diff --git a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/docker.md b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/docker.md index b9360600626dae9b96e155f425bbc03ffbb6b725..31033f2d6b8337a162ceb9fa9d31de44e2d2f8ff 100644 --- a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/docker.md +++ b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/docker.md @@ -8,7 +8,7 @@ $ cd docker $ ./run.sh --docker-gpu 0 --docker-egs chime4/asr1 --docker-folders /export/corpora4/CHiME4/CHiME3 --dlayers 1 --ngpu 1 ``` Optionally, you can set the CUDA version with the arguments `--docker-cuda` respectively (default version set at CUDA=9.1). The docker container can be built based on the CUDA installed in your computer if you empty this arguments. -By default, all GPU-based images are built with NCCL v2 and CUDNN v7. +By default, all GPU-based images are built with NCCL v2 and cuDNN v7. The arguments required for the docker configuration have a prefix "--docker" (e.g., `--docker-gpu`, `--docker-egs`, `--docker-folders`). `run.sh` accept all normal ESPnet arguments, which must be followed by these docker arguments. All docker containers are executed using the same user as your login account. If you want to run the docker in root access, add the flag `--is-root` to command line. In addition, you can pass any environment variable using `--docker-env` (e.g., `--docker-env "foo=path"`) @@ -43,7 +43,7 @@ $ cd docker $ ./run.sh --docker-gpu 0 --docker-egs chime4/asr1 --docker-folders /export/corpus/CHiME4,/export/corpus/LDC/LDC93S6B,/export/corpus/LDC/LDC94S13B --docker-env "CHIME4_CORPUS=/export/corpus/CHiME4/CHiME3,WSJ0_CORPUS=/export/corpus/LDC/LDC93S6B,WSJ1_CORPUS=/export/corpus/LDC/LDC94S13B" --ngpu 1 ``` -Remember that for some recipes, you first need to download the Corpus before running the experiments, such as CHiME, WSJ, and LDC corporas. You will need to set the directories where these were downloaded and replace them in the recipe (e.g.: `CHIME4_CORPUS=//CHiME4/CHiME3`) +Remember that for some recipes, you first need to download the Corpus before running the experiments, such as CHiME, WSJ, and LDC corpora. You will need to set the directories where these were downloaded and replace them in the recipe (e.g.: `CHIME4_CORPUS=//CHiME4/CHiME3`) ### Using CPU-based container @@ -68,7 +68,7 @@ However, in some cases, "local" builds are preferable, that are built based on t The script `build.sh` supports making local builds for this purpose. During the docker build process, the local espnet source code is imported through a git archive based on git HEAD (the previous commit), and copied over within a file. -For example, a local build that the base image from Docker Hub (`espnet/espnet:runtime`, based on Ubuntu 16), that already contains a kaldi installation, using Cuda 10.0: +For example, a local build that the base image from Docker Hub (`espnet/espnet:runtime`, based on Ubuntu 16), that already contains a kaldi installation, using CUDA 10.0: ``` ./build.sh local 10.0 ``` @@ -78,7 +78,7 @@ Also, docker images can also be built based on the Ubuntu version specified in ` ./build.sh fully_local cpu ``` -Local container builds then are started by adding the flag `--is-local` when using `run.sh`, e.g., for the Cuda 10.0 image: +Local container builds then are started by adding the flag `--is-local` when using `run.sh`, e.g., for the CUDA 10.0 image: ``` $ ./run.sh --is-local --docker_cuda 10.0 --docker_gpu 0 ... ``` diff --git a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/espnet2_tutorial.md b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/espnet2_tutorial.md index 5bdec078cc5ae269401926583b4fd812956a9b94..ac30618c7afcfebcb3bac2e186045f0db1792626 100644 --- a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/espnet2_tutorial.md +++ b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/espnet2_tutorial.md @@ -2,7 +2,7 @@ We are planning a super major update, called `ESPnet2`. The developing status is still **under construction** yet, so please be very careful to use with understanding following cautions: - There might be fatal bugs related to essential parts. -- We haven't achieved comparable results to espnet1 on each task yet. +- We haven't achieved comparable results to ESPnet1 on each task yet. ## Main changing from ESPnet1 diff --git a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/parallelization.md b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/parallelization.md index 63a4b167b26ec53db9dc0b13eb13c33f893cf98a..eae19ff750eded66c4478998dae6b475e4460ecc 100644 --- a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/parallelization.md +++ b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/doc/parallelization.md @@ -2,7 +2,7 @@ Our recipes support some Job scheduling systems, SGE, PBS/Torque, and Slurm, according to [Parallelization in Kaldi](https://kaldi-asr.org/doc/queue.html). -By default, the job runs at local machine. If there are any Job scheduling systems in your environment, +By default, the Job runs at local machine. If there are any Job scheduling systems in your environment, you can submit more number of Jobs with multiple machines. Please ask the administrator to install it if you have multiple machines. @@ -31,9 +31,9 @@ nj=4 ${cmd} JOB=1:${nj} JOB.log echo JOB ``` -`JOB=1:${nj}` indicates the parallelization, which is known as "array-job", with `${nj}` number of jobs. -`JOB.log` is a destination of the stdout and stderr from jobs. -The string of `JOB` will be changed to the job number +`JOB=1:${nj}` indicates the parallelization, which is known as "array-job", with `${nj}` number of Jobs. +`JOB.log` is a destination of the stdout and stderr from Jobs. +The string of `JOB` will be changed to the Job number if it's included in the log file name or command line arguments. i.e. The following commands are almost equivalent to the above: @@ -46,7 +46,7 @@ wait ``` ## Configuration -You also need to modify the configuration file for a specific job scheduler to change command-line options to submit jobs e.g. queue setting, resource request, etc. +You also need to modify the configuration file for a specific Job scheduler to change command-line options to submit Jobs e.g. queue setting, resource request, etc. The following text is an example of `conf/queue.conf`. diff --git a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/docker/README.md b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/docker/README.md index f71c34c1d8847328931d8d64bc4fa0c93c781f49..326a46f9895d7023339e482323f88c87d6da0f0d 100644 --- a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/docker/README.md +++ b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/docker/README.md @@ -13,11 +13,11 @@ See https://espnet.github.io/espnet/docker.html ### Ubuntu 18.04 -Pytorch 1.3.1, No warp-ctc: +PyTorch 1.3.1, No warp-ctc: - [`cuda10.1-cudnn7` (*docker/prebuilt/gpu/10.1/cudnn7/Dockerfile*)](https://github.com/espnet/espnet/tree/master/docker/prebuilt/devel/gpu/10.1/cudnn7/Dockerfile) -Pytorch 1.0.1, warp-ctc: +PyTorch 1.0.1, warp-ctc: - [`cuda10.0-cudnn7` (*docker/prebuilt/gpu/10.0/cudnn7/Dockerfile*)](https://github.com/espnet/espnet/tree/master/docker/prebuilt/devel/gpu/10.0/cudnn7/Dockerfile) - [`cpu-u18` (*docker/prebuilt/devel/Dockerfile*)](https://github.com/espnet/espnet/tree/master/docker/prebuilt/devel/Dockerfile) diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/UIO.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/UIO.md index 9443b416acf8d86b081be0b803d667ac02db319d..10af86c049dc0076f6f78c8d6ed977665b4d8fe7 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/UIO.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/UIO.md @@ -2,9 +2,9 @@ In order to support the model training of industrial tens of millions of hours of speech dataset, the data processing method UIO (Unified IO) has been updated in WeNet. The document will introduce UIO from the following sections: -Necessity of upgrading IO mothod, System design of UIO, Validation experiments, Usage of UIO, Q&A. +Necessity of upgrading IO method, System design of UIO, Validation experiments, Usage of UIO, Q&A. -## Necessity of upgrading IO mothod +## Necessity of upgrading IO method The old IO method in WeNet is based on Pytorch's native Dataset. During training, it need to load all training audio paths and correspondingly labels into the memory at one time, then randomly read data. In the case of industrial-grade ultra-large-scale data (egs: more than 50,000 hours or 50 million or more audio), this method will cause the training diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/context.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/context.md index 881d119ea05e4a591d42cba0d4efd123fc56a80f..c30b40d7cfc4e892c4914676c94ba9b7793d49aa 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/context.