620cf - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-3f597f2c3735ec2ef59457a5c8d0e82cc329fdd9)\n\n**new alerts:**\n\n* 3 for Unused import" ]

[]

[]

[ "Perturbation augmentation in NeMo is applied if probability at random sample > prob of augmentation selected. So total num of samples (iterations) remains the same. I suggest you train for a higher number of epochs for effective usage of augmentation. " ]

[]

[]

[ "Wich kind of LM? N-gram? If so, read [this](https://github.com/NVIDIA/NeMo/blob/main/tutorials/asr/Offline_ASR.ipynb)\r\n\r\nIf you are talking about transformer based LM, I am struggling the same. I opened a discussion about this [here](https://github.com/NVIDIA/NeMo/discussions/1736)", "Any other reference to add transformer based LM for the citrinet model ?", "@kruthikakr Could you get meaningful results with adding language model to Citrinet? I could not achieve: https://github.com/NVIDIA/NeMo/issues/2142", "Example and instructions to setup language model for ASR beam search with KenLM and neural language models are here - https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/asr_language_modeling.html" ]

[]

[]

[]

[ { "id": 1485815557, "node_id": "MDU6TGFiZWwxNDg1ODE1NTU3", "url": "https://api.github.com/repos/NVIDIA/NeMo/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": "Something isn't working" } ]

[]

[ "Could you share your training config file for QuartzNet ? The defaults for the pytorch lightning training logs at every step, which is excessive, and can be a cause for such a significant drop in training speed. ", "I am also wondering about `num_workers` parameter for data layers", "[quartznet.yaml.txt](https://github.com/NVIDIA/NeMo/files/5980973/quartznet.yaml.txt)\r\n\r\nAnd I start the training with these parameters:\r\n\r\n hydra.run.dir=\".\" \\\r\n trainer.gpus=4 \\\r\n +trainer.max_steps=200000 \\\r\n trainer.log_every_n_steps=100 \\\r\n model.train_ds.batch_size=32 \\\r\n +trainer.precision=16 \\\r\n +trainer.amp_level=O1 \\\r\n trainer.accumulate_grad_batches=2 \\\r\n trainer.val_check_interval=2000\\\r\n +model.validation_ds.num_workers=1 \\\r\n +model.train_ds.num_workers=8 \\\r\n +model.train_ds.pin_memory=True \\\r\n model.optim.lr=0.01\r\n\r\nRegarding the ``num_workers``: I have good experience with having two workers per GPU.", "The config and overrides seem fine. I see you are using amp_level=O1, please note that since Pytorch 1.6, the recommended implementation of AMP is Pytorch Native AMP. I doubt thats the cause of any slowdown though.\r\n\r\n" ]

[]

[]

[ "No, you should be able to get much higher accuracies for 'megatron-bert-345m-uncased' on Squadv2 dataset.\r\nThis what I got after 2 epochs of training em=81.35%, f1=84.75%.\r\nI used batch size=12, precision=16, amp_level=01", "Actually, I think I know the reason for the low numbers. We had a recent bug in QA model's code which was just corrected. Please take the latest update from Nemo 'main' branch and it should train properly for you.", "@avikpal00 could you share your omegaconfig of your settings that is printed at the beginning of execution if the suggestions by @vladgets do not solve your problem?", "@avikpal00 closing due to inactivity " ]

[]

[]

[]

[]

[]

[]

[]

[]

[ "This pull request **introduces 1 alert** when merging 39fe1a548baa62507241a391b05d471ab05d5d76 into 719e21d402679e0edbdb7fdcc41c5bcd4af295d8 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-5b66e21ee1c4104cca572d02b00a48546f1e7c3b)\n\n**new alerts:**\n\n* 1 for Unused local variable" ]

[]

[]

[]

[]

[]

[]

[ { "id": 1485815557, "node_id": "MDU6TGFiZWwxNDg1ODE1NTU3", "url": "https://api.github.com/repos/NVIDIA/NeMo/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": "Something isn't working" } ]

[]

