ValueError when setting finetune to True

[Arminius4]

Hi there,

when running

python -m basic.cli --mode train --noload --debug --finetune True

I get

'word_vocab_size': 71939} 0%| | 0/2 [00:00<?, ?it/s] Traceback (most recent call last): File "/anaconda3/envs/py35/lib/python3.5/runpy.py", line 193, in _run_module_as_main "main", mod_spec) File "/anaconda3/envs/py35/lib/python3.5/runpy.py", line 85, in _run_code exec(code, run_globals) File "/Users/i863717/Desktop/bi-att-flow/basic/cli.py", line 109, in tf.app.run() File "/anaconda3/envs/py35/lib/python3.5/site-packages/tensorflow/python/platform/app.py", line 30, in run sys.exit(main(sys.argv[:1] + flags_passthrough)) File "/Users/i863717/Desktop/bi-att-flow/basic/cli.py", line 106, in main m(config) File "/Users/i863717/Desktop/bi-att-flow/basic/main.py", line 23, in main _train(config) File "/Users/i863717/Desktop/bi-att-flow/basic/main.py", line 99, in _train loss, summary, train_op = trainer.step(sess, batches, get_summary=get_summary) File "/Users/i863717/Desktop/bi-att-flow/basic/trainer.py", line 69, in step sess.run([self.loss, self.summary, self.train_op], feed_dict=feed_dict) File "/anaconda3/envs/py35/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 717, in run run_metadata_ptr) File "/anaconda3/envs/py35/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 894, in _run % (np_val.shape, subfeed_t.name, str(subfeed_t.get_shape()))) ValueError: Cannot feed value of shape (0,) for Tensor 'model_0/new_emb_mat:0', which has shape '(?, 100)'.

However,

python -m basic.cli --mode train --noload --debug

(which is the first command from README section 2) works perfectly fine!

Lukas

[dengyuning]

same problem

[Arminius4]

I was able to avoid this problem by setting known_if_glove to False

python -m basic.cli --mode train --noload --debug --finetune True --known_if_glove False

in basic/read_data.py line 188

if count > config.word_count_th or (config.known_if_glove and word in word2vec_dict))}

this parameter is checked in the condition for filling the shared dictionary.

In basic/cli.py there seems to be a mistake

flags.DEFINE_bool("known_if_glove", True, "consider as known if present in glove [False]")

as the default given in the explanatory string does not match the actual code, opposed to all the other lines. After changing it to False, it worked!

Lukas