44 Commits (6f90a33f1f02f917035ba860aa22201424c3dc18)

Author	SHA1	Message	Date
Xinghai Sun	86811af7a3	Reset default value of batch_size, nun_proc_data and fix an invalid url for DS2.	7 years ago
Xinghai Sun	2e5e9b8c11	Turn on rnn_use_batch of Paddle for accelartion. ... Improve xmap_reader_mp by adding a flush thread.	7 years ago
Xinghai Sun	be1fbc68a1	Set process daemon property and reset default value of num_proc_data arguments.	7 years ago
Yibing Liu	3f608b3b81	enable the log of gradient clipping in training	7 years ago
Xinghai Sun	7f45752a13	Add profile.sh script for multi-gpu profiling.	7 years ago
Xinghai Sun	ac56a2f249	Update READMD.md and other details by following reviewers comments.	7 years ago
Xinghai Sun	ae7ef7929a	Rename some folders and update examples.	7 years ago
Xinghai Sun	0bbb9c3ee2	Re-organize folder structure and hierarchy for DS2.	7 years ago
Xinghai Sun	e8f7a8fde1	Update argument naming following Yibing's reviews.	7 years ago
Xinghai Sun	9571b6fc0e	Add back utils.py.	7 years ago
Xinghai Sun	8b64ef29c8	Re-style the config codes for tools in DS2.	7 years ago
Xinghai Sun	792129166a	Sort the config lines to make it look better.	7 years ago
Xinghai Sun	805846ce67	Reduce the config parsing codes for DS2 and make it looks cleaner.	7 years ago
Xinghai Sun	2aa4af1c29	Revert back to support input-hidden weights sharing between bi-directional RNNs. ... 1. Add options to enable and disable RNN weights sharing. 2. Set rnn_layer_size to 2048 by default. 3. Revert back the striding steps of 1st conv layer from 2 to 3. 4. Revert back to BRelu. Above follows DS2 papers.	7 years ago
Xinghai Sun	177af05953	Fix a bug in use_gru argument parsing.	7 years ago
Xinghai Sun	8f89a9bdd4	Print log to pfs for DS cloud training and set use_gru to False by default.	7 years ago
Xinghai Sun	5a63275845	Add GRU support.	7 years ago
Xinghai Sun	75719fea22	Fix an incorrect usage of is_local argument.	7 years ago
Xinghai Sun	0e79ee37a4	Bug fix and refine cloud training for DS2. ... Summary: 1. Add missing is_local argument (when set False, use pserver). 2. Add exception thrown if cp failed. 3. Add cloud mkdir if not cloud path for uploading does not exist. 4. Fix a bug using common path ./local_manifest for all nodes. (convert to /local_manifest) 5. Refine coding style.	7 years ago
Xinghai Sun	1b787411d3	Update noise and impulse augmentor according to code review.	7 years ago
Xinghai Sun	99e819e8ea	Add ImpulseResponseAugmentor and augmentation.config file.	7 years ago
Xinghai Sun	6df0f9bc44	Reset default multi-thread/process number to half of cpu count() for speedup.	7 years ago
Xinghai Sun	526e18b119	Add function docs for layer.py and model.py and update other details.	8 years ago
Xinghai Sun	8122dd9c29	Simplify train.py, evaluate.py, infer.py and tune.py by adding DeepSpeech2Model class.	8 years ago
Yibing Liu	ee5abbe37d	add mfcc feature for DS2	8 years ago
Xinghai Sun	13f708739b	Improve audio featurizer and add shift augmentor. ... 1. Improve audio featurizer. 2. Add shift augmentor. 3. Update default argument to be the current best seggestion. 4. Add checkpoints with pass id.	8 years ago
Xinghai Sun	d104eccf67	Update the default num_threads for DS2 data generator.	8 years ago
Xinghai Sun	1d8cc4a5a9	Add multi-threading support for DS2 data generator.	8 years ago
Xinghai Sun	ed5f04afb8	Add shuffle type of instance_shuffle and batch_shuffle_clipped.	8 years ago
Xinghai Sun	04a225ae4f	Enable min_batch_num in train.py and update train info print.	8 years ago
Xinghai Sun	b07ee84a1d	Add function, class and module docs for data parts in DS2.	8 years ago
Xinghai Sun	cd3617aeb4	Refactor whole data preprocessor for DS2 (re-design classes, re-organize dir, add augmentaion interfaces etc.). ... 1. Refactor data preprocessor with new added class AudioSegment, SpeechSegment, TextFeaturizer, AudioFeaturizer, SpeechFeaturizer. 2. Add data augmentation interfaces and class AugmentorBase, AugmentationPipeline, VolumnPerturbAugmentor etc.. 3. Seperate normalizer's mean and std computing from training, by adding FeatureNormalizer and a seperate tool compute_mean_std.py. 4. Re-organize directory.	8 years ago
dangqingqing	9c27b1d14e	add more comments and update train.py	8 years ago
dangqingqing	bf73540067	Merge branch 'develop' of https://github.com/PaddlePaddle/models into ds2	8 years ago
yangyaming	f49eab5fec	Change assert to exception raising.	8 years ago
dangqingqing	7c85e0fdb5	Support variable input batch and sortagrad.	8 years ago
yangyaming	d2e467385d	Add loading model function for train.py.	8 years ago
Xinghai Sun	2a83486500	Refactor decoder interfaces and add ./data directory.	8 years ago
Xinghai Sun	8313895e85	1. Fix incorrect decoder result printing. ... 2. Fix incorrect batch-norm usage in RNN. 3. Fix overlapping train/dev/test manfests. 4. Update README.md and requirements.txt. 5. Expose more arguments to users in argparser. 6. Update all other details.	8 years ago
Xinghai Sun	e6a349992b	Refactor data utils into a class and add feature normalization.	8 years ago
Xinghai Sun	9c3cd3c704	Update some parameters and comments.	8 years ago
Xinghai Sun	0babc5c4d7	Add function docs.	8 years ago
Xinghai Sun	70a343a499	Add infererence and add SortaGrad for only first pass.	8 years ago
Xinghai Sun	3fc94427db	Add librispeech dataset, audio data provider and simplfied DeepSpeech2 model configuration. ... Bug exists when run training.	8 years ago