abu shahid
abu shahid
It is a filtering issue. Just turn it off by `--data_filtering_off` when training
I have tried training the model ob various indic scripts and although, the training accuracies are decent; evaluation is straight up bad for some reason.
Dropping in because I too am interested.
Can you please elaborate on what you mean by how to use? You can use pre-trained weights and `read.py`. You can also change the charset in test.py to numeric characters...
So I tinkered a lot and this is perhaps due to 'label_length' in main.yaml and the image size you are giving to the model. In my case, with the model...
I did something similar, but instead of concating, i used a for loop. Got the exact same error. Were you able to resolve?
Yes. Can we have weights? It will be very helpful.
Faced the same issue. Awaiting fix.
@hinthornw Apoogies. I must have mixed things up. I was referring to this: https://python.langchain.com/v0.1/docs/modules/model_io/llms/llm_caching/ And since I am using langsmith for orchestration and calling of prompts, got confused. I am...
Is this issue resolved? I am facing the same error. Apparently this issue was there before as well and the authors claim to have fixed it. Not really sure why...