nzinfo

对隐式借用的形成方式描述有错误。

val_loss frozen at 0.2000

1

When I started training process, the val_loss fixed. Epoch 00001: val_loss improved from inf to 0.20000, saving model to models/model.01-0.2000.hdf5 Epoch 2/15 211/1539 [===>..........................] - ETA: 48:51 - loss: 0.1927...

There's no padding_mask in TransformerBlock

1

When I repeat the training process, and got an error ``` Using TensorFlow backend. mask is Tensor("lambda_1/MatMul:0", shape=(?, 150, 150), dtype=float32) Traceback (most recent call last): File "/data/nlp/transformer-word-segmenter/tf_segmenter/__init__.py", line 321,...

1. Upgrade dependency. 2. Change `struct File` to `struct File

Improve article 2305.15852

1

**Exact location of issue** Article content is missing. **Problem details** The [conversion log](https://ar5iv.labs.arxiv.org/log/2305.15852) suggests that `iclr2024_conference.sty` and `soulutf8.sty` need new support for latexml.