Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / sanchit-gandhi/seq2seq-speech issues and pull requests

#98 - May we know the version of jax/flax/transformers?

Issue - State: open - Opened by snoop2head about 1 year ago

#97 - do not have file requirements.txt

Issue - State: open - Opened by pphuc25 about 1 year ago

#96 - error while finetuning wav2vec2 with bart.

Issue - State: open - Opened by arvindmn01 about 1 year ago - 1 comment

#95 - whisper

Issue - State: open - Opened by surajpachouri over 1 year ago

#93 - Analyse data script

Pull Request - State: closed - Opened by patrickvonplaten about 2 years ago

#92 - [E22] Add missing junk token (<silence>)

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#91 - [Whisper] Training

Pull Request - State: closed - Opened by patrickvonplaten about 2 years ago

#90 - [SWBD] Torchaudio resampler (12x faster than Librosa)

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#89 - [Earnings22] Black box systems

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#88 - [SWBD+Fisher] Black-box pre-processing

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#87 - More scripts for ngram

Pull Request - State: closed - Opened by patrickvonplaten about 2 years ago

#86 - [black-box] Add scripts for ERSR black-box runs and code changes

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#85 - [Error Correction] Add SPGISpeech and tidy up comments

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#83 - [Ngram] compute ngrams & evaluate

Pull Request - State: closed - Opened by patrickvonplaten about 2 years ago - 3 comments

#82 - [prepare_dataset] Improve error correction vs normalisation

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#81 - [CTC] SWBD Tokenizer

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#80 - Improve error correction vs. normalization

Pull Request - State: closed - Opened by patrickvonplaten about 2 years ago - 1 comment

#79 - [n-gram] Add dummy script

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#78 - [n-gram] Update pre-processing to align with CTC + S2S

Pull Request - State: closed - Opened by sanchit-gandhi about 2 years ago

#77 - Eval samples filtering

Issue - State: closed - Opened by mutiann over 2 years ago - 1 comment

#76 - [Seq2Seq] Fix pad_shard_unpad kwarg bug

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#75 - Random `TypeError` with NumPy array with all values -100

Issue - State: closed - Opened by versae over 2 years ago - 3 comments

#74 - Recovering from crashed run

Issue - State: open - Opened by versae over 2 years ago - 3 comments

#73 - [Seq2Seq] Fix pad_shard_unpad kwarg bug

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#72 - Add RNN-T Model & Training Script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago - 1 comment

#71 - [Seq2Seq] Template scripts

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#70 - [Imports] Fix Flax model paths (for TPU) and omit RNN-T (for GPU)

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#69 - [Pre-processing] Make ds specific (for ERSR)

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#68 - [Clean] Tidy up imports

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#67 - [CTC] Add regularisation

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#66 - [Dataloader] Handle incomplete final batch

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#65 - [`_do_init`] Update modeling files to handle `_do_init` kwarg

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#64 - [Earnings] Final preprocessing

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#63 - [CTC] Fix missing backslash in dummy CTC run script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#62 - [Run] Fix bugs in dummy run scripts

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#61 - [Datasets] fix hashing/caching bug

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#60 - [CTC] Filter by min target length

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#59 - [CTC] Fix eval/test split bug

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#58 - [GigaSpeech] Add preprocessing

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#57 - [Earnings 22] Add preprocessing

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#56 - [CTC] Update dummy run script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#55 - Update dummy run scripts

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#54 - [Eval] Fix eval steps bug

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#53 - [SWB] Finalise preprocessing

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#52 - [AMI] Add Seq2Seq + CTC training scripts

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#51 - [Seq2Seq Train] Beam at each eval steps, greedy at end of epoch

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#50 - [wandb] Fix push to hub, log run id

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#49 - [Preprocess] Gigaspeech and SWB

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#47 - [CTC Tokenizer] Gigaspeech, SWB and AMI

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#46 - [Eval] Update logger to log wer or cer (depending on data args)

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#45 - [Eval] Evaluate with wer (default) or cer metrics

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#44 - Remove spaced apostrophes and <unk> token

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#43 - [Eval] Change default split name + set steps to max_steps

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#42 - make style

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#41 - [Train] Fix bug in reference to epochs

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#40 - [CTC] Add beam search decoder

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#39 - [Train] Remove speech disfluencies and clean data

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#38 - [TEDLIUM] Remove speech disfluencies

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#37 - [CTC] Remove unnecessary arg `ignore_mismatched_size`

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#36 - [CTC] Fix bug in model-tokenizer vocab size mismatch

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#34 - Train for N steps, eval every M and save every K

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#33 - [CTC] Remove punctuation from train split only

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#32 - [CTC] Facilitate removal of punctuation from tokenizer

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#31 - Fix bug in data arg

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#30 - [CTC] Enable removal of punctuation

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#29 - Update CTC notebook

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#28 - Negative Losses in CTC Training

Issue - State: closed - Opened by sanchit-gandhi over 2 years ago - 2 comments

#27 - Check Negative CTC Loss

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#26 - Update run_ctc_dummy.sh

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#25 - Make CTC tokenizer `do_lower_case` attribute water-tight

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#24 - Fix `do_lower_case` bug in CTC train script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#23 - CTC tokenizer returns <unk> tokens with `do_lower_case=True`

Issue - State: closed - Opened by sanchit-gandhi over 2 years ago - 1 comment

#22 - Add `--do_predict` to CTC script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#21 - Fix`train_metrics` bug in CTC script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#20 - Pad target to max length if int is not specified

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#19 - Run train/eval/test in one-script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#18 - File to create CTC Tokenizer

Pull Request - State: closed - Opened by patrickvonplaten over 2 years ago

#17 - Wav2Vec2 CTC in Flax/JAX

Pull Request - State: closed - Opened by patrickvonplaten over 2 years ago

#16 - Training script clean-up

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#15 - Beam search with scan

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#14 - Fuse matmul operations

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#13 - Gradient checkpointing and scan

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#12 - Correct Mixed Precision Training

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#11 - [train] Optimise gradient accumulation loop

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#10 - [feat] Save and use feature encoder outputs during training

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#9 - [FlaxWav2Vec2Model] Fix bug in attention mask

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#8 - [FlaxSpeechEncoderDecoder] Fix input shape bug in weights init

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#7 - (feat) Clip Gradients by Global Norm

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#6 - Update README.md

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#5 - Convert Flax training script to a standalone training script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#4 - Convert Flax SpeechEncoderDecoderModel to a standalone model script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#3 - Convert Flax Wav2Vec2 to a standalone model script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#2 - Convert Flax Bart to a standalone model script

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago

#1 - Add modelling files from Transformers 🤗

Pull Request - State: closed - Opened by sanchit-gandhi over 2 years ago