google-research / tensorflow-coder Goto Github PK

The proposal is to have the solver recognize that, say these two
calls are equivalent, so once you've tried the first you don't need to
try the second:

 tf.multiply(tf.constant(2), tf.range(5)
 tf.multiply(tf.constant(5), tf.range(2)

i.e., multiply(a,b) == multiply(b,a) and ditto for add.
Maybe a and be have to be of the same dtype however.

This is the problem I gave it (and of note is that adding constant 2 is necessary):

inputs = {
    'data': [[0, 1, 1, 0, 0],
             [0, 0, 0, 0, 0],
             [1, 1, 1, 1, 1]]
}

output = [[0, 3, 5, 6, 8],
          [0, 2, 4, 6, 8],
          [1, 3, 5, 7, 9]]

constants = [2]

And these are the solutions - of course only the first is interesting since tf-coder
is just reordering args.

Found solution: tf.add(data, tf.multiply(tf.constant(2), tf.range(5)))
Found solution: tf.add(data, tf.multiply(tf.range(5), tf.constant(2)))
Found solution: tf.add(tf.multiply(tf.constant(2), tf.range(5)), data)
Found solution: tf.add(tf.multiply(tf.range(5), tf.constant(2)), data)

Hello there.
Thank you very much for this good tool.
I have a question for you. If I want to train on this model, which file should I start with?
Forgive me for this very basic question.

Hello,I found a performance issue in the definition of load_data ,
tf_coder/datasets/github/data_loader.py,
dataset = dataset.map(parse_example_proto) was called without num_parallel_calls.
I think it will increase the efficiency of your program if you add this.

Here is the documemtation of tensorflow to support this thing.

Looking forward to your reply. Btw, I am very glad to create a PR to fix it if you are too busy.

Hi.
Thanks a lot for this great tool.
I'm trying to use it for a simple op, where the input is a Tensor of shape (num_boxes, 4) - where num_boxes can be any value, and the output would have a fixed shape (10, 4). Basically, it would either pad if num_boxes is less than 10, or slice the first 10 otherwise.

Am not clear how I could set various examples for inputs, since they would have different shapes.

Could you please give me a hint ?

Thanks a lot ! :)