-
I have a strange problem.
When running spm.SentencePieceTrainer.Train(" ".join(sys.argv[1:])),
model_type: UNIGRAM
vocab_size: 600
I get an error
unigram_model_trainer.cc(247) LOG(INFO) Making s…
-
**Describe the bug**
In latest Unigram version 9.3 (7970), it is not possible to send a message silently to any topic. It is actual for groups that enabled "Topics" (aka "Forum") feature.
**To Rep…
-
**Describe the bug**
When I switch to another app with Alt+Tab and back to Unigram, sometimes it doesn't take any input after that switch
I have to minimize and maximize Unigram, or close/open it, t…
-
I downloaded unigram through windows store and I changed the download storage path to another drive, it just stopped working just blank screen and crashes. if I reset and restore to old settings it wo…
-
Expand the standard tutorial to include the new segmenters.
- [ ] tweet segmenter
- [ ] unigram segmenter
- [ ] word segmenter cascades
-
满足以下要求:
1、记录下特征,包括记录下原来的句子,及一些信息,
以csv记录
2、组合模式,能收集到所有使用的Feature名字
-
**Describe the bug**
I am training a conformer-transducer medium model and on the validation set I reach `val_wer=0.0157-epoch=13.ckpt`, where val_wer is actualy ChER as i use `use_cer=True`. When …
-
The memory footprint of TR is too large, it requires at least 20GB to run. I believe this is due to a couple word embedding models we load in memory for grounding.
Can we mmap them such that the ov…
-
Command ['/usr/local/bin/python3.11', '-m', 'pip', 'install', '--use-pep517', '--disable-pip-version-check', '--prefix', '/usr/local', '--no-deps', '/root/.cache/pypoetry/artifacts/fa/e3/33/3b69b842d7…
-
### Steps to reproduce
1. Start the app.
2. Connection goes forever.
### Expected behaviour
Connection stage should pass and app should show channels.
### Actual behaviour
Connection goes for…