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/context.md @@ -12,7 +12,7 @@ where, $\lambda$ is a tunable hyperparameter controlling how much the contextual ### Context Graph -If we want to improve the score of the word "cat", and the biasing score $\lambda\,\mathrm{log}\,P_C(\mathbf y)$ of each character is 0.25. The context graph can be constructed as follow: +If we want to improve the score of the word "cat", and the biasing score $\lambda\,\mathrm{log}\,P_C(\mathbf y)$ of each character is 0.25. The context graph can be constructed as follows: ![context graph](images/context_graph.png) @@ -57,7 +57,7 @@ In the process of CTC prefix beam search, each prefix needs to record the hot wo WeNet adopts the Lattice Faster Online Decoder from Kaldi for WFST beam search. We have to modify the `lattice-faster-decoder.cc` to support context biasing. -WFST beam search decodes in the TLG graph according to the CTC outputs. If we bias the input label of the TLG, we need to compose the context graph with the Token graph. Finally, we decide to bias TLG's output towards the contextual fst. We need to modify the `ProcessEmitting` and `ProcessNonemitting` functions as follow: +WFST beam search decodes in the TLG graph according to the CTC outputs. If we bias the input label of the TLG, we need to compose the context graph with the Token graph. Finally, we decide to bias TLG's output towards the contextual fst. We need to modify the `ProcessEmitting` and `ProcessNonemitting` functions as follows: ```c++ Elem *e_next = diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/jit_in_wenet.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/jit_in_wenet.md index 650090d9e6f6f19f0081df912f750eee3820a713..c2c5fc98f6aa4ce76f70b88e06c7d90e623c57a1 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/jit_in_wenet.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/jit_in_wenet.md @@ -19,7 +19,7 @@ script_model = torch.jit.script(model) script_model.save(os.path.join(args.model_dir, 'init.zip')) ``` -Two principles should be taken into consideration when we contribute our python code +Two principles should be taken into consideration when we contribute our Python code to WeNet, especially for the subclass of torch.nn.Module, and for the forward function. 1. Know what is allowed and what is disallowed. diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/tutorial_librispeech.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/tutorial_librispeech.md index 223f3b6a913def973a5ce3feb9b95d73ab9b491d..49f5e6a39c8f66d63094d874d753fe9dbb11c3a2 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/tutorial_librispeech.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/docs/tutorial_librispeech.md @@ -79,7 +79,7 @@ In this stage, `local/data_prep_torchaudio.sh` organizes the original data into If you want to train using your customized data, just organize the data into two files `wav.scp` and `text`, and start from `stage 1`. -#### Stage 1: Extract optinal cmvn features +#### Stage 1: Extract optional cmvn features ``` sh if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then @@ -194,7 +194,7 @@ Here is an example of the `data.list`, and please see the generated training fea {"key": "1455-134435-0002", "wav": "/mnt/nfs/ptm1/open-data/LibriSpeech/train-clean-100/1455/134435/1455-134435-0002.flac", "txt": "BUT LOUISE COULD NOT BE MADE HAPPY SHE FLEW INTO HALF INSANE FITS OF TEMPER DURING WHICH SHE WAS SOMETIMES SILENT SOMETIMES NOISY AND QUARRELSOME SHE SWORE AND CRIED OUT IN HER ANGER SHE GOT A KNIFE FROM THE KITCHEN AND THREATENED HER HUSBAND'S LIFE"} ``` -We aslo design another format for `data.list` named `shard` which is for big data training. +We also design another format for `data.list` named `shard` which is for big data training. Please see [gigaspeech](https://github.com/wenet-e2e/wenet/tree/main/examples/gigaspeech/s0)(10k hours) or [wenetspeech](https://github.com/wenet-e2e/wenet/tree/main/examples/wenetspeech/s0)(10k hours) for how to use `shard` style `data.list` if you want to apply WeNet on big data set(more than 5k). diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/aishell/NST/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/aishell/NST/README.md index 75fb1e43a9398ee1826a617882029e09e25f3b93..f972e0c39f3df0008eaaa425dac3154a4534084e 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/aishell/NST/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/aishell/NST/README.md @@ -2,7 +2,7 @@ Noisy Student Training (NST) has recently demonstrated extremely strong performance in Automatic Speech Recognition (ASR). -Here, we provide a recipe to run NST with `LM filter` strategy using AISHELL-1 as supervised data and WenetSpeech as unsupervised data from [this paper](https://arxiv.org/abs/2211.04717), where hypotheses with and without Language Model are generated and CER differences between them are utilized as a filter threshold to improve the ASR performances of non-target domain datas. +Here, we provide a recipe to run NST with `LM filter` strategy using AISHELL-1 as supervised data and WenetSpeech as unsupervised data from [this paper](https://arxiv.org/abs/2211.04717), where hypotheses with and without Language Model are generated and CER differences between them are utilized as a filter threshold to improve the ASR performances of non-target domain data. ## Table of Contents diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/openasr2021/s0/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/openasr2021/s0/README.md index 7b904dbfb7af31ac08fccc27095631ece8173762..3eda9c9693fc3216f58571c856684ac8996bac0a 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/openasr2021/s0/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/examples/openasr2021/s0/README.md @@ -1,6 +1,6 @@ # w2v-conformer based end-to-end model for Openasr2021 challenge -This is a example to use unsupervised pretrained w2v-conformer model to fintune [OpenASR2021](https://www.nist.gov/itl/iad/mig/openasr-challenge) constrained-plus tasks. +This is a example to use unsupervised pretrained w2v-conformer model to fine-tune [OpenASR2021](https://www.nist.gov/itl/iad/mig/openasr-challenge) constrained-plus tasks. We pretrain conformer encoders using wav2vec 2.0 pre-training method , which we called ch-w2v-conformer. The original pre-training works take raw waveforms as input. Unlike these works, we use MFCC features as inputs. diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/README.md index 3bece6d8e3fe71d01bf09922cb6f6569da189e5f..430ab16338c5ca5dccfc804665d80fd72ed61a53 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/README.md @@ -6,9 +6,9 @@ We are going to support the following platforms: 1. Various deep learning inference engines, such as LibTorch, ONNX, OpenVINO, TVM, and so on. 2. Various OS, such as android, iOS, Harmony, and so on. -3. Various AI chips, such as GPU, Horzion BPU, and so on. -4. Various hardware platforms, such as Raspberry Pi. -5. Various language binding, such as python and go. +3. Various AI chips, such as GPU, Horizion BPU, and so on. +4. Various hardware platforms, such as Raspberry pi. +5. Various language binding, such as Python and Go. Feel free to volunteer yourself if you are interested in trying out some items(they do not have to be on the list). diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/binding/python/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/binding/python/README.md index 993e65cb0f1d199da34854a1e9ebc6f25f4abb42..9555653787bb7bdeb591744f11f19edc11664693 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/binding/python/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/binding/python/README.md @@ -1,6 +1,6 @@ # Python Binding -This is a python binding of WeNet. +This is a Python binding of WeNet. WeNet is a production first and production ready end-to-end speech recognition toolkit. diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/core/kaldi/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/core/kaldi/README.md index 4eb9c9173b747686f00b658afc5e1e0dfdc17e68..47ffaad3d4f4a7b7b155e976b7ce55483b0638d7 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/core/kaldi/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/core/kaldi/README.md @@ -2,7 +2,7 @@ We use Kaldi decoder to implement TLG based language model integration, so we copied related files to this directory. The main changes are: -1. To minimize the change, we use the same directories tree as Kaldi. +1. To minimize the change, we use the same directory tree as Kaldi. 