[ "What exactly do you mean by \"Method 1\" and \"Method 2\". Could you please paste code snippets? ", "\"Method 1. \"\r\n```\r\nprint(first_asr_model.transcribe(paths2audio_files=['pass path to .wav file']))\r\n```\r\n\"Method 2.\"\r\n```\r\n# Bigger batch-size = bigger throughput\r\nparams['model']['validation_ds']['batch_size'] = 16\r\n\r\n# Setup the test data loader and make sure the model is on GPU\r\nfirst_asr_model.setup_test_data(test_data_config=params['model']['validation_ds'])\r\nfirst_asr_model.cuda()\r\n\r\n# We will be computing Word Error Rate (WER) metric between our hypothesis and predictions.\r\n# WER is computed as numerator/denominator.\r\n# We'll gather all the test batches' numerators and denominators.\r\nwer_nums = []\r\nwer_denoms = []\r\n\r\n# Loop over all test batches.\r\n# Iterating over the model's `test_dataloader` will give us:\r\n# (audio_signal, audio_signal_length, transcript_tokens, transcript_length)\r\n# See the AudioToCharDataset for more details.\r\nfor test_batch in first_asr_model.test_dataloader():\r\n test_batch = [x.cuda() for x in test_batch]\r\n targets = test_batch[2]\r\n targets_lengths = test_batch[3] \r\n log_probs, encoded_len, greedy_predictions = first_asr_model(\r\n input_signal=test_batch[0], input_signal_length=test_batch[1]\r\n )\r\n # Notice the model has a helper object to compute WER\r\n first_asr_model._wer.update(greedy_predictions, targets, targets_lengths)\r\n _, wer_num, wer_denom = first_asr_model._wer.compute()\r\n wer_nums.append(wer_num.detach().cpu().numpy())\r\n wer_denoms.append(wer_denom.detach().cpu().numpy())\r\n\r\n# We need to sum all numerators and denominators first. Then divide.\r\nprint(f\"WER = {sum(wer_nums)/sum(wer_denoms)}\")\r\n```\r\nDid not changed and took it from tutorial.", "@okuchaiev any comments why is it so?", "Could this be because of Batch Normalization?", "In method two, I dont see any line where you call `model.eval()`. Assuming the notebook is being run multiple times, the model can predict random items if its still in train mode when being called manually - batch effects will be in play since batchnorm will try updating running statistics.\r\n\r\nTo go one step further wrap the first and second method with :\r\n```python\r\n\r\nasr_model.freeze()\r\nwith torch.no_grad():\r\n ...\r\n```\r\n\r\nThis is generally not necessary though, and simply calling model.eval() along with torch.no_grad() should do the trick.", "@topusOctopus Theres also a PR which should (hopefully) simplify transcription for all NeMo ASR models in the future - can you take my branch and give your eval a try? https://github.com/NVIDIA/NeMo/pull/1786", "@titu1994 I will try your suggestion today" ]

[]

[]

[]

[]

[]

[ "This pull request **introduces 3 alerts** when merging 23de8602c7e623882a7e62ac314fec5fddd3cc0a into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-3402b02b11a89a42f2072d0329605bc57ba587a6)\n\n**new alerts:**\n\n* 3 for Unused import", "This pull request **introduces 3 alerts** when merging 41d4e9c4764c782783a70c06c52220a89bfd0a43 into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-be6468e500d0c314d04164485e00c974e0328f26)\n\n**new alerts:**\n\n* 3 for Unused import", "This pull request **introduces 3 alerts** when merging 6d8e51d6a3796f5fb5199e985c94d8505c9dbbd9 into dacdb1ad4286731f50fa33c0a74b04089d373f44 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-2e6fae1825d5fe05559444575e75b34e4f416b6e)\n\n**new alerts:**\n\n* 3 for Unused import", "This pull request **introduces 3 alerts** when merging 2e54cf28e61fd41793c6dcd517887aa49e7a305b into 85224dbdc40af41be5f68c2b60004073c91d9a91 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-e09b8fef5acf5d84018cad8572998bed652fb1d6)\n\n**new alerts:**\n\n* 3 for Unused import", "This pull request **introduces 3 alerts** when merging 665f5fed8b1d424e61c2d87f66457ffce69dfed3 into fc4017a08db4cf2aaad2288dd158e9c287f7107a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-1c10b51593551fac77c26655b1f78183cdfd7b5f)\n\n**new alerts:**\n\n* 3 for Unused import" ]

