From a0eaae31b5753ff4274a298e52f3ef7b8f48a5b6 Mon Sep 17 00:00:00 2001 From: speech_asr <wangjiaming.wjm@alibaba-inc.com> Date: 星期三, 15 二月 2023 16:12:25 +0800 Subject: [PATCH] update docs --- docs_cn/modelscope_usages.md | 10 ++-- docs/modelscope_usages.md | 65 ++++++++++++++++---------------- 2 files changed, 38 insertions(+), 37 deletions(-) diff --git a/docs/modelscope_usages.md b/docs/modelscope_usages.md index d42416a..5a4dc76 100644 --- a/docs/modelscope_usages.md +++ b/docs/modelscope_usages.md @@ -8,44 +8,45 @@ - `infer_after_finetune.py`: perform inference on the specified dataset based on the finetuned model ## Inference -We provide `infer.py` to achieve the inference. Based on this file, users can preform inference on the specified dataset based on our provided model and obtain the corresponding recognition results. If the transcript is given, the `CER` will be calculated at the same time. Before performing inference, users can specify the following parameters to modify the inference configuration: -* `data_dir`锛氭暟鎹泦鐩綍銆傜洰褰曚笅搴旇鍖呮嫭闊抽鍒楄〃鏂囦欢`wav.scp`鍜屾妱鏈枃浠禶text`(鍙��)锛屽叿浣撴牸寮忓彲浠ュ弬瑙乕蹇�熷紑濮媇(./get_started.md)涓殑璇存槑銆傚鏋渀text`鏂囦欢瀛樺湪锛屽垯浼氱浉搴旂殑璁＄畻CER锛屽惁鍒欎細璺宠繃銆� -* `output_dir`锛氭帹鐞嗙粨鏋滀繚瀛樼洰褰� -* `batch_size`锛氭帹鐞嗘椂鐨刡atch澶у皬 -* `ctc_weight`锛氶儴鍒嗘ā鍨嬪寘鍚獵TC妯″潡锛屽彲浠ヨ缃鍙傛暟鏉ユ寚瀹氭帹鐞嗘椂锛孋TC妯″潡鐨勬潈閲� +We provide `infer.py` to achieve the inference. Based on this file, users can preform inference on the specified dataset based on our provided model and obtain the corresponding recognition results. If the transcript is given, the `CER` will be calculated at the same time. Before performing inference, users can set the following parameters to modify the inference configuration: +* `data_dir`锛歞ataset directory. The directory should contain the wav list file `wav.scp` and the transcript file `text` (optional). For the format of these two files, please refer to the instructions in [Quick Start](./get_started.md). If the `text` file exists, the CER will be calculated accordingly, otherwise it will be skipped. +* `output_dir`锛歵he directory for saving the inference results +* `batch_size`锛歜atch size during the inference +* `ctc_weight`锛歴ome models contain a CTC module, users can set this parameter to specify the weight of the CTC module during the inference -闄や簡鐩存帴鍦╜infer.py`涓缃弬鏁板锛岀敤鎴蜂篃鍙互閫氳繃鎵嬪姩淇敼妯″瀷涓嬭浇鐩綍涓嬬殑`decoding.yaml`鏂囦欢涓殑鍙傛暟鏉ヤ慨鏀规帹鐞嗛厤缃�� +In addition to directly setting parameters in `infer.py`, users can also manually set the parameters in the `decoding.yaml` file in the model download directory to modify the inference configuration. -## 妯″瀷寰皟 -鎴戜滑鎻愪緵浜哷finetune.py`鏉ュ疄鐜版ā鍨嬪井璋冦�傚熀浜庢鏂囦欢锛岀敤鎴峰彲浠ュ熀浜庢垜浠彁渚涚殑妯″瀷浣滀负鍒濆妯″瀷锛屽湪鎸囧畾鐨勬暟鎹泦涓婅繘琛屽井璋冿紝浠庤�屽湪鐗瑰緛棰嗗煙鍙栧緱鏇村ソ鐨勬�ц兘銆傚湪寰皟寮�濮嬪墠锛岀敤鎴峰彲浠ユ寚瀹氬涓嬪弬鏁版潵淇敼寰皟閰嶇疆锛� -* `data_path`锛氭暟鎹洰褰曘�傝鐩綍涓嬪簲璇ュ寘鎷瓨鏀捐缁冮泦鏁版嵁鐨刞train`鐩綍鍜屽瓨鏀鹃獙璇侀泦鏁版嵁鐨刞dev`鐩綍銆傛瘡涓洰褰曚腑闇�瑕佸寘鎷煶棰戝垪琛ㄦ枃浠禶wav.scp`鍜屾妱鏈枃浠禶text` -* `output_dir`锛氬井璋冪粨鏋滀繚瀛樼洰褰� -* `dataset_type`锛氬浜庡皬鏁版嵁闆嗭紝璁剧疆涓篳small`锛涘綋鏁版嵁閲忓ぇ浜�1000灏忔椂鏃讹紝璁剧疆涓篳large` -* `batch_bins`锛歜atch size锛屽鏋渄ataset_type璁剧疆涓篳small`锛宐atch_bins鍗曚綅涓篺bank鐗瑰緛甯ф暟锛涘鏋渄ataset_type=`large`锛宐atch_bins鍗曚綅涓烘绉� -* `max_epoch`锛氭渶澶х殑璁粌杞暟 +## Finetuning +We provide `finetune.py` to achieve the finetuning. Based on this file, users can finetune on the specified dataset based on our provided model as the initial model to achieve better performance in the specificed domain. Before finetuning, users can set the following parameters to modify the finetuning configuration: +* `data_path`锛歞ataset directory銆俆his directory should contain the `train` directory for saving the training set and the `dev` directory for saving the validation set. Each directory needs to contain the wav list file `wav.scp` and the transcript file `text` +* `output_dir`锛歵he directory for saving the finetuning results +* `dataset_type`锛歠or small dataset锛宻et as `small`锛沠or dataset larger than 1000 hours锛宻et as `large` +* `batch_bins`锛歜atch size锛宨f dataset_type is set as `small`锛宼he unit of batch_bins is the number of fbank feature frames; if dataset_type is set as `large`, the unit of batch_bins is milliseconds +* `max_epoch`锛歵he maximum number of training epochs -浠ヤ笅鍙傛暟涔熷彲浠ヨ繘琛岃缃�備絾鏄鏋滄病鏈夌壒鍒殑闇�姹傦紝鍙互蹇界暐锛岀洿鎺ヤ娇鐢ㄦ垜浠粰瀹氱殑榛樿鍊硷細 -* `accum_grad`锛氭搴︾疮绉� -* `keep_nbest_models`锛氶�夋嫨鎬ц兘鏈�濂界殑`keep_nbest_models`涓ā鍨嬬殑鍙傛暟杩涜骞冲潎锛屽緱鍒版�ц兘鏇村ソ鐨勬ā鍨� -* `optim`锛氳缃井璋冩椂鐨勪紭鍖栧櫒 -* `lr`锛氳缃井璋冩椂鐨勫涔犵巼 -* `scheduler`锛氳缃涔犵巼璋冩暣绛栫暐 -* `scheduler_conf`锛氬涔犵巼璋冩暣绛栫暐鐨勭浉鍏冲弬鏁� -* `specaug`锛氳缃氨澧炲箍 -* `specaug_conf`锛氳氨澧炲箍鐨勭浉鍏冲弬鏁� +The following parameters can also be set. However, if there is no special requirement, users can ignore these parameters and use the default value we provided directly: +* `accum_grad`锛歵he accumulation of the