2. We replace Kaldi log system with glog in the following way. diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/cuda_decoders/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/cuda_decoders/README.md index f63307ff79b1e2dcf621ccf0154add09955e5313..5dc11a9e916b8c1af6c3d4aeb518cba965fafff1 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/cuda_decoders/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/cuda_decoders/README.md @@ -18,7 +18,7 @@ bash run.sh ### TODO: Performance of Small Offline ASR Model using Different Decoders -Benchmark(small offline conformer onnx fp16 model trained on Aishell1) based on Aishell1 test set with V100, the total audio duration is 36108.919 seconds. +Benchmark(small offline conformer ONNX FP16 model trained on Aishell1) based on Aishell1 test set with V100, the total audio duration is 36108.919 seconds. (Note: 80 concurrent tasks, service has been fully warm up.) |Decoding Method | decoding time(s) | WER (%) | diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt/README.md index 4ecca2dfcb6e5bd7a34c964c5192eb976573a13d..af42c630e391add51398d71c6ed0ed5b3e6132fb 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt/README.md @@ -1,4 +1,4 @@ -### Using Tensorrt for Triton ASR Server +### Using TensorRT for Triton ASR Server ```sh # using docker image runtime/gpu/Dockerfile/Dockerfile.server @@ -16,16 +16,16 @@ bash run_streaming_small_model.sh #### Performance of Small u2pp Model for Streaming ASR -Benchmark(small u2pp onnx) based on Aishell1 test set with server-A10 (16vCPU 60GB Memory)/client(4vCPU 16GB Memory), the total audio duration is 36108.919 seconds. +Benchmark(small u2pp ONNX) based on Aishell1 test set with server-A10 (16vCPU 60GB Memory)/client(4vCPU 16GB Memory), the total audio duration is 36108.919 seconds. -(Note: using non-simulate-streaming mode) +(Note: using non-simulated-streaming mode) |concurrent-tasks | processing time(s) | |----------|--------------------| -| 20 (onnx fp16) | 123.796 | -| 40 (onnx fp16) | 84.557 | -| 60 (onnx fp16) | 73.232 | -| 80 (onnx fp16) | 66.862 | -| 20 (trt fp16+layernorm plugin)| 90.582 | -| 40 (trt fp16+layernorm plugin)| 75.411 | -| 60 (trt fp16+layernorm plugin)| 69.602 | -| 80 (trt fp16+layernorm plugin)| 65.603 | \ No newline at end of file +| 20 (ONNX FP16) | 123.796 | +| 40 (ONNX FP16) | 84.557 | +| 60 (ONNX FP16) | 73.232 | +| 80 (ONNX FP16) | 66.862 | +| 20 (TRT FP16+layernorm plugin)| 90.582 | +| 40 (TRT FP16+layernorm plugin)| 75.411 | +| 60 (TRT FP16+layernorm plugin)| 69.602 | +| 80 (TRT FP16+layernorm plugin)| 65.603 | \ No newline at end of file diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt_fastertransformer/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt_fastertransformer/README.md index 0c0a2be84afae41ccffcddd8484b45566e202264..46e5bdebeacfbb4b3d360e51f77838226649bb07 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt_fastertransformer/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/gpu/tensorrt_fastertransformer/README.md @@ -23,7 +23,7 @@ With help of FasterTransformer wenet tensorrt plugins, the overall throughput co #### Performance of Small Model for AIShell2 -The following benchmakr shows the performance on T4 of a small Conformer model for AIShell2 case as defined in [WeNet AIShell2 Example](https://github.com/wenet-e2e/wenet/tree/main/examples/aishell2/s0). +The following benchmark shows the performance on T4 of a small Conformer model for AIShell2 case as defined in [WeNet AIShell2 Example](https://github.com/wenet-e2e/wenet/tree/main/examples/aishell2/s0). diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/openvino/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/openvino/README.md index 9e0009247e8b328c9426db50e9211be15a0dc9bc..fe95fff9a1e43613f51645e05aebc09c719f67e7 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/openvino/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/openvino/README.md @@ -38,7 +38,7 @@ cmake -DOPENVINO=ON -DTORCH=OFF -DWEBSOCKET=OFF -DGRPC=OFF .. make --jobs=$(nproc --all) ``` -(Optional) Some users may cannot easily download OpenVINO™ binary package from server due to firewall or proxy issue. If you failed to download by CMake script, you can download OpenVINO™ package by your selves and put the package to below path: +(Optional) Some users may cannot easily download OpenVINO™ binary package from server due to firewall or proxy issue. If you failed to download by CMake script, you can download OpenVINO™ package by yourselves and put the package to below path: ``` sh ${wenet_path}/runtime/openvino/fc_base/openvino-subbuild/openvino-populate-prefix/src/l_openvino_toolkit_ubuntu20_2022.3.0.9052.9752fafe8eb_x86_64.tgz diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/raspberrypi/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/raspberrypi/README.md index 9cb1943e7e7763cb5e17ae62037b2f67e8f1b417..ce22350f6bd43567ffec51147c96b6244ef63689 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/raspberrypi/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/runtime/raspberrypi/README.md @@ -1,4 +1,4 @@ -# WeNet & Raspberry PI (Cross Compile) +# WeNet & Raspberry pi (Cross Compile) * Step 1. Install cross compile tools in the PC. @@ -25,7 +25,7 @@ python -m wenet.bin.export_onnx_cpu \ # We use the quantified to speed up the inference, so rename it without the suffix `.quant` ``` -* Step 3. Build. The build requires cmake 3.14 or above. and Send the binary and libraries to Raspberry PI. +* Step 3. Build. The build requires cmake 3.14 or above. and Send the binary and libraries to Raspberry pi. ``` sh cmake -B build -DONNX=ON -DTORCH=OFF -DWEBSOCKET=OFF -DGRPC=OFF -DCMAKE_TOOLCHAIN_FILE=toolchains/aarch64-linux-gnu.toolchain.cmake @@ -34,7 +34,7 @@ scp build/bin/decoder_main pi@xxx.xxx.xxx:/path/to/wenet scp fc_base/onnxruntime-src/lib/libonnxruntime.so* pi@xxx.xxx.xxx:/path/to/wenet ``` -* Step 4. Testing, the RTF(real time factor) is shown in Raspberry PI's console. +* Step 4. Testing, the RTF(real time factor) is shown in Raspberry pi's console. ``` sh cd /path/to/wenet diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/wenet/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/wenet/README.md index 48d491aea1d587d87870c41f76006f0b3dfdcd39..1c86f04c8ba5cd8391f8abea307f953620b83378 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/wenet/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/wenet/README.md @@ -17,7 +17,7 @@ Here is a brief introduction of each module(directory). * `cif`: Continuous Integrate-and-Fire implemented, please refer [paper](https://arxiv.org/pdf/1905.11235.pdf) `transducer`, `squeezeformer`, `efficient_conformer`, and `cif` are all based on `transformer`, -they resue a lot of the common blocks of `tranformer`. +they reuse a lot of the common blocks of `transformer`. **If you want to contribute your own x-former, please reuse the current code as much as possible**. diff --git a/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/public_address_statement.md b/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/public_address_statement.md index 61dcdd53cafcb2d6296fa4b03c401b471b0bee9e..6b33289bcf3a91703d6c6000bc35a88dd915a134 100644 --- a/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/public_address_statement.md +++ b/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/public_address_statement.md @@ -1,6 +1,6 @@ | 类型 | 开源代码地址 | 文件名 | 公网IP地址/公网URL地址/域名/邮箱地址 | 用途说明 | | ---- | ------------ | ------ | ------------------------------------ | -------- | -| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/timm/utils/model_ema.py|EfficientNetV2_for_PyTorch/fused_ema.py | https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAvera | 模型相关说明 | -| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/train.py|EfficientNetV2_for_PyTorch/train.py | https://github.com/pytorch/examples/tree/master/imagen | 源码实现 | -| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/train.py|EfficientNetV2_for_PyTorch/train.py | https://github.com/NVIDIA/apex/tree/master/examples/imagen | 源码实现 | -| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/.github/ISSUE_TEMPLATE/config.yml|EfficientNetV2_for_PyTorch/train.py | https://github.com/rwightm | 源码实现 | \ No newline at end of file +| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/timm/utils/model_ema.py|EfficientNetV2_for_PyTorch/fused_ema.py | https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAverage | 模型相关说明 | +| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/train.py|EfficientNetV2_for_PyTorch/train.py | https://github.com/pytorch/examples/tree/master/imagenet | 源码实现 | +| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/train.py|EfficientNetV2_for_PyTorch/train.py | https://github.com/NVIDIA/apex/tree/master/examples/imagenet | 源码实现 | +| 开源代码引入 | https://github.com/rwightman/pytorch-image-models/blob/master/.github/ISSUE_TEMPLATE/config.yml|EfficientNetV2_for_PyTorch/train.py | https://github.com/rwightman | 源码实现 | \ No newline at end of file diff --git a/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README_raw.md b/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README_raw.md index 8c3bd3bb9b288ea01b099accb3fc06b57b0414cc..200b8f6a84b3092ea188d38516a988e92d57872d 100644 --- a/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README_raw.md +++ b/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README_raw.md @@ -20,7 +20,7 @@ the following parameters: ### AlexNet and VGG Since `AlexNet` and the original `VGG` architectures do not include batch -normalization, the default initial learning rate `--lr 0.1` is to high. +normalization, the default initial learning rate `--lr 0.1` is too high. ``` python -m torch.distributed.launch --nproc_per_node=8 --use_env train.py\ @@ -31,16 +31,16 @@ Here `$MODEL` is one of `alexnet`, `vgg11`, `vgg13`, `vgg16` or `vgg19`. Note that `vgg11_bn`, `vgg13_bn`, `vgg16_bn`, and `vgg19_bn` include batch normalization and thus are trained with the default parameters. -### ResNext-50 32x4d +### ResNeXt-50 32x4d ``` python -m torch.distributed.launch --nproc_per_node=8 --use_env train.py\ --model resnext50_32x4d --epochs 100 ``` -### ResNext-101 32x8d +### ResNeXt-101 32x8d -On 8 nodes, each with 8 GPUs (for a total of 64 GPUS) +On 8 nodes, each with 8 GPUs (for a total of 64 GPUs) ``` python -m torch.distributed.launch --nproc_per_node=8 --use_env train.py\ --model resnext101_32x8d --epochs 100 diff --git a/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md b/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md index cb17cf8151b6ed90d8696df2eba883c79121dbaa..8bd62a3a34144ecdfc9abbfb1f288510d3a76e1e 100644 --- a/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md @@ -1,4 +1,4 @@ -# Resnet50 for PyTorch +# ResNet50 for PyTorch - [概述](#概述) - [准备训练环境](#准备训练环境) diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md index 1f762baa0f524680cfee5638e97bce570570566d..1dd724ab1f31826d1e0ca65da5d3324cd5c23a8b 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md @@ -201,7 +201,7 @@ MMClassification 是一款基于 PyTorch 的开源图像分类工具箱，是 Op - 挂载vNPU，并声明shm内存（避免容器内存不足无法拉起训练） ``` docker run -it \ - --device=/dev/vdavinci100:/dev/davinci100 \ # 挂载切分好的vNPU + --device=/dev/davinci100:/dev/davinci100 \ # 挂载切分好的vNPU --device=/dev/davinci_manager \ --device=/dev/devmm_svm \ --device=/dev/hisi_hdc \ diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/efficientnet/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/efficientnet/README.md index 832f5c6b2f9d65acb9dcc920547a4e82292a4da8..5baa683420467e01d02a13364753f30b849e72bf 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/efficientnet/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/efficientnet/README.md @@ -52,7 +52,7 @@ Note: In MMClassification, we support training with AutoAugment, don't support A ``` @inproceedings{tan2019efficientnet, - title={Efficientnet: Rethinking model scaling for convolutional neural networks}, + title={EfficientNet: Rethinking model scaling for convolutional neural networks}, author={Tan, Mingxing and Le, Quoc}, booktitle={International Conference on Machine Learning}, pages={6105--6114}, diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mlp_mixer/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mlp_mixer/README.md index 5ec98871b6da4406551100c617200104f478860d..0571169940ef0d3bc0429363a396dde18b632ea5 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mlp_mixer/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mlp_mixer/README.md @@ -1,4 +1,4 @@ -# Mlp-Mixer +# MLP-Mixer > [MLP-Mixer: An all-MLP Architecture for Vision](https://arxiv.org/abs/2105.01601) diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mobilenet_v3/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mobilenet_v3/README.md index 737c4d32ec01e65f464f73cc68cbb245021c3a99..81f5b079e5202409011d40c5b3766f5d1face0a3 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mobilenet_v3/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/mobilenet_v3/README.md @@ -1,4 +1,4 @@ -# MobileNet V3 +# MobileNetV3 > [Searching for MobileNetV3](https://arxiv.org/abs/1905.02244) diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/repmlp/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/repmlp/README.md index 453346352cd5ed36f5b9fb82242b72952b0ab0d3..f817d1acdc510bc397d4400d7691f46dfcd3a52e 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/repmlp/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/repmlp/README.md @@ -85,7 +85,7 @@ classifier.backbone.switch_to_deploy() ``` @article{ding2021repmlp, - title={Repmlp: Re-parameterizing convolutions into fully-connected layers for image recognition}, + title={RepMLP: Re-parameterizing convolutions into fully-connected layers for image recognition}, author={Ding, Xiaohan and Xia, Chunlong and Zhang, Xiangyu and Chu, Xiaojie and Han, Jungong and Ding, Guiguang}, journal={arXiv preprint arXiv:2105.01883}, year={2021} diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/seresnet/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/seresnet/README.md index ccfd1d156ed3dc0c2d392069f80d0f88669bd2e7..c7965f002e9128a2dce5abebe33a7b115d9b4121 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/seresnet/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/seresnet/README.md @@ -25,7 +25,7 @@ The central building block of convolutional neural networks (CNNs) is the convol ``` @inproceedings{hu2018squeeze, - title={Squeeze-and-excitation networks}, + title={Squeeze-and-Excitation Networks}, author={Hu, Jie and Shen, Li and Sun, Gang}, booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition}, pages={7132--7141}, diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/shufflenet_v1/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/shufflenet_v1/README.md index fd131279210a4e9c8f44f372b02daa76d8c88a15..a1d13b52b58c059339a2de3511423b6252f1e770 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/shufflenet_v1/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/shufflenet_v1/README.md @@ -24,7 +24,7 @@ We introduce an extremely computation-efficient CNN architecture named ShuffleNe ``` @inproceedings{zhang2018shufflenet, - title={Shufflenet: An extremely efficient convolutional neural network for mobile devices}, + title={ShuffleNet: An extremely efficient convolutional neural network for mobile devices}, author={Zhang, Xiangyu and Zhou, Xinyu and Lin, Mengxiao and Sun, Jian}, booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition}, pages={6848--6856}, diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/twins/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/twins/README.md index 87e72941f4a44519efd1232d0651dc9203986caa..5759c6a91fe46aa71d521136d79f2b187fd162fa 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/twins/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/configs/twins/README.md @@ -34,6 +34,6 @@ Very recently, a variety of vision transformer architectures for dense predictio title={Twins: Revisiting spatial attention design in vision transformers}, author={Chu, Xiangxiang and Tian, Zhi and Wang, Yuqing and Zhang, Bo and Ren, Haibing and Wei, Xiaolin and Xia, Huaxia and Shen, Chunhua}, journal={arXiv preprint arXiv:2104.13840}, - year={2021}altgvt + year={2021} } ``` diff --git a/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md b/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md index fb4a23ce2f27ca646a61aa5cb9196da3bc316c0d..88f452040ed0d530159b2415ee31107a85116a2d 100644 --- a/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md @@ -181,7 +181,7 @@ DB(Differentiable Binarization)是一种使用可微分二值图来实时文字 --addr //主机地址 --num_workers //加载数据进程数 --epochs //重复训练次数 - --batch-size //训练批次大小，默认：240 + --batch_size //训练批次大小，默认：240 --lr //初始学习率 --amp //是否使用混合精度 ``` diff --git a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/datasets/README.md b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/datasets/README.md index b888ee9714697fcbc03aa6b08f920edd8924fec3..6ea714874c18e43c8ffc6ee76c4b4c3524112633 100755 --- a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/datasets/README.md +++ b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/datasets/README.md @@ -2,11 +2,11 @@ A dataset can be used by accessing [DatasetCatalog](https://detectron2.readthedocs.io/modules/data.html#detectron2.data.DatasetCatalog) for its data, or [MetadataCatalog](https://detectron2.readthedocs.io/modules/data.html#detectron2.data.MetadataCatalog) for its metadata (class names, etc). -This document explains how to setup the builtin datasets so they can be used by the above APIs. +This document explains how to setup the built-in datasets so they can be used by the above APIs. [Use Custom