search

mingzhu0527 / MASHQA

Apache License 2.0
26 stars 8 forks source link

Error while training #2

Open Ibtastic opened 3 years ago

Ibtastic commented 3 years ago

When doing sh sentqa_GPU.sh , I get this error OSError: Not found: "xlnet_cased_L-24_H-1024_A-16/spiece.model": No such file or directory Error #2`. Can you tell how can I solve this?

mingzhu0527 commented 3 years ago

You can download it from https://storage.googleapis.com/xlnet/released_models/cased_L-24_H-1024_A-16.zip

Ibtastic commented 3 years ago

What I am supposed to put in model_dir=experiment/squad_consec directory , which cannot be empty? Also, to preprocess the data the shell script runs the run_sentqa_att.py , there is no such file in your repo.

mingzhu0527 commented 3 years ago

Just added run_sentqa_att.py file. Sorry I didn't upload it before

mingzhu0527 commented 3 years ago

Can you provide more information about the model_dir related question? What kind of error message did you get?

Ibtastic commented 3 years ago

The tf.contrib.tpu.RunConfig function in model_utils.py has model as a parameter and throws this error ValueError: model_dir should be non-empty.

You pass model_dir as an argument for sentqa_GPU_att.py

mingzhu0527 commented 3 years ago

To train the model, you should use: sh scripts/sentqa_GPU.sh

Ibtastic commented 3 years ago

Yes, did that. In the python script run_sentqa_att_sparse_hier_weight_10_self_att.py at line 715 estimator.train(input_fn=train_input_fn, max_steps=FLAGS.train_steps) , I get this error ValueError: Tensor conversion requested dtype string for Tensor with dtype float32: <tf.Tensor 'args_0:0' shape=() dtype=float32>

lizc1992 commented 2 years ago

I get also this error-

ValueError: Tensor conversion requested dtype string for Tensor with dtype float32: <tf.Tensor 'args_0:0' shape=() dtype=float32>

@Ibtastic did you success to solve this?