| Commit message (Collapse) | Author | Age |
|
|
|
|
|
|
|
| |
self.test_session() has been deprecated in 9962eb5e84b15e309410071b06c2ed2d6148ed44 as its name confuses readers of the test. Moving to cached_session() instead which is more explicit about:
* the fact that the session may be reused.
* the session is not closed even when doing a "with self.test_session()" statement.
PiperOrigin-RevId: 213944355
|
|
|
|
| |
PiperOrigin-RevId: 190878279
|
|
|
|
| |
PiperOrigin-RevId: 173889798
|
|
|
|
|
|
|
|
|
|
| |
tf.contrib.text.skip_gram_sample_with_text_vocab op. This argument mirrors `corpus_size` argument used by tf.contrib.text.skip_gram_sample op.
It allows vocab subsampling to work properly when vocabulary files have been preprocessed to eliminate all infrequent tokens (where frequency < vocab_min_count) to save the memory for those unused tokens. This preprocessing will help to train skip-gram models on a large vocabulary where memory usage of the internal lookup table could become a performance bottleneck.
If `corpus_size` is needed but not supplied, then it will be calculated from `vocab_freq_file`, which is the same as the existing implementation.
PiperOrigin-RevId: 166873744
|
|
|
|
| |
PiperOrigin-RevId: 159591663
|
|
|
|
| |
PiperOrigin-RevId: 157783952
|
|
PiperOrigin-RevId: 157655970
|