From c3e2ca4b73d1d798cd7ec25263e06c357d9d9a2f Mon Sep 17 00:00:00 2001 From: zhifu gao <zhifu.gzf@alibaba-inc.com> Date: Wed, 8 Nov 2023 18:46:29 +0800 Subject: [PATCH 1/5] Update README.md --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index e51d8bdeb..3f6b4349e 100644 --- a/README.md +++ b/README.md @@ -17,7 +17,7 @@ | [**Quick Start**](#quick-start) | [**Runtime**](./runtime/readme.md) | [**Model Zoo**](#model-zoo) -| [**Contact**](#community-communication) +| [**Contact**](#contact) <a name="highlights"></a> @@ -126,7 +126,7 @@ FunASR supports deploying pre-trained or further fine-tuned models for service. For more detailed information, please refer to the [service deployment documentation](runtime/readme.md). -<a name="Community Communication"></a> +<a name="contact"></a> ## Community Communication If you encounter problems in use, you can directly raise Issues on the github page. From 71f95f256bfa10a6b67539700e17f1eda45414dd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=9B=BE=E8=81=AA?= <wucong.lyb@alibaba-inc.com> Date: Thu, 9 Nov 2023 10:56:22 +0800 Subject: [PATCH 2/5] update funasr-onnx-2pass.cpp --- runtime/onnxruntime/bin/funasr-onnx-2pass-rtf.cpp | 2 +- runtime/onnxruntime/bin/funasr-onnx-2pass.cpp | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/runtime/onnxruntime/bin/funasr-onnx-2pass-rtf.cpp b/runtime/onnxruntime/bin/funasr-onnx-2pass-rtf.cpp index 5bf8f18ae..c55c888d4 100644 --- a/runtime/onnxruntime/bin/funasr-onnx-2pass-rtf.cpp +++ b/runtime/onnxruntime/bin/funasr-onnx-2pass-rtf.cpp @@ -279,7 +279,7 @@ int main(int argc, char** argv) // hotword file unordered_map<string, int> hws_map; std::string nn_hotwords_ = ""; - std::string hotword_path = model_path.at(HOTWORD); + std::string hotword_path = hotword.getValue(); LOG(INFO) << "hotword path: " << hotword_path; funasr::ExtractHws(hotword_path, hws_map, nn_hotwords_); diff --git a/runtime/onnxruntime/bin/funasr-onnx-2pass.cpp b/runtime/onnxruntime/bin/funasr-onnx-2pass.cpp index d0154999d..5af0b414a 100644 --- a/runtime/onnxruntime/bin/funasr-onnx-2pass.cpp +++ b/runtime/onnxruntime/bin/funasr-onnx-2pass.cpp @@ -113,7 +113,7 @@ int main(int argc, char** argv) // hotword file unordered_map<string, int> hws_map; std::string nn_hotwords_ = ""; - std::string hotword_path = model_path.at(HOTWORD); + std::string hotword_path = hotword.getValue(); LOG(INFO) << "hotword path: " << hotword_path; funasr::ExtractHws(hotword_path, hws_map, nn_hotwords_); From 9954fae3875c562df348789dc837ffd49335d9e8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=9B=BE=E8=81=AA?= <wucong.lyb@alibaba-inc.com> Date: Thu, 9 Nov 2023 11:04:21 +0800 Subject: [PATCH 3/5] tokenizer.cpp --- runtime/onnxruntime/src/tokenizer.cpp | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/runtime/onnxruntime/src/tokenizer.cpp b/runtime/onnxruntime/src/tokenizer.cpp index a111b9157..f56601a95 100644 --- a/runtime/onnxruntime/src/tokenizer.cpp +++ b/runtime/onnxruntime/src/tokenizer.cpp @@ -17,8 +17,12 @@ CTokenizer::CTokenizer():m_ready(false) CTokenizer::~CTokenizer() { - delete jieba_dict_trie_; - delete jieba_model_; + if (jieba_dict_trie_){ + delete jieba_dict_trie_; + } + if (jieba_model_){ + delete jieba_model_; + } } void CTokenizer::SetJiebaRes(cppjieba::DictTrie *dict, cppjieba::HMMModel *hmm) { From b0c108279275a5c0011dfcf76b3e725b3263209b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=9B=BE=E8=81=AA?= <wucong.lyb@alibaba-inc.com> Date: Thu, 9 Nov 2023 11:05:51 +0800 Subject: [PATCH 4/5] update tokenizer.h --- runtime/onnxruntime/src/tokenizer.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/runtime/onnxruntime/src/tokenizer.h b/runtime/onnxruntime/src/tokenizer.h index 149161ba3..166061bd4 100644 --- a/runtime/onnxruntime/src/tokenizer.h +++ b/runtime/onnxruntime/src/tokenizer.h @@ -17,8 +17,8 @@ class CTokenizer { vector<string> m_id2token,m_id2punc; map<string, int> m_token2id,m_punc2id; - cppjieba::DictTrie *jieba_dict_trie_; - cppjieba::HMMModel *jieba_model_; + cppjieba::DictTrie *jieba_dict_trie_=nullptr; + cppjieba::HMMModel *jieba_model_=nullptr; cppjieba::Jieba jieba_processor_; public: From d90de51e76099a308e731e3575f76a71cc684f95 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=8C=97=E5=BF=B5?= <lzr265946@alibaba-inc.com> Date: Thu, 9 Nov 2023 11:06:07 +0800 Subject: [PATCH 5/5] fix paraformer finetune bug --- funasr/bin/build_trainer.py | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/funasr/bin/build_trainer.py b/funasr/bin/build_trainer.py index 61af7663d..bda83ec87 100644 --- a/funasr/bin/build_trainer.py +++ b/funasr/bin/build_trainer.py @@ -548,7 +548,10 @@ def build_trainer(modelscope_dict, init_param = modelscope_dict['init_model'] cmvn_file = modelscope_dict['cmvn_file'] seg_dict_file = modelscope_dict['seg_dict'] - bpemodel = modelscope_dict['bpemodel'] + if 'bpemodel' in modelscope_dict: + bpemodel = modelscope_dict['bpemodel'] + else: + bpemodel = None # overwrite parameters with open(config) as f: @@ -582,7 +585,7 @@ def build_trainer(modelscope_dict, args.seg_dict_file = seg_dict_file else: args.seg_dict_file = None - if os.path.exists(bpemodel): + if bpemodel is not None and os.path.exists(bpemodel): args.bpemodel = bpemodel else: args.bpemodel = None