Hello,
I'm trying to train model using unsupervised.py with each possible pairing in the set {English, German, Spanish, French, Italian, Portuguese}. The line of code I'm running each time is
python unsupervised.py --src_lang en --tgt_lang es --src_emb data/wiki.en.vec --tgt_emb data/wiki.es.vec --n_refinement 5 --export pth
(obviously changing the parameters for language and embeddings each time)
This has worked fine for 12 of the pairings, and I have saved 12 fitted models, but for some reason when training French-Italian, French-Portuguese, or Italian-Portuguese, I get the following error at the end of the first Adversarial Training epoch:
Traceback (most recent call last):
File "unsupervised.py", line 139, in
evaluator.all_eval(to_log)
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/evaluator.py", line 217, in all_eval
self.word_translation(to_log)
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/evaluator.py", line 116, in word_translation
results = get_word_translation_accuracy(
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/word_translation.py", line 96, in get_word_translation_accuracy
dico = load_dictionary(path, word2id1, word2id2)
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/word_translation.py", line 63, in load_dictionary
word1, word2 = parts
ValueError: too many values to unpack (expected 2)
I've gone through it multiple times and can't see why this only occurs for these language pairs.
Hello, I'm trying to train model using unsupervised.py with each possible pairing in the set {English, German, Spanish, French, Italian, Portuguese}. The line of code I'm running each time is
python unsupervised.py --src_lang en --tgt_lang es --src_emb data/wiki.en.vec --tgt_emb data/wiki.es.vec --n_refinement 5 --export pth(obviously changing the parameters for language and embeddings each time) This has worked fine for 12 of the pairings, and I have saved 12 fitted models, but for some reason when training French-Italian, French-Portuguese, or Italian-Portuguese, I get the following error at the end of the first Adversarial Training epoch:Traceback (most recent call last): File "unsupervised.py", line 139, in
evaluator.all_eval(to_log)
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/evaluator.py", line 217, in all_eval
self.word_translation(to_log)
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/evaluator.py", line 116, in word_translation
results = get_word_translation_accuracy(
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/word_translation.py", line 96, in get_word_translation_accuracy
dico = load_dictionary(path, word2id1, word2id2)
File "/mnt/batch/tasks/shared/LS_root/mounts/clusters/cbs11/code/Users/cbs1/muse/src/evaluation/word_translation.py", line 63, in load_dictionary
word1, word2 = parts
ValueError: too many values to unpack (expected 2)
I've gone through it multiple times and can't see why this only occurs for these language pairs.