gradient +* `keep_nbest_models`锛歴elect the `keep_nbest_models` models with the best performance and average the parameters + of these models to get a better model +* `optim`锛歴et the optimizer +* `lr`锛歴et the learning rate +* `scheduler`锛歴et learning rate adjustment strategy +* `scheduler_conf`锛歴et the related parameters of the learning rate adjustment strategy +* `specaug`锛歴et for the spectral augmentation +* `specaug_conf`锛歴et related parameters of the spectral augmentation -闄や簡鐩存帴鍦╜finetune.py`涓缃弬鏁板锛岀敤鎴蜂篃鍙互閫氳繃鎵嬪姩淇敼妯″瀷涓嬭浇鐩綍涓嬬殑`finetune.yaml`鏂囦欢涓殑鍙傛暟鏉ヤ慨鏀瑰井璋冮厤缃�� +In addition to directly setting parameters in `finetune.py`, users can also manually set the parameters in the `finetune.yaml` file in the model download directory to modify the finetuning configuration. -## 鍩轰簬寰皟鍚庣殑妯″瀷鎺ㄧ悊 -鎴戜滑鎻愪緵浜哷infer_after_finetune.py`鏉ュ疄鐜板熀浜庣敤鎴疯嚜宸卞井璋冨緱鍒扮殑妯″瀷杩涜鎺ㄧ悊銆傚熀浜庢鏂囦欢锛岀敤鎴峰彲浠ュ熀浜庡井璋冨悗鐨勬ā鍨嬶紝瀵规寚瀹氱殑鏁版嵁闆嗚繘琛屾帹鐞嗭紝寰楀埌鐩稿簲鐨勮瘑鍒粨鏋溿�傚鏋滃悓鏃剁粰瀹氫簡鎶勬湰锛屽垯浼氬悓鏃惰绠桟ER銆傚湪寮�濮嬫帹鐞嗗墠锛岀敤鎴峰彲浠ユ寚瀹氬涓嬪弬鏁版潵淇敼鎺ㄧ悊閰嶇疆锛� -* `data_dir`锛氭暟鎹泦鐩綍銆傜洰褰曚笅搴旇鍖呮嫭闊抽鍒楄〃鏂囦欢`wav.scp`鍜屾妱鏈枃浠禶text`(鍙��)銆傚鏋渀text`鏂囦欢瀛樺湪锛屽垯浼氱浉搴旂殑璁＄畻CER锛屽惁鍒欎細璺宠繃銆� -* `output_dir`锛氭帹鐞嗙粨鏋滀繚瀛樼洰褰� -* `batch_size`锛氭帹鐞嗘椂鐨刡atch澶у皬 -* `ctc_weight`锛氶儴鍒嗘ā鍨嬪寘鍚獵TC妯″潡锛屽彲浠ヨ缃鍙傛暟鏉ユ寚瀹氭帹鐞嗘椂锛孋TC妯″潡鐨勬潈閲� -* `decoding_model_name`锛氭寚瀹氱敤浜庢帹鐞嗙殑妯″瀷鍚� +## Inference after Finetuning +We provide `infer_after_finetune.py` to achieve the inference based on the model finetuned by users. Based on this file, users can preform inference on the specified dataset based on the finetuned model and obtain the corresponding recognition results. If the transcript is given, the `CER` will be calculated at the same time. Before performing inference, users can set the following parameters to modify the inference configuration: +* `data_dir`锛歞ataset directory銆俆he directory should contain the wav list file `wav.scp` and the transcript file `text` (optional). If the `text` file exists, the CER will be calculated accordingly, otherwise it will be skipped. +* `output_dir`锛歵he directory for saving the inference results +* `batch_size`锛歜atch size during the inference +* `ctc_weight`锛歴ome models contain a CTC module, users can set this parameter to specify the weight of the CTC module during the inference +* `decoding_model_name`锛歴et the name of the model used for the inference -浠ヤ笅鍙傛暟涔熷彲浠ヨ繘琛岃缃�備絾鏄鏋滄病鏈夌壒鍒殑闇�姹傦紝鍙互蹇界暐锛岀洿鎺ヤ娇鐢ㄦ垜浠粰瀹氱殑榛樿鍊硷細 -* `modelscope_model_name`锛氬井璋冩椂浣跨敤鐨勫垵濮嬫ā鍨� +The following parameters can also be set. However, if there is no special requirement, users can ignore these parameters and use the default value we provided directly: +* `modelscope_model_name`锛歵he initial model name used when finetuning * `required_files`锛氫娇鐢╩odelscope鎺ュ彛杩涜鎺ㄧ悊鏃堕渶瑕佺敤鍒扮殑鏂囦欢 ## 娉ㄦ剰浜嬮」 diff --git a/docs_cn/modelscope_usages.md b/docs_cn/modelscope_usages.md index 43d0c0f..3d9080a 100644 --- a/docs_cn/modelscope_usages.md +++ b/docs_cn/modelscope_usages.md @@ -21,14 +21,14 @@ * `data_path`锛氭暟鎹洰褰曘�傝鐩綍涓嬪簲璇ュ寘鎷瓨鏀捐缁冮泦鏁版嵁鐨刞train`鐩綍鍜屽瓨鏀鹃獙璇侀泦鏁版嵁鐨刞dev`鐩綍銆傛瘡涓洰褰曚腑闇�瑕佸寘鎷煶棰戝垪琛ㄦ枃浠禶wav.scp`鍜屾妱鏈枃浠禶text` * `output_dir`锛氬井璋冪粨鏋滀繚瀛樼洰褰� * `dataset_type`锛氬浜庡皬鏁版嵁闆嗭紝璁剧疆涓篳small`锛涘綋鏁版嵁閲忓ぇ浜�1000灏忔椂鏃讹紝璁剧疆涓篳large` -* `batch_bins`锛歜atch size锛屽鏋渄ataset_type璁剧疆涓篳small`锛宐atch_bins鍗曚綅涓篺bank鐗瑰緛甯ф暟锛涘鏋渄ataset_type=`large`锛宐atch_bins鍗曚綅涓烘绉� +* `batch_bins`锛歜atch size锛屽鏋渄ataset_type璁剧疆涓篳small`锛宐atch_bins鍗曚綅涓篺bank鐗瑰緛甯ф暟锛涘鏋渄ataset_type璁剧疆涓篳large`锛宐atch_bins鍗曚綅涓烘绉� * `max_epoch`锛氭渶澶х殑璁粌杞暟 浠ヤ笅鍙傛暟涔熷彲浠ヨ繘琛岃缃�備絾鏄鏋滄病鏈夌壒鍒殑闇�姹傦紝鍙互蹇界暐锛岀洿鎺ヤ娇鐢ㄦ垜浠粰瀹氱殑榛樿鍊硷細 * `accum_grad`锛氭搴︾疮绉� * `keep_nbest_models`锛氶�夋嫨鎬ц兘鏈�濂界殑`keep_nbest_models`涓ā鍨嬬殑鍙傛暟杩涜骞冲潎锛屽緱鍒版�ц兘鏇村ソ鐨勬ā鍨� -* `optim`锛氳缃井璋冩椂鐨勪紭鍖栧櫒 -* `lr`锛氳缃井璋冩椂鐨勫涔犵巼 +* `optim`锛氳缃紭鍖栧櫒 +* `lr`锛氳缃涔犵巼 * `scheduler`锛氳缃涔犵巼璋冩暣绛栫暐 * `scheduler_conf`锛氬涔犵巼璋冩暣绛栫暐鐨勭浉鍏冲弬鏁� * `specaug`锛氳缃氨澧炲箍 @@ -37,7 +37,7 @@ 闄や簡鐩存帴鍦╜finetune.py`涓缃弬鏁板锛岀敤鎴蜂篃鍙互閫氳繃鎵嬪姩淇敼妯″瀷涓嬭浇鐩綍涓嬬殑`finetune.yaml`鏂囦欢涓殑鍙傛暟鏉ヤ慨鏀瑰井璋冮厤缃�� ## 鍩轰簬寰皟鍚庣殑妯″瀷鎺ㄧ悊 -鎴戜滑鎻愪緵浜哷infer_after_finetune.py`鏉ュ疄鐜板熀浜庣敤鎴疯嚜宸卞井璋冨緱鍒扮殑妯″瀷杩涜鎺ㄧ悊銆傚熀浜庢鏂囦欢锛岀敤鎴峰彲浠ュ熀浜庡井璋冨悗鐨勬ā鍨嬶紝瀵规寚瀹氱殑鏁版嵁闆嗚繘琛屾帹鐞嗭紝寰楀埌鐩稿簲鐨勮瘑鍒粨鏋溿�傚鏋滃悓鏃剁粰瀹氫簡鎶勬湰锛屽垯浼氬悓鏃惰绠桟ER銆傚湪寮�濮嬫帹鐞嗗墠锛岀敤鎴峰彲浠ユ寚瀹氬涓嬪弬鏁版潵淇敼鎺ㄧ悊閰嶇疆锛� +鎴戜滑鎻愪緵浜哷infer_after_finetune.py`鏉ュ疄鐜板熀浜庣敤鎴疯嚜宸卞井璋冨緱鍒扮殑妯″瀷杩涜鎺ㄧ悊銆傚熀浜庢鏂囦欢锛岀敤鎴峰彲浠ュ熀浜庡井璋冨悗鐨勬ā鍨嬶紝瀵规寚瀹氱殑鏁版嵁闆嗚繘琛屾帹鐞嗭紝寰楀埌鐩稿簲鐨勮瘑鍒粨鏋溿�傚鏋滅粰瀹氫簡鎶勬湰锛屽垯浼氬悓鏃惰绠桟ER銆傚湪寮�濮嬫帹鐞嗗墠锛岀敤鎴峰彲浠ユ寚瀹氬涓嬪弬鏁版潵淇敼鎺ㄧ悊閰嶇疆锛� * `data_dir`锛氭暟鎹泦鐩綍銆傜洰褰曚笅搴旇鍖呮嫭闊抽鍒楄〃鏂囦欢`wav.scp`鍜屾妱鏈枃浠禶text`(鍙��)銆傚鏋渀text`鏂囦欢瀛樺湪锛屽垯浼氱浉搴旂殑璁＄畻CER锛屽惁鍒欎細璺宠繃銆� * `output_dir`锛氭帹鐞嗙粨鏋滀繚瀛樼洰褰� * `batch_size`锛氭帹鐞嗘椂鐨刡atch澶у皬 @@ -45,7 +45,7 @@ * `decoding_model_name`锛氭寚瀹氱敤浜庢帹鐞嗙殑妯″瀷鍚� 浠ヤ笅鍙傛暟涔熷彲浠ヨ繘琛岃缃�備絾鏄鏋滄病鏈夌壒鍒殑闇�姹傦紝鍙互蹇界暐锛岀洿鎺ヤ娇鐢ㄦ垜浠粰瀹氱殑榛樿鍊硷細 -* `modelscope_model_name`锛氬井璋冩椂浣跨敤鐨勫垵濮嬫ā鍨� +* `modelscope_model_name`锛氬井璋冩椂浣跨敤鐨勫垵濮嬫ā鍨嬪悕 * `required_files`锛氫娇鐢╩odelscope鎺ュ彛杩涜鎺ㄧ悊鏃堕渶瑕佺敤鍒扮殑鏂囦欢 ## 娉ㄦ剰浜嬮」 -- Gitblit v1.9.1