Commits

Commits on Mar 25, 2019

fix store+load model (to use RELEASE param correctly); fix transform layer to not use bias (now #trainable_params equals to orig tf model); use tf.train.AdamOptimizer() as optimizer instead keras "…
eladn
committed

Commits on Mar 24, 2019

base model redactor; OOV+PAD special words instead of NoSuch; impl save+load keras model; reader refactor
eladn
committed

Commits on Mar 22, 2019

export the `topk` param into config.TOP_K_WORDS_CONSIDERED_DURING_PREDICTION
eladn
committed
impl keras Words Subtoken Metrics (Precision, Recall, F1)
eladn
committed
export word prediction calculation from model into WordPredictionLayer
eladn
committed
keras model: add prediction tf graph, add evaluation f1 metric; pass reader.dataset directly to model.fit() instead of using reader.dataset.iterator; rename "Model" ==> "Code2VecModel"; use common.…
eladn
committed
model_base: use common.SpecialDictWords
eladn
committed
add common.tf_get_first_true()
eladn
committed
export Config to config.py; common.SpecialDictWords(Enum);
eladn
committed

Commits on Mar 21, 2019

keras model: use `tf.data` reader [now training works]
eladn
committed

Commits on Feb 28, 2019

Commits on Feb 26, 2019

doc: add to the README a way to get the trained (but not release) java14m model
danglotb
committed

Commits on Feb 23, 2019

in the scenario of "load and continue training (on possibly another dataset)": initialize NUM_EXAMPLES as the number of examples in the new dataset,
urialon
committed

Commits on Feb 16, 2019

Renaming variables to make it clear where the code vectors are
urialon
committed

Commits on Jan 18, 2019

Commits on Jan 17, 2019

README
urialon
committed

Commits on Jan 16, 2019

Commits on Jan 10, 2019

Fixing memory leak in JavaExtractor
urialon
committed

Commits on Jan 9, 2019

Duplicate path contexts in C# extractor (#7)

ASvyatkovskiy
authored and
urialon
committed