Datasets](https://detectron2.readthedocs.io/tutorials/datasets.html) gives a deeper dive on how to use `DatasetCatalog` and `MetadataCatalog`, and how to add new datasets to them. -Detectron2 has builtin support for a few datasets. +Detectron2 has built-in support for a few datasets. The datasets are assumed to exist in a directory specified by the environment variable `DETECTRON2_DATASETS`. Under this directory, detectron2 will look for datasets in the structure described below, if needed. @@ -18,11 +18,11 @@ $DETECTRON2_DATASETS/ VOC20{07,12}/ ``` -You can set the location for builtin datasets by `export DETECTRON2_DATASETS=/path/to/datasets`. +You can set the location for built-in datasets by `export DETECTRON2_DATASETS=/path/to/datasets`. If left unset, the default is `./datasets` relative to your current working directory. The [model zoo](https://github.com/facebookresearch/detectron2/blob/master/MODEL_ZOO.md) -contains configs and models that use these builtin datasets. +contains configs and models that use these built-in datasets. ## Expected dataset structure for COCO instance/keypoint detection: @@ -37,7 +37,7 @@ coco/ You can use the 2014 version of the dataset as well. -Some of the builtin tests (`dev/run_*_tests.sh`) uses a tiny version of the COCO dataset, +Some of the built-in tests (`dev/run_*_tests.sh`) uses a tiny version of the COCO dataset, which you can download with `./prepare_for_tests.sh`. ## Expected dataset structure for PanopticFPN: diff --git a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/detectron2/utils/README.md b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/detectron2/utils/README.md index 9765b24a730b77556104187ac3ef5439ab0859fd..7e9249f2997c6a6eabb371637de883e855ee0d20 100755 --- a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/detectron2/utils/README.md +++ b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/detectron2/utils/README.md @@ -1,5 +1,5 @@ # Utility functions -This folder contain utility functions that are not used in the +This folder contains utility functions that are not used in the core library, but are useful for building models or training code using the config system. diff --git a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/notes/changelog.md b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/notes/changelog.md index f83eee64ce149969bc7a1fb7762e6fa876f75ea6..f274f4d4f3e26df8cf9d724da77d85c068e6166d 100755 --- a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/notes/changelog.md +++ b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/notes/changelog.md @@ -14,7 +14,7 @@ But we try to reduce users' disruption by the following ways: otherwise noted in the documentation. They are less likely to be broken, but if needed, will trigger a deprecation warning for a reasonable period before getting broken, and will be documented in release logs. -* Others functions/classses/attributes are considered internal, and are more likely to change. +* Others functions/classes/attributes are considered internal, and are more likely to change. However, we're aware that some of them may be already used by other projects, and in particular we may use them for convenience among projects under `detectron2/projects`. For such APIs, we may treat them as stable APIs and also apply the above strategies. diff --git a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/datasets.md b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/datasets.md index 8926cd7fc6f12a2453f497f208150f9c07e693f6..2b4b092703978195106a3eab11bb2d89a02d2c4c 100755 --- a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/datasets.md +++ b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/datasets.md @@ -4,7 +4,7 @@ This document explains how the dataset APIs ([DatasetCatalog](../modules/data.html#detectron2.data.DatasetCatalog), or [MetadataCatalog](../modules/data.html#detectron2.data.MetadataCatalog)) work, and how to use them to add custom datasets. -Datasets that have builtin support in detectron2 are listed in [builtin datasets](builtin_datasets.md). +Datasets that have built-in support in detectron2 are listed in [built-in datasets](builtin_datasets.md). If you want to use a custom dataset while also reusing detectron2's data loaders, you will need to: @@ -37,7 +37,7 @@ The function must return the same data if called multiple times. The registration stays effective until the process exits. The function can do arbitrary things and should returns the data in either of the following formats: -1. Detectron2's standard dataset dict, described below. This will make it work with many other builtin +1. Detectron2's standard dataset dict, described below. This will make it work with many other built-in features in detectron2, so it's recommended to use it when it's sufficient. 2. Any custom format. You can also return arbitrary dicts in your own format, such as adding extra keys for new tasks. @@ -142,7 +142,7 @@ from detectron2.data import MetadataCatalog MetadataCatalog.get("my_dataset").thing_classes = ["person", "dog"] ``` -Here is a list of metadata keys that are used by builtin features in detectron2. +Here is a list of metadata keys that are used by built-in features in detectron2. If you add your own dataset without these metadata, some features may be unavailable to you: @@ -180,7 +180,7 @@ Some additional metadata that are specific to the evaluation of certain datasets * `json_file`: The COCO annotation json file. Used by COCO evaluation for COCO-format datasets. * `panoptic_root`, `panoptic_json`: Used by panoptic evaluation. -* `evaluator_type`: Used by the builtin main training script to select +* `evaluator_type`: Used by the built-in main training script to select evaluator. Don't use it in a new training script. You can just provide the [DatasetEvaluator](../modules/evaluation.html#detectron2.evaluation.DatasetEvaluator) for your dataset directly in your main script. diff --git a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/training.md b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/training.md index 3b99662bbf720ae9a18ac331c75a487b8ff96304..744a055bc0c4241679ecdacaa6596bc60f1293f8 100755 --- a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/training.md +++ b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/docs/tutorials/training.md @@ -14,7 +14,7 @@ Any customization on the training logic is then easily controlled by the user. ### Trainer Abstraction -We also provide a standarized "trainer" abstraction with a +We also provide a standardized "trainer" abstraction with a hook system that helps simplify the standard training behavior. It includes the following two instantiations: diff --git a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/tests/README.md b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/tests/README.md index f560384045ab4f6bc2beabef1170308fca117eb3..8f3155f2e665b31c37a2f254298234c72fb6f998 100755 --- a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/tests/README.md +++ b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/tests/README.md @@ -1,6 +1,6 @@ ## Unit Tests -To run the unittests, do: +To run the unittest, do: ``` cd detectron2 python -m unittest discover -v -s ./tests diff --git a/PyTorch/built-in/cv/semantic_segmentation/DeepLabv3+_ID1695_for_PyTorch/public_address_statement.md b/PyTorch/built-in/cv/semantic_segmentation/DeepLabv3+_ID1695_for_PyTorch/public_address_statement.md index 16b8e56faaf6aa7a01824436cf7171cd3ae7f952..538746f054e45b3e5d65f7a562d3812a2c2a546f 100644 --- a/PyTorch/built-in/cv/semantic_segmentation/DeepLabv3+_ID1695_for_PyTorch/public_address_statement.md +++ b/PyTorch/built-in/cv/semantic_segmentation/DeepLabv3+_ID1695_for_PyTorch/public_address_statement.md @@ -7,22 +7,22 @@ | 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/9135e104a7a51ea9effa9c6676a2fcffe6a6a2e6/modeling/backbone/mobilenet.py | DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/mobilenet.py | http://jeff95.me/models/mobilenet_v2-6a65762b.pth | 下载权重文件 | | 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/9135e104a7a51ea9effa9c6676a2fcffe6a6a2e6/modeling/backbone/resnet.py | DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/resnet.py | https://download.pytorch.org/models/resnet101-5d3b4d8f.pth | 下载权重文件 | | 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/9135e104a7a51ea9effa9c6676a2fcffe6a6a2e6/modeling/backbone/xception.py | DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/xception.py | http://data.lip6.fr/cadene/pretrainedmodels/xception-b5690688.pth | 下载权重文件 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_resnet.py|DeepLabv3+_ID1695_for_PyTorch/doc/deeplab_resnet.py | https://download.pytorch.org/models/resnet101-5d3b4d8f.p | 模型相关说明 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_xception.py|DeepLabv3+_ID1695_for_PyTorch/doc/deeplab_xception.py | http://data.lip6.fr/cadene/pretrainedmodels/xception-b5690688.p | 模型相关说明 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/utils/lr_scheduler.py|DeepLabv3+_ID1695_for_PyTorch/utils/lr_scheduler.py | zhang.hang@rutgers.e | 开发者邮箱配置 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/backbone/drn.