-
Dear,
我是想在您原有的pytorch ocr识别模型ocr-lstm.pth基础上,再增加自己的训练集,自己准备好了行图片与行文本,训练代码参考我在网上找了只有https://github.com/Aurora11111/crnn-train-pytorch与https://github.com/meijieru/crnn.pytorch?里面有不少坑,比如不用warp_ctc_p…
-
Log:
```
(bonitox) usamec@bonito3train:~$ bonito train --multi-gpu bonitobase
[loading data]
[loading model]
[0/1184825]: 0%| …
-
@ku21fan, I did some plots of STN results and seems like flat angles are detected, but roundings not. Maybe 20 point are too much, or we need more such examples (or more focused loss)?
![STN](https…
-
Hi @ku21fan
Just want to know which hyperparameters (no. of iterations, lr, batch size, optimizer) are used to train TPS-ResNet-BiLSTM-Attn-case-sensitive model? also what was the final training los…
-
Hi,
I'm trying to run the conformer_ctc recipe for LibriSpeech. If I use a single gpu (i.e., world-size = 1), the recipe does work without any issue.
If I use multiple GPUs, the reciped gets st…
-
modelscope版本:1.15.0
python版本:3.8.0
运行CTC的demo程序:
```
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks
kws = pipeline(
Tasks.keyword_spotting,
…
-
i got these error when installing ctcdecode in window
"
PS F:\code\VAC_CSLR-main1\ctcdecode> pip install .
Processing f:\code\vac_cslr-main1\ctcdecode
Preparing metadata (setup.py) ... done
…
-
- [ ] make sure errors are being handled
- [x] implement verifyToken for all routes
Also quick note: there's a [bug](https://docs.google.com/document/d/1keGlsJzReiEsfDA2yowRC5pJLx7qyitGGP18jkfZCyg/ed…
-
After running `ctc setup` to config my RPC url, and then running `llama pools`, I got the following error:
```
ConnectionKey(host='yields.llama.fi', port=443, is_ssl=True, ssl=None, proxy=None, pr…
-
你好,请问能否提供一个预训练导出的onnx模型进行测试。谢谢