-
**System information**
- TensorFlow version: 2.4.0
- Python version: 3.6.2
**Problem**
I am trying to upgrade the LAS model from tensorflow 1.8.0 version to 2.4.0. There is no problem in trainin…
-
From nn benchmark, there is a problem that such models are stuck at 1-st iteration.
-
（1）
转换ctc的模型时候，使用的命令是，
python3 tools/export_model.py -c configs/rec/rec_chinese_lite_train.yml -o Global.checkpoints=./ch_lite/rec_mv3_crnn/best_accuracy \
Global.save_inference_dir=./infer…
aceyw updated
3 years ago
-
I am trying to convert a keras biLSTM model to ONNX.
def gelu(x):
cdf = 0.5 * (1.0 + tf.tanh(
(np.sqrt(2 / np.pi) * (x + 0.044715 * tf.pow(x, 3)))))
return x * cdf
modelfile = "…
-
## Check List
Thanks for considering to open an issue. Before you submit your issue, please confirm these boxes are checked.
**You can post pictures, but if specific text or code is required to re…
-
Hi,
It s me again, I opened a new issue because I tried to add a **BILSTM** layer to see if it ganna enhance the performance or not
Actually the enhancement was not impressive ~2% between two mode…
-
F:\PaddleOCR>python tools/train.py -c configs/rec/rec_icdar15_train.yml
[2021/02/20 17:42:08] root INFO: Architecture :
[2021/02/20 17:42:08] root INFO: Backbone :
[2021/02/20 17:42:08] root IN…
-
The Cudnn LSTM is supposed to make the training much more efficient compared to the dynamic RNN. However I am using variable length sequences with the dynamic RNN using padded batches and passing the …
ghost updated
3 years ago
-
This is my training code
```
logger.info('Training....')
best_val_loss = []
stored_loss = 100000000
def train(train_dataloader=train_dataloader):
model.train()
start_time = datetime.…
-
看了下最后输出关键词及类别的过程，好像是将全部文字组成了一个长序列，然后在里头提取出了各字的标签，再将连续同标签的字组成知名并打个此标签。这个过程中OCR部分识别为一个box里的词组的前后部分都可能被分配到其他组。
比如OCR出了三块"XXX","YYY","ZZZ“，序列标注的拆分结果可能变成了"XX","XYY","YZZ","Z"。
假设OCR的BOX组合的过程准确率很高，那如何更好利用…