Training fails with tensor2tensor current version (1.9.0)

[Hugo-W]

The documentation mentioned

Tensor2Tensor version 1.6.6 or higher

Though, with tensor2tensor version 1.9.0, the following error occured when trying to train a new model:

Traceback (most recent call last):
  File "/home/hw2512/anaconda/envs/tensorflow/bin/g2p-seq2seq", line 11, in <module>
    load_entry_point('g2p-seq2seq==6.2.2a0', 'console_scripts', 'g2p-seq2seq')()
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/g2p_seq2seq-6.2.2a0-py2.7.egg/g2p_seq2seq/app.py", line 113, in main
    g2p_model.train()
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/g2p_seq2seq-6.2.2a0-py2.7.egg/g2p_seq2seq/g2p.py", line 259, in train
    execute_schedule(self.exp, self.params)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/g2p_seq2seq-6.2.2a0-py2.7.egg/g2p_seq2seq/g2p.py", line 589, in execute_schedule
    getattr(exp, params.schedule)()
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensor2tensor/utils/trainer_lib.py", line 331, in continuous_train_and_eval
    self._eval_spec)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/estimator/training.py", line 471, in train_and_evaluate
    return executor.run()
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/estimator/training.py", line 610, in run
    return self.run_local()
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/estimator/training.py", line 711, in run_local
    saving_listeners=saving_listeners)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/estimator/estimator.py", line 356, in train
    loss = self._train_model(input_fn, hooks, saving_listeners)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/estimator/estimator.py", line 1181, in _train_model
    return self._train_model_default(input_fn, hooks, saving_listeners)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/estimator/estimator.py", line 1211, in _train_model_default
    features, labels, model_fn_lib.ModeKeys.TRAIN, self.config)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/estimator/estimator.py", line 1169, in _call_model_fn
    model_fn_results = self._model_fn(features=features, **kwargs)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensor2tensor/utils/t2t_model.py", line 1225, in wrapping_model_fn
    decode_hparams=decode_hparams)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensor2tensor/utils/t2t_model.py", line 1322, in estimator_model_fn
    loss, num_async_replicas=num_async_replicas, use_tpu=use_tpu)
  File "/home/hw2512/anaconda/envs/tensorflow/lib/python2.7/site-packages/tensor2tensor/utils/t2t_model.py", line 1349, in estimator_spec_train
    if self._hparams.warm_start_from:
AttributeError: 'HParams' object has no attribute 'warm_start_from'

The interactive and classic decoding was working without error though.

The training error disappeared when I downgraded tensor2tensor to version 1.6.6.

[huangruizhe]

Got the same problem, and fixed in the same way. Thanks!