[]

[]

[ "This pull request **introduces 1 alert** when merging a3c78c222796b9fcffcfebdabf589b9d6abeb2bd into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-65d55c914da1b1fc018109d5c222d0a94573de41)\n\n**new alerts:**\n\n* 1 for Unreachable code", "Agreed about docs - maybe in a notebook specifically for types ? Or maybe the api docs in RTD" ]

[]

[]

[]

[]

[]

[]

[]

[]

[ "This pull request **introduces 4 alerts** and **fixes 8** when merging dca86d9367a59e969c14dd3bac2321db6e9a7b16 into 25abffdd37efb3a9f5a6e236d910f045271ae08f - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-ec901cb9fdf53607236db93f64f526687b49cea5)\n\n**new alerts:**\n\n* 2 for Unused local variable\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 8 for Unused import", "This pull request **introduces 2 alerts** and **fixes 8** when merging 67d8abe00438996aabf3312be19582154179cdf5 into dacdb1ad4286731f50fa33c0a74b04089d373f44 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-9972ff061575ee90629df37c3370ed78841afaaa)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 8 for Unused import", "@borisfom is this still WIP?", "This pull request **introduces 2 alerts** and **fixes 8** when merging 7ff3e50ac34475c813c0e4ca54f61771905f0a7b into 16b87d1f7d22126b8eb471ca0538f75ade9620cf - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-31349133ef976f22408b5b7f7a72f3f40edba1b3)\n\n**new alerts:**\n\n* 2 for Unused import\n\n**fixed alerts:**\n\n* 8 for Unused import" ]

[]

[]

[ "@ekmb is this only docs? Should it go to r1.0.0rc1 then?" ]

[]

[]