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/drn.py | http://dl.yf.io/dr | 模型相关说明 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/backbone/drn.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/drn.py | https://download.pytorch.org/models/resnet50-19c8e357.p | 模型相关说明 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/backbone/mobilenet.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/mobilenet.py | http://jeff95.me/models/mobilenet_v2-6a65762b.p | 模型相关说明 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/batchnorm.py | maojiayuan@gmail.c | 开发者邮箱配置 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/batchnorm.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTor | 源码实现 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_resnet.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/resnet.py | https://download.pytorch.org/models/resnet101-5d3b4d8f.p | 模型相关说明 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_resnet.py|DeepLabv3+_ID1695_for_PyTorch/doc/deeplab_resnet.py | https://download.pytorch.org/models/resnet101-5d3b4d8f.pth | 模型相关说明 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_xception.py|DeepLabv3+_ID1695_for_PyTorch/doc/deeplab_xception.py | http://data.lip6.fr/cadene/pretrainedmodels/xception-b5690688.pth | 模型相关说明 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/utils/lr_scheduler.py|DeepLabv3+_ID1695_for_PyTorch/utils/lr_scheduler.py | zhang.hang@rutgers.edu | 开发者邮箱配置 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/backbone/drn.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/drn.py | http://dl.yf.io/drn | 模型相关说明 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/backbone/drn.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/drn.py | https://download.pytorch.org/models/resnet50-19c8e357.pth | 模型相关说明 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/backbone/mobilenet.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/mobilenet.py | http://jeff95.me/models/mobilenet_v2-6a65762b.pth | 模型相关说明 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/batchnorm.py | maojiayuan@gmail.com | 开发者邮箱配置 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/batchnorm.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTorch | 源码实现 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_resnet.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/resnet.py | https://download.pytorch.org/models/resnet101-5d3b4d8f.pth | 模型相关说明 | | 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/batchnorm.py | http://tetexiao.co | 模型相关说明 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/comm.py | maojiayuan@gmail.c | 开发者邮箱配置 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/comm.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTor | 源码实现 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/replicate.py | maojiayuan@gmail.c | 开发者邮箱配置 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/replicate.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTor | 源码实现 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_xception.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/xception.py | http://data.lip6.fr/cadene/pretrainedmodels/xception-b5690688.p | 模型相关说明 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/unittest.py | maojiayuan@gmail.c | 开发者邮箱配置 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/unittest.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTor | 源码实现 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/__init__.py | maojiayuan@gmail.c | 开发者邮箱配置 | -| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/__init__.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTor | 源码实现 | \ No newline at end of file +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/comm.py | maojiayuan@gmail.com | 开发者邮箱配置 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/comm.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTorch | 源码实现 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/replicate.py | maojiayuan@gmail.com | 开发者邮箱配置 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/replicate.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTorch | 源码实现 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/doc/deeplab_xception.py|DeepLabv3+_ID1695_for_PyTorch/modeling/backbone/xception.py | http://data.lip6.fr/cadene/pretrainedmodels/xception-b5690688.pth | 模型相关说明 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/unittest.py | maojiayuan@gmail.com | 开发者邮箱配置 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/unittest.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTorch | 源码实现 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/__init__.py | maojiayuan@gmail.com | 开发者邮箱配置 | +| 开源代码引入 | https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/modeling/sync_batchnorm/batchnorm.py|DeepLabv3+_ID1695_for_PyTorch/modeling/sync_batchnorm/__init__.py | https://github.com/vacancy/Synchronized-BatchNorm-PyTorch | 源码实现 | \ No newline at end of file diff --git a/PyTorch/built-in/docs/dist_train_demo/train_8p_shell.py b/PyTorch/built-in/docs/dist_train_demo/train_8p_shell.py index 153e9d19c3a03e3285b3556295e73c4bfe2af7b5..c38719f25a52e9ab8344502fae77c533273ff0b3 100644 --- a/PyTorch/built-in/docs/dist_train_demo/train_8p_shell.py +++ b/PyTorch/built-in/docs/dist_train_demo/train_8p_shell.py @@ -99,7 +99,7 @@ def train(local_rank, world_size, args): def main(): args = get_train_args() - world_size = int(os.environ("WORLD_SIZE")) + world_size = int(os.environ["WORLD_SIZE"]) local_rank = int(os.environ["LOCAL_RANK"]) train(local_rank, world_size, args) diff --git a/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/public_address_statement.md b/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/public_address_statement.md index 3dd99c229ccc0abe5c76ddadb432ba9d7634d2f5..e44c6c5c0e71f05e737f7c3b0691ea720cd11da3 100644 --- a/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/public_address_statement.md +++ b/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/public_address_statement.md @@ -31,7 +31,7 @@ | 开源代码引入 | https://github.com/NVIDIA/DeepLearningExamples/blob/master/PyTorch/LanguageModeling/BERT/data/SquadDownloader.py |Bert-Squad_ID0470_for_PyTorch/data/SquadDownloader.py |https://rajpurkar.github.io/SQuAD-explorer/dataset/dev-v2.0.json |SQuAD模型在开源社区上的dev-v2.0.json的下载链接| | 开源代码引入 | https://github.com/NVIDIA/DeepLearningExamples/blob/master/PyTorch/LanguageModeling/BERT/data/SquadDownloader.py |Bert-Squad_ID0470_for_PyTorch/data/SquadDownloader.py |https://worksheets.codalab.org/rest/bundles/0x6b567e1cf2e041ec80d7098f031c5c9e/contents/blob/'|SQuAD模型在开源社区上的evaluate-v2.0.py的下载链接| | 开源代码引入 | https://github.com/NVIDIA/DeepLearningExamples/blob/master/PyTorch/LanguageModeling/BERT/data/WikiDownloader.py | Bert-Squad_ID0470_for_PyTorch/data/WikiDownloader.py |https://dumps.wikimedia.org/enwiki/latest/enwiki-latest-pages-articles.xml.bz2|enwiki在开源社区上的enwiki-latest-pages-articles.xml.bz2的下载链接| -| 开源代码引入 | https://github.com/NVIDIA/DeepLearningExamples/blob/master/PyTorch/LanguageModeling/BERT/data/WikiDownloader.py |Bert-Squad_ID0470_for_PyTorch/data/WikiDownloader.py |https://dumps.wikimedia.org/zhwiki/latest/zhwiki-latest-pages-articles.xml.bz2|zhwiki在开源社区上的zhwiki-latest-pages-articles.xml.bz的下载链接| +| 开源代码引入 | https://github.com/NVIDIA/DeepLearningExamples/blob/master/PyTorch/LanguageModeling/BERT/data/WikiDownloader.py |Bert-Squad_ID0470_for_PyTorch/data/WikiDownloader.py |https://dumps.wikimedia.org/zhwiki/latest/zhwiki-latest-pages-articles.xml.bz2|zhwiki在开源社区上的zhwiki-latest-pages-articles.xml.bz2的下载链接| | 开源代码引入 | https://github.com/huggingface/transformers/blob/main/transformers/docs/source/en/main_classes/processors.mdx |Bert-Squad_ID0470_for_PyTorch/data/squad/squad_download.sh | https://rajpurkar.github.io/SQuAD-explorer/dataset/train-v1.1.json | SQuAD_train-v1.1在开源社区上的json下载链接| | 开源代码引入 | https://github.com/huggingface/transformers/blob/main/transformers/docs/source/en/main_classes/processors.mdx |Bert-Squad_ID0470_for_PyTorch/data/squad/squad_download.sh | https://rajpurkar.github.io/SQuAD-explorer/dataset/dev-v1.1.json | SQuAD_dev-v1.1在开源社区上的json下载链接| | 开发引入 | / |Bert-Squad_ID0470_for_PyTorch/url.ini | https://worksheets.codalab.org/rest/bundles/0xbcd57bee090b421c982906709c8c27e1/contents |SQuAD_evaluate-v1.1在开源社区上的py下载链接| diff --git a/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md b/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md