NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
https://nvidia.github.io/OpenSeq2Seq
Apache License 2.0
1.54k stars 372 forks source link

Interactive Inference Example problem in loading model #521

Open NatBrian opened 4 years ago

NatBrian commented 4 years ago

I downloaded the checkpoints for SpeechToText https://drive.google.com/file/d/12CQvNrTvf0cjTsKjbaWWvdaZb7RxWI6X/view and TextToSpeech https://drive.google.com/file/d/1Ddf7nDI2PpgaxvZMm7bd8N_Evk_ExTwg/view

I extract everything and put each checkpoint folder inside infer_S2T and infer_T2S so, inside infer_S2T folder there are

similarly with infer_T2S the code is not abe to load the model, how should I fix the checkpoint?

WARNING:tensorflow:From D:\GitHub\OpenSeq2Seq\open_seq2seq\parts\transformer\attention_layer.py:24: The name tf.layers.Layer is deprecated. Please use tf.compat.v1.layers.Layer instead.

WARNING:tensorflow:From D:\GitHub\OpenSeq2Seq\open_seq2seq\parts\cnns\tcn.py:8: The name tf.layers.Conv1D is deprecated. Please use tf.compat.v1.layers.Conv1D instead.

*** Restoring from the latest checkpoint WARNING:tensorflow:UnknownError: NewRandomAccessFile failed to Create/Open: Infer_S2T/checkpoint : Access is denied. ; Input/output error WARNING:tensorflow:Infer_S2T/checkpoint: Checkpoint ignored

OSError Traceback (most recent call last)

in 30 return model, checkpoint 31 ---> 32 model_S2T, checkpoint_S2T = get_model(args_S2T, "S2T") 33 model_T2S, checkpoint_T2S = get_model(args_T2S, "T2S") 34 in get_model(args, scope) 26 with tf.variable_scope(scope): 27 args, base_config, base_model, config_module = get_base_config(args) ---> 28 checkpoint = check_logdir(args, base_config) 29 model = create_model(args, base_config, config_module, base_model, None) 30 return model, checkpoint D:\GitHub\OpenSeq2Seq\open_seq2seq\utils\utils.py in check_logdir(args, base_config, restore_best_checkpoint) 692 raise IOError( 693 "There is no valid TensorFlow checkpoint in the " --> 694 "{} directory. Can't load model".format(ckpt_dir) 695 ) 696 else: OSError: There is no valid TensorFlow checkpoint in the Infer_S2T/ directory. Can't load model