[ "This pull request **introduces 7 alerts** and **fixes 1** when merging 39305db0bbb9b46913d7af27263078f7a3e0d051 into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-9d19b73f6a2e17f867cc19d0bc33bfa1fed8951a)\n\n**new alerts:**\n\n* 4 for Unused import\n* 3 for NotImplemented is not an Exception\n\n**fixed alerts:**\n\n* 1 for NotImplemented is not an Exception", "This pull request **introduces 7 alerts** and **fixes 1** when merging c8771cd94d853f0e95bb10837a8151c74aefb017 into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-21eb996b3547d2da4c9db33b8b98c3e4e47b40c3)\n\n**new alerts:**\n\n* 4 for Unused import\n* 3 for NotImplemented is not an Exception\n\n**fixed alerts:**\n\n* 1 for NotImplemented is not an Exception", "This pull request **introduces 7 alerts** and **fixes 1** when merging ee310bde853aa5bde392fa700bbe7c2d859a66c6 into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-af39994aab0a30451f880e37c982b9d02cfcb99e)\n\n**new alerts:**\n\n* 4 for Unused import\n* 3 for NotImplemented is not an Exception\n\n**fixed alerts:**\n\n* 1 for NotImplemented is not an Exception", "This pull request **introduces 7 alerts** and **fixes 1** when merging 415d89b4bd54c104728493471adf5d6a749d4774 into dacdb1ad4286731f50fa33c0a74b04089d373f44 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-787da5da4a13eb3e92830b36c7d672d8ea97f7ed)\n\n**new alerts:**\n\n* 4 for Unused import\n* 3 for NotImplemented is not an Exception\n\n**fixed alerts:**\n\n* 1 for NotImplemented is not an Exception", "This pull request **introduces 7 alerts** and **fixes 1** when merging e58ce898f08bd6a05791f2c06a25e0fbfe975608 into dacdb1ad4286731f50fa33c0a74b04089d373f44 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-e03037e6d8fdca6358617d98f91d19b1c291c1a9)\n\n**new alerts:**\n\n* 4 for Unused import\n* 3 for NotImplemented is not an Exception\n\n**fixed alerts:**\n\n* 1 for NotImplemented is not an Exception", "This pull request **introduces 7 alerts** and **fixes 1** when merging ea2059bdd67e734aacd14a7d8cbb4b81841ed80a into 85224dbdc40af41be5f68c2b60004073c91d9a91 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-d4a50594e7480874d6701c8ae0a48e05ba857d61)\n\n**new alerts:**\n\n* 4 for Unused import\n* 3 for NotImplemented is not an Exception\n\n**fixed alerts:**\n\n* 1 for NotImplemented is not an Exception", "This pull request **introduces 8 alerts** and **fixes 1** when merging 3af0fbf020445f9bc0f7dc26b1a2b44fc3fb8632 into 85224dbdc40af41be5f68c2b60004073c91d9a91 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-07995c215b3500c2750daa2c065604cacfbecbab)\n\n**new alerts:**\n\n* 5 for Unused import\n* 3 for NotImplemented is not an Exception\n\n**fixed alerts:**\n\n* 1 for NotImplemented is not an Exception", "This pull request **introduces 1 alert** and **fixes 2** when merging 007cdb97bf3e14feac42b696e120bbd4430b0099 into 85224dbdc40af41be5f68c2b60004073c91d9a91 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-e287075de9ec0672b66629e0f7ef9b34d7dacdd2)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This pull request **introduces 1 alert** and **fixes 2** when merging 7c2a531b871532e71bfe08e1273a5e78930c9b8e into 4518b0030485bdd0634831a957cd6f3dd5c52f60 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-0b459ac05528cebb1c9eb47566536d9743d75959)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This pull request **introduces 1 alert** and **fixes 2** when merging be18efa852e5f86ca9fe110db733213129ec50a3 into fc4017a08db4cf2aaad2288dd158e9c287f7107a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-e9a3285512df8ba91cb05fb5406557007d0fc3a1)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This pull request **introduces 1 alert** and **fixes 2** when merging a120d6f76a06a5e8b54dd8ad14a2412367c8552f into fc4017a08db4cf2aaad2288dd158e9c287f7107a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-110ce275b7fb2de6c634a9a7de0df585f9f1db06)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This PR is good, but @ericharper please add tests (in this or subsequent PRs) for this workflow with some arbitrary small random data - \r\n1) from scratch tokenize, tar, train flow\r\n2) restore model - in different env than original env\r\n3) setup train tokekizer, etc after restore c\r\n\r\nNMT side has woefully few tests for tarred dataset and model checkpoint testing in general ", "This pull request **introduces 1 alert** and **fixes 2** when merging 1e22f113606a7fc007511dd39551484584cac6e4 into fc4017a08db4cf2aaad2288dd158e9c287f7107a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-6d860e01bf512f9daaac61759de758320115eb91)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This pull request **introduces 1 alert** and **fixes 2** when merging 7a2095807c916c54c224f3320521cb209c53df75 into fc4017a08db4cf2aaad2288dd158e9c287f7107a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-f820a3d19835b685676282308a050280697d5c6b)\n\n**new alerts:**\n\n* 1 for Unused import\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This pull request **fixes 2 alerts** when merging 972ffeb15e5e297b955051e6f4e0d1bed950780c into fc4017a08db4cf2aaad2288dd158e9c287f7107a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-8df142ef1715cd2747ce6ce31a690804e0cbea40)\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This pull request **fixes 2 alerts** when merging 10a45a12ea8fdd28e475f251c61a44f0e265c7a2 into fc4017a08db4cf2aaad2288dd158e9c287f7107a - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-360464a32a3857a4e406632ca47503627f5d168d)\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception", "This pull request **fixes 2 alerts** when merging bee4236db27a78cf7ace6b38e3e6fc57e346338d into e6439786239044e3c4b803b296b8013feb142772 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-5bf19714013d312906024d42216a831fbf06f665)\n\n**fixed alerts:**\n\n* 2 for NotImplemented is not an Exception" ]

[]

[]

