test infer

app.py

@@ -53,11 +53,12 @@ TASK_PROMPT_MAPPING = {
     "STTC (Speech to Text Chat)": "首先将语音转录为文字，然后对语音内容进行回复，转录和文字之间使用<开始回答>分割。"
 }
 
+@spaces.GPU
 def init_model_my():
     logging.basicConfig(level=logging.DEBUG,
                         format='%(asctime)s %(levelname)s %(message)s')
-    config_path = "/home/node54_tmpdata/xlgeng/code/wenet_undersdand_and_speech_xlgeng/examples/wenetspeech/whisper/exp/update_data/epoch_1_with_token/epoch_11.yaml"
-    checkpoint_path = "/home/work_nfs15/asr_data/ckpt/understanding_model/epoch_13_with_asr-chat_full_data/step_32499/step_32499.pt"
+    config_path = "train.yaml"
+    checkpoint_path = "step_32499.pt"
     args = SimpleNamespace(**{
         "checkpoint": checkpoint_path,
     })
@@ -68,7 +69,7 @@ def init_model_my():
     print(model)
     return model, tokenizer
 
-# model, tokenizer = init_model_my()
+model, tokenizer = init_model_my()
 print("model init success")
 def do_resample(input_wav_path, output_wav_path):
     """"""
@@ -120,13 +121,13 @@ def true_decode_fuc(input_wav_path, input_prompt):
     model = None
     res_text = model.generate(wavs=feat, wavs_len=feat_lens, prompt=input_prompt)[0]
     print("耿雪龙哈哈：", res_text)
-    return res_text, now_file_tmp_path_resample
+    return res_text
 @spaces.GPU
 def do_decode(input_wav_path, input_prompt):
     print(f'input_wav_path= {input_wav_path}, input_prompt= {input_prompt}')
     # 省略处理逻辑
-    # output_res, now_file_tmp_path_resample= true_decode_fuc(input_wav_path, input_prompt)
-    output_res = f"耿雪龙哈哈：测试结果, input_wav_path= {input_wav_path}, input_prompt= {input_prompt}"
+    output_res= true_decode_fuc(input_wav_path, input_prompt)
+    # output_res = f"耿雪龙哈哈：测试结果, input_wav_path= {input_wav_path}, input_prompt= {input_prompt}"
     return output_res
 
 def save_to_jsonl(if_correct, wav, prompt, res):