-
### Description
The detailed figure 4 in appendix seems to do not follow the iterative equations (4)(5) in the paper. If I follow the figure, it should be H^t = LayerNorm(A^t+Transition(A^t)), and A^…
-
The same simple query in engine:
```
21 rows in set (2 min 9.63 sec)
```
native:
```
21 rows in set (3.69 sec)
```
According to my Activity Monitor it happens because of FS mapping int…
-
The current implementation from tensor2tensor is not compatible to the seq2seq library. Another possible implementation in openNMT also write the helper function like `dynamic_decode` itself. This is …
-
### Description
I tried training a transformer model for `translate_enfr_wmt32k_packed` on a single TPU-v2 (8 cores) using the default `transformer-transformer_big_enfr_tpu` hparams and noticed th…
-
Hi, I was wondering what is the 'cleanest' way to specify a GPU index to run on.
(e.g., if you have a machine with 8 GPUs, and you have been allocated GPU number 5)
I am currently able to do that …
-
**Prerequisites**
> Please fill in by replacing `[ ]` with `[x]`.
* [x] Are you running the latest `bert-as-service`?
* [x] Did you follow [the installation](https://github.com/hanxiao/bert-as-…
-
### 1. What I did:
I'm trying to train Breakout on google colab using the A3C implementation. Whenever I run
`!python train-atari.py --env Breakout-v0 --gpu 0`
It causes an error;
Traceback…
-
Move away from adopting the naming scheme (arbitrarily) chosen by authors, and establish names with consistent semantics across attacks.
Proposed naming scheme:
- "eps": the maximum perturbation…
-
This may not directly related to tensor2tensor, but I am curious about to what extend NER could improve translation quality in general. Here are some examples where NER could apply.
1. numbers, espec…
lkluo updated
5 years ago
-
I cannot generate tflite file which may be caused that frozon pb file still have training ops.
The operation like below:
use
bazel-bin/tensorflow/python/tools/freeze_graph \
--input_graph=/da…