index 329de42a4b9f490b3a943b8a912fd4f975786503..75eacb95ebdc275af9b109794ff309dc51f9b393 100644 --- a/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md +++ b/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md @@ -309,6 +309,8 @@ BERT的全称是Bidirectional Encoder Representation from Transformers，即双 # Bert_Base_Chinese模型-推理指导 + 最新的推理指导参考[Bert_Base_Chinese模型-推理指导](https://gitee.com/ascend/ModelZoo-PyTorch/blob/master/ACL_PyTorch/built-in/nlp/Bert_Base_Chinese_for_Pytorch/README.md) + - [概述](#ZH-CN_TOPIC_0000001172161501) - [输入输出数据](#section540883920406) @@ -333,7 +335,7 @@ BERT的全称是Bidirectional Encoder Representation from Transformers，即双 ```shell url=https://huggingface.co/bert-base-chinese commit_id=38fda776740d17609554e879e3ac7b9837bdb5ee - mode_name=Bert_Base_Chinese + model_name=Bert_Base_Chinese ``` ### 输入输出数据 @@ -361,7 +363,7 @@ BERT的全称是Bidirectional Encoder Representation from Transformers，即双 | 配套 | 版本 | 环境准备指导 | | ------------------------------------------------------------ | ------- | ------------------------------------------------------------ | -| 固件与驱动 | 1.0.17 | [Pytorch框架推理环境准备](https://www.hiascend.com/document/detail/zh/ModelZoo/pytorchframework/pies) | +| 固件与驱动 | 1.0.17 | [PyTorch框架推理环境准备](https://www.hiascend.com/document/detail/zh/ModelZoo/pytorchframework/pies) | | CANN | 6.0.RC1 | - | | Python | 3.7.5 | - | | PyTorch | 1.5.0+ | - | @@ -377,7 +379,7 @@ BERT的全称是Bidirectional Encoder Representation from Transformers，即双 ``` git clone https://gitee.com/ascend/ModelZoo-PyTorch.git # 克隆仓库的代码 git checkout master # 切换到对应分支 - cd ACL_PyTorch/built-in/nlp/Bert_Base_Chinese_for_Pytorch # 切换到模型的代码仓目录 + cd /ModelZoo-PyTorch/ACL_PyTorch/built-in/nlp/Bert_Base_Chinese_for_Pytorch # 切换到模型的代码仓目录 ``` 2. 安装依赖。 @@ -565,7 +567,7 @@ BERT的全称是Bidirectional Encoder Representation from Transformers，即双 ## 其他下游任务 -+ [序列标注(Sequence Labeling)](downstream_tasks/sequence_labeling/README.md) ++ [序列标注(Sequence Labeling)](https://gitee.com/ascend/ModelZoo-PyTorch/blob/master/ACL_PyTorch/built-in/nlp/Bert_Base_Chinese_for_Pytorch/downstream_tasks/sequence_labeling/README.md) # 公网地址说明 diff --git a/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README_raw.md b/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README_raw.md index 047e1b768642ed9b012d5dc761439456c8689442..9a899afeec2671b73f9150383ae593d3a0242de3 100644 --- a/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README_raw.md +++ b/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README_raw.md @@ -7,7 +7,7 @@ Latest ReleaseBuild StatusDocumentation Status - CicleCI Status + CircleCI Status

-------------------------------------------------------------------------------- @@ -42,7 +42,7 @@ We provide reference implementations of various sequence modeling papers: + [RoBERTa: A Robustly Optimized BERT Pretraining Approach (Liu et al., 2019)](examples/roberta/README.md) + [Facebook FAIR's WMT19 News Translation Task Submission (Ng et al., 2019)](examples/wmt19/README.md) + [Jointly Learning to Align and Translate with Transformer Models (Garg et al., 2019)](examples/joint_alignment_translation/README.md ) - + [Multilingual Denoising Pre-training for Neural Machine Translation (Liu et at., 2020)](examples/mbart/README.md) + + [Multilingual Denoising Pre-training for Neural Machine Translation (Liu et al., 2020)](examples/mbart/README.md) + [Neural Machine Translation with Byte-Level Subwords (Wang et al., 2020)](examples/byte_level_bpe/README.md) + [Unsupervised Quality Estimation for Neural Machine Translation (Fomicheva et al., 2020)](examples/unsupervised_quality_estimation/README.md) + [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations (Baevski et al., 2020)](examples/wav2vec/README.md) @@ -55,9 +55,9 @@ We provide reference implementations of various sequence modeling papers: + [Robust wav2vec 2.0: Analyzing Domain Shift in Self-Supervised Pre-Training (Hsu, et al., 2021)](https://arxiv.org/abs/2104.01027) + [Unsupervised Speech Recognition (Baevski, et al., 2021)](https://arxiv.org/abs/2105.11084) + [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition (Xu et al., 2021)](https://arxiv.org/abs/2109.11680) - + [VideoCLIP: Contrastive Pre-training for Zero-shot Video-Text Understanding (Xu et. al., 2021)](https://arxiv.org/pdf/2109.14084.pdf) - + [VLM: Task-agnostic Video-Language Model Pre-training for Video Understanding (Xu et. al., 2021)](https://aclanthology.org/2021.findings-acl.370.pdf) - + [NormFormer: Improved Transformer Pretraining with Extra Normalization (Shleifer et. al, 2021)](examples/normformer/README.md) + + [VideoCLIP: Contrastive Pre-training for Zero-shot Video-Text Understanding (Xu et al., 2021)](https://arxiv.org/pdf/2109.14084.pdf) + + [VLM: Task-agnostic Video-Language Model Pre-training for Video Understanding (Xu et al., 2021)](https://aclanthology.org/2021.findings-acl.370.pdf) + + [NormFormer: Improved Transformer Pretraining with Extra Normalization (Shleifer et al., 2021)](examples/normformer/README.md) * **Non-autoregressive Transformers** + Non-Autoregressive Neural Machine Translation (Gu et al., 2017) + Deterministic Non-Autoregressive Neural Sequence Modeling by Iterative Refinement (Lee et al. 2018) diff --git a/PyTorch/contrib/CONTRIBUTING.md b/PyTorch/contrib/CONTRIBUTING.md index 2fbd07b96d018ad1d0706e5c953bb9a9571c326e..6322fda47a1805b7466f1781b10915cf032e3abe 100644 --- a/PyTorch/contrib/CONTRIBUTING.md +++ b/PyTorch/contrib/CONTRIBUTING.md @@ -12,7 +12,7 @@ Ascend ModelZoo，欢迎各位开发者 **一、源码** -1、训练及在线推理请使用python代码实现，Ascend平台离线推理请使用C++或python代码，符合第四部分编码规范 +1、训练及在线推理请使用Python代码实现，Ascend平台离线推理请使用C++或Python代码，符合第四部分编码规范 2、参考[sample](https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow/Official/nlp/Transformer_for_TensorFlow) @@ -287,7 +287,7 @@ PerfStatus:PERFECT/OK/POK/NOK - 规范标准 -1、C++代码遵循google编程规范：Google C++ Coding Guidelines；单元测测试遵循规范： Googletest Primer。 +1、C++代码遵循Google编程规范：Google C++ Coding Guidelines；单元测试遵循规范： Google Test Primer。 2、Python代码遵循PEP8规范：Python PEP 8 Coding Style；单元测试遵循规范： pytest diff --git a/PyTorch/contrib/ENVIRONMENTS.md b/PyTorch/contrib/ENVIRONMENTS.md index 05916b9fab90a43abd5fbf939891842633985c61..9f5be1db12e52baf6c215485deb1b7c08b8fbc8a 100644 --- a/PyTorch/contrib/ENVIRONMENTS.md +++ b/PyTorch/contrib/ENVIRONMENTS.md @@ -15,7 +15,7 @@ Shell脚本中，访问时用提供变量即可 | CV | COCO_val | $COCO_val | | CV | SBD | $SBD | | CV | VOC2012 | $VOC2012 | -| CV | Cityscape | $cityscapes_dataset | +| CV | Cityscapes | $cityscapes_dataset | | CV | icdar2015 | $icdar2015_train | | NLP | Wikipedia_CN | $Wikipedia_CN | | NLP | WMT_ENDE | $WMT_ENDE | diff --git a/PyTorch/contrib/audio/wav2vec2.0/examples/wav2vec/unsupervised/README.md b/PyTorch/contrib/audio/wav2vec2.0/examples/wav2vec/unsupervised/README.md index 0b213fd202d04bce2149936ec149c23c6d483745..15679d4e06337b4c03a6eff3b222935b19f85d99 100644 --- a/PyTorch/contrib/audio/wav2vec2.0/examples/wav2vec/unsupervised/README.md +++ b/PyTorch/contrib/audio/wav2vec2.0/examples/wav2vec/unsupervised/README.md @@ -1,6 +1,6 @@ # wav2vec Unsupervised (wav2vec-U) -Wav2vec Unsupervised (wav2vec-U) is a framework for building speech recognition systems without any labeled training data as described in [Unsupervised Speech Recognition (Baevski et al., 2021)](https://ai.facebook.com/research/publications/unsupervised-speech-recognition). The model takes as input wav2vec 2.0 or XLSR representations (see [pretrained models](https://github.com/pytorch/fairseq/blob/main/examples/wav2vec)) as well as unlabeled speech and text data. +wav2vec Unsupervised (wav2vec-U) is a framework for building speech recognition systems without any labeled training data as described in [Unsupervised Speech Recognition (Baevski et al., 2021)](https://ai.facebook.com/research/publications/unsupervised-speech-recognition). The model takes as input wav2vec 2.0 or XLSR representations (see [pretrained models](https://github.com/pytorch/fairseq/blob/main/examples/wav2vec)) as well as unlabeled speech and text data. The wav2vec-U training procedure consists of three consecutive main steps: * Preparation of speech representations and text data diff --git a/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md b/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md index d0c1281073da775bcb837bfa32ccf8408bf33c88..4cb1846075c7139339f7c618821c25bc59ea7666 100644 --- a/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md +++ b/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md @@ -193,8 +193,8 @@ MAE的设计虽然简单，但已被证明是一个强大的、可扩展的视 ``` 公共参数： --data_path // 数据集路径 - --finetune_pth // 预训练模型路径 - --resume_pth // finetuned模型路径 + --finetune_path // 预训练模型路径 + --resume_path // finetuned模型路径 ``` diff --git a/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README_raw.md