[ "This pull request **introduces 4 alerts** when merging f5a2dbfb6f1f1bd595fd00682be5e54e902ef96a into 25abffdd37efb3a9f5a6e236d910f045271ae08f - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-c72ad221ab5286086e5b409c23e1e96df83bfe0c)\n\n**new alerts:**\n\n* 3 for Unused import\n* 1 for Module is imported more than once", "This pull request **introduces 4 alerts** when merging 6320fbe535405c66cae2e2bf50570be0b3b546ad into 25abffdd37efb3a9f5a6e236d910f045271ae08f - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-b530df5982cf2d083f00241b63c2a314e63a6cbb)\n\n**new alerts:**\n\n* 3 for Unused import\n* 1 for Module is imported more than once", "This pull request **introduces 4 alerts** when merging d7f55f28566ae6e790642fa91ebe03948aa31c38 into 25abffdd37efb3a9f5a6e236d910f045271ae08f - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-be3e449c660487cf53a6d2d6c4eee8e653aa5bb5)\n\n**new alerts:**\n\n* 4 for Unused import", "This pull request **introduces 1 alert** when merging 375acd96a2b1c7ca9afdb4b4b9e8ab17bb84221c into 25abffdd37efb3a9f5a6e236d910f045271ae08f - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-af93b8d91ce11fda7cca63699183e7ff74c19f92)\n\n**new alerts:**\n\n* 1 for Unused import", "This pull request **introduces 1 alert** when merging a9cae4f63640088e0490f45b20a4c6565d26b369 into 25abffdd37efb3a9f5a6e236d910f045271ae08f - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-cee1a14c2d4a960d316751115b4a204418f1cc29)\n\n**new alerts:**\n\n* 1 for Unused import", "This pull request **introduces 1 alert** when merging b1cfc5a28c9f161ad9537f479dc0f35b2f07a391 into 87e0466f7e3d4e8904b28b447b5a059fbdffa28b - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-a7ac88eb4ff732e1d6150f3f61a40d21f08390cf)\n\n**new alerts:**\n\n* 1 for Unused import", "This pull request **introduces 2 alerts** when merging 3a52f04ed2fb76537c8a2d5037904b2a48b58891 into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-50a322c0bf93e9c19aadabcf9f84649993778599)\n\n**new alerts:**\n\n* 1 for Unused import\n* 1 for Syntax error", "This pull request **introduces 1 alert** when merging 1a38d7298d1f744af24ffd607ff05f37ca919a15 into 3c288d8a7caf667c95444c39434e3ebc5f53d911 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-a1da6057e9885912d064606cc3b59b135286c90a)\n\n**new alerts:**\n\n* 1 for Variable defined multiple times" ]

[]

[]

[]

[]

[]

[ "You should use:\r\n```\r\npip install nemo-toolkit[all]==1.0.0b3\r\n```\r\nOr in your requirements.txt file\r\n```\r\nnemo-toolkit[all]==1.0.0b3\r\n```", "Thanks a lot for the quick reply @okuchaiev . I have seen that nemo 1.0.0b4 was recently released and I am looking forward to incorporating the latest changes in that version. Do you plan to upload this latest version to PyPI as well?\r\n\r\nCheers,\r\nEshref" ]

[]

[]

[ "Doing this for every single step in beam search process would be too computationally expensive.\r\nInstead you can re-score candidate beams from beam search like we did with transformer XL in this paper:\r\nhttps://arxiv.org/pdf/1904.03288.pdf " ]

[]

[]

[ "We should close this PR - its sole purpose was to enable ONNX export for Megatron (in Nemo b3 environment) - which we have found is broken anyway in b4 (due to Pytorch version).\r\nSo it does not solve the issue for our standard containers and introduces risks.", "please note I still need EFF." ]

[]

[]

[]

[]

[]

[ "This pull request **introduces 5 alerts** when merging 661c91f16cd7c94e0d9c6e358c457530c80338d0 into 0e6159141b45336607dc137389bc0d6839a6606c - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-a1cbb8519895cc2212edbefc60557dccc5da00ee)\n\n**new alerts:**\n\n* 5 for Unused import", "This pull request **introduces 1 alert** when merging abdd3bf12198089666cf0a842f3a5612aa93bb6e into 6c34fe9d44d0c398dd9a87404f9fa45616a13647 - [view on LGTM.com](https://lgtm.com/projects/g/NVIDIA/NeMo/rev/pr-6724a601044c3fe1d5d9ce0bd8fcea3b2281f90e)\n\n**new alerts:**\n\n* 1 for Unused import" ]

[]

[]

[]