Add the CSV hash table in Hash layer and fix a bug.

[dengc367]

• delete the Lambda sublayer in LocalActivationUnit Layer class

• add vocabulary_path in the SparseFeat to support the csv HashTable functionality

• update docs and add examples in doc

• Remove trailing whitespace

[dengc367]

this PR will be merged into release branch, and so you need merge into master branch when this PR is merged. @shenweichen

[codecov]

Codecov Report

Merging #385 (8804e7e) into release (0df401c) will increase coverage by 0.32%.
The diff coverage is 95.45%.

@@             Coverage Diff             @@
##           release     #385      +/-   ##
===========================================
+ Coverage    94.66%   94.99%   +0.32%     
===========================================
  Files           54       54              
  Lines         2924     2935      +11     
===========================================
+ Hits          2768     2788      +20     
+ Misses         156      147       -9     

Flag	Coverage Δ
pytest	94.99% <95.45%> (+0.32%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
deepctr/inputs.py	95.83% <66.66%> (ø)
deepctr/feature_column.py	92.92% <100.00%> (+0.12%)	⬆️
deepctr/layers/core.py	99.09% <100.00%> (-0.01%)	⬇️
deepctr/layers/sequence.py	90.58% <100.00%> (+0.47%)	⬆️
deepctr/layers/utils.py	91.42% <100.00%> (+6.04%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0df401c...8804e7e. Read the comment docs.

[dengc367]

I added some pytest code of hash vocabulary_path parameter

[dengc367]

@shenweichen I used your layer_test function in pytest, and when that run the line model.predict and throws the exception Failed precondition: Table not initialized., I think the pytest environment is the tfv1.

[dengc367]

It's strange, I have tested successfully in my local linux, but it always failed when using CI action with exception: Failed precondition: Table not initialized..

[shenweichen]

@shenweichen I used your layer_test function in pytest, and when that run the line model.predict and throws the exception Failed precondition: Table not initialized., I think the pytest environment is the tfv1.

i think you need to test it in your local machine in both tf1 and tf2

[dengc367]

I have modified some code to fix a bug An op outside of the function building code is being passed a "Graph" tensor in PostionalEncoding Layer, which substituted the tf.concat with np.zeros when zero_pad=True, and remove the tf.compat.v1.disable_eager_execution() in sequence_test.py, so all the unit tests in test/layers directory are successful.

[shenweichen]

committers are not allowed to merge their own branches to release, we will review your pull request soon.

[dengc367]

OK. Thanks.