mv the intermediate files to the assets directory

wenet-e2e · Oct 23, 2023 · e8fa013 · e8fa013
1 parent 852165c
commit e8fa013
Show file tree

Hide file tree

Showing 5 changed files with 2 additions and 17 deletions.
diff --git a/wenet/paraformer/ali_paraformer/config.yaml → ...aformer/ali_paraformer/assets/config.yaml b/wenet/paraformer/ali_paraformer/config.yaml → ...aformer/ali_paraformer/assets/config.yaml
@@ -14,6 +14,8 @@ encoder_conf:
     kernel_size: 11
     sanm_shfit: 0
 
+input_dim: 80
+output_dim: 8404
 paraformer: true
 # decoder related
 decoder: SanmDecoder

diff --git a/wenet/paraformer/ali_paraformer/global_cmvn → ...aformer/ali_paraformer/assets/global_cmvn b/wenet/paraformer/ali_paraformer/global_cmvn → ...aformer/ali_paraformer/assets/global_cmvn
diff --git a/wenet/paraformer/ali_paraformer/units.txt → ...araformer/ali_paraformer/assets/units.txt b/wenet/paraformer/ali_paraformer/units.txt → ...araformer/ali_paraformer/assets/units.txt
diff --git a/...t/paraformer/ali_paraformer/export_jit.py → ...raformer/ali_paraformer/test_infer_jit.py b/...t/paraformer/ali_paraformer/export_jit.py → ...raformer/ali_paraformer/test_infer_jit.py
@@ -43,24 +43,8 @@ def main():
     char_dict = {v: k for k, v in symbol_table.items()}
     with open(args.config, 'r') as fin:
         configs = yaml.load(fin, Loader=yaml.FullLoader)
-
-    # mean, istd = load_cmvn(args.cmvn, is_json=True)
-    # global_cmvn = GlobalCMVN(
-    #     torch.from_numpy(mean).float(),
-    #     torch.from_numpy(istd).float())
-    # configs['encoder_conf']['input_size'] = 80 * 7
-    # encoder = SanmEncoder(global_cmvn=global_cmvn, **configs['encoder_conf'])
-    # configs['decoder_conf']['vocab_size'] = len(char_dict)
-    # configs['decoder_conf']['encoder_output_size'] = encoder.output_size()
-    # decoder = SanmDecoer(**configs['decoder_conf'])
-
-    # # predictor = PredictorV3(**configs['predictor_conf'])
-    # predictor = Predictor(**configs['predictor_conf'])
-    # model = AliParaformer(encoder, decoder, predictor)
     configs['cmvn_file'] = args.cmvn
     configs['is_json_cmvn'] = True
-    configs['input_dim'] = 80
-    configs['output_dim'] = len(char_dict)
     model = init_model(configs)
     load_checkpoint(model, args.ali_paraformer)
     model.eval()

diff --git a/wenet/utils/init_model.py b/wenet/utils/init_model.py
@@ -131,7 +131,6 @@ def init_model(configs):
         if isinstance(encoder, SanmEncoder):
             assert isinstance(decoder, SanmDecoer)
             # NOTE(Mddct): only support inference for now
-            print('hello world')
             model = AliParaformer(encoder, decoder, predictor)
         else:
             model = Paraformer(vocab_size=vocab_size,