b/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README_raw.md index ec7bce0faa9de139dbb482b948bdab9f1b554d36..cc0c94549889a76ef7da52a6eb42f9a9e065c364 100644 --- a/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README_raw.md +++ b/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README_raw.md @@ -111,21 +111,21 @@ By fine-tuning these pre-trained models, we rank #1 in these classification task - + - + - + diff --git a/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md b/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md index 33cb1b0b25e0b0c933e1ace161cdff65659f00d4..5cb60a608107b91b8029beffdd5a870332811615 100644 --- a/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md +++ b/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md @@ -436,7 +436,7 @@ https://gitee.com/ascend/ModelZoo-PyTorch/tree/master/ACL_PyTorch/contrib/nlp/al ``` # 以bs32为例 - python3 -m ais_bench --model outputs/albert_seq128_bs32.om --input ./preprocessed_data_seq128/input_ids,./preprocessed_data_seq128/attention_mask,./preprocessed_data_seq128/token_type_ids --output results --output_dirname seq128_bs32 --outfmt NPY --batchsize 32 + python3 -m ais_bench --model outputs/albert_seq128_bs32.om --input ./preprocessed_data_seq128/input_ids,./preprocessed_data_seq128/attention_mask,./preprocessed_data_seq128/token_type_ids --output results --output_dirname seq128_bs32 --outfmt NPY --batch_size 32 ``` - 参数说明： diff --git a/PyTorch/contrib/nlp/roberta_for_PyTorch/README_raw.md b/PyTorch/contrib/nlp/roberta_for_PyTorch/README_raw.md index 4b008af1fd363788775e9470ec59d2211c907a8d..0a73cb13cb85494ff0dc2c2d5618aab9c0f15893 100644 --- a/PyTorch/contrib/nlp/roberta_for_PyTorch/README_raw.md +++ b/PyTorch/contrib/nlp/roberta_for_PyTorch/README_raw.md @@ -40,7 +40,7 @@ We provide reference implementations of various sequence modeling papers: + [RoBERTa: A Robustly Optimized BERT Pretraining Approach (Liu et al., 2019)](examples/roberta/README.md) + [Facebook FAIR's WMT19 News Translation Task Submission (Ng et al., 2019)](examples/wmt19/README.md) + [Jointly Learning to Align and Translate with Transformer Models (Garg et al., 2019)](examples/joint_alignment_translation/README.md ) - + [Multilingual Denoising Pre-training for Neural Machine Translation (Liu et at., 2020)](examples/mbart/README.md) + + [Multilingual Denoising Pre-training for Neural Machine Translation (Liu et al., 2020)](examples/mbart/README.md) + [Neural Machine Translation with Byte-Level Subwords (Wang et al., 2020)](examples/byte_level_bpe/README.md) + [Unsupervised Quality Estimation for Neural Machine Translation (Fomicheva et al., 2020)](examples/unsupervised_quality_estimation/README.md) + [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations (Baevski et al., 2020)](examples/wav2vec/README.md) diff --git a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.custom_classification.md b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.custom_classification.md index 6f4c2cf3155b4d485d1d93b6b0ef7639daf8ba57..146e030a31cb7bee5c180cf037a6ea32c13ad8dd 100644 --- a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.custom_classification.md +++ b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.custom_classification.md @@ -1,4 +1,4 @@ -# Finetuning RoBERTa on a custom classification task +# Fine-tuning RoBERTa on a custom classification task This example shows how to finetune RoBERTa on the IMDB dataset, but should illustrate the process for most classification tasks. @@ -12,7 +12,7 @@ tar zxvf aclImdb_v1.tar.gz ### 2) Format data -`IMDB` data has one data-sample in each file, below python code-snippet converts it one file for train and valid each for ease of processing. +`IMDB` data has one data sample in each file, below python code snippet converts it one file for train and valid each for ease of processing. ```python import argparse import os diff --git a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.glue.md b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.glue.md index fcb56e9850f6ae7f4861167404141d362899688e..e53f09396bb510419e1b086585423da6773e493a 100644 --- a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.glue.md +++ b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.glue.md @@ -1,4 +1,4 @@ -# Finetuning RoBERTa on GLUE tasks +# Fine-tuning RoBERTa on GLUE tasks ### 1) Download the data from GLUE website (https://github.com/pytorch/fairseq) using following commands: ```bash @@ -27,9 +27,9 @@ There are additional config files for each of the GLUE tasks in the examples/rob **Note:** -a) Above cmd-args and hyperparams are tested on one Nvidia `V100` GPU with `32gb` of memory for each task. Depending on the GPU memory resources available to you, you can use increase `--update-freq` and reduce `--batch-size`. +a) Above cmd-args and hyperparameters are tested on one Nvidia `V100` GPU with `32gb` of memory for each task. Depending on the GPU memory resources available to you, you can use increase `--update-freq` and reduce `--batch-size`. -b) All the settings in above table are suggested settings based on our hyperparam search within a fixed search space (for careful comparison across models). You might be able to find better metrics with wider hyperparam search. +b) All the settings in above table are suggested settings based on our hyperparameter search within a fixed search space (for careful comparison across models). You might be able to find better metrics with wider hyperparameter search. ### Inference on GLUE task After training the model as mentioned in previous step, you can perform inference with checkpoints in `checkpoints/` directory using following python code snippet: diff --git a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.race.md b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.race.md index 13c917e8eca6621e91dce541c7e41436b38cbdc1..98f7b47a4465139bdacc351936ba950c1c3bef56 100644 --- a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.race.md +++ b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/README.race.md @@ -1,4 +1,4 @@ -# Finetuning RoBERTa on RACE tasks +# Fine-tuning RoBERTa on RACE tasks ### 1) Download the data from RACE website (http://www.cs.cmu.edu/~glai1/data/race/) @@ -46,9 +46,9 @@ CUDA_VISIBLE_DEVICES=0,1 fairseq-train $DATA_DIR --ddp-backend=legacy_ddp \ a) As contexts in RACE are relatively long, we are using smaller batch size per GPU while increasing update-freq to achieve larger effective batch size. -b) Above cmd-args and hyperparams are tested on one Nvidia `V100` GPU with `32gb` of memory for each task. Depending on the GPU memory resources available to you, you can use increase `--update-freq` and reduce `--batch-size`. +b) Above cmd-args and hyperparameters are tested on one Nvidia `V100` GPU with `32GB` of memory for each task. Depending on the GPU memory resources available to you, you can use increase `--update-freq` and reduce `--batch-size`. -c) The setting in above command is based on our hyperparam search within a fixed search space (for careful comparison across models). You might be able to find better metrics with wider hyperparam search. +c) The setting in above command is based on our hyperparameter search within a fixed search space (for careful comparison across models). You might be able to find better metrics with wider hyperparameter search. ### 4) Evaluation: diff --git a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/commonsense_qa/README.md b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/commonsense_qa/README.md index 05c6f841a8966d2b74a8d3fe73bca22694fe9a8a..e13780228e5ac72b976f28976a95225b08337fea 100644 --- a/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/commonsense_qa/README.md +++ b/PyTorch/contrib/nlp/roberta_for_PyTorch/examples/roberta/commonsense_qa/README.md @@ -1,6 +1,6 @@ -# Finetuning RoBERTa on Commonsense QA +# Fine-tuning RoBERTa on Commonsense QA -We follow a similar approach to [finetuning RACE](../README.race.md). Specifically +We follow a similar approach to [fine-tuning RACE](../README.race.md). Specifically for each question we construct five inputs, one for each of the five candidate answer choices. Each input is constructed by concatenating the question and candidate answer. We then encode each input and pass the resulting "[CLS]" @@ -23,7 +23,7 @@ development set after 100 trials. bash examples/roberta/commonsense_qa/download_cqa_data.sh ``` -### 2) Finetune +### 2) Fine-tune ```bash MAX_UPDATES=3000 # Number of training steps.
following are transfer learning by fine-tuning the pre-trained MAE on the target dataset:
iNaturalists 2017
iNaturalist 2017 70.5 75.7 79.3 83.4 75.4
iNaturalists 2018
iNaturalist 2018 75.4 80.1 83.0 86.8 81.2
iNaturalists 2019
iNaturalist 2019 80.5 83.4 85.7