Jp

clicking on an image to see it "full-screen" yields a blank windows until CTRL-R is pressed

14

I have this weird issue, both on my freshly install Windows 11 laptop and my linux (Pop OS/Ubuntu 22.04) laptop. When in a chat, someone sends me a picture, if...

[question/feature-request] prune trailing whitespaces at read time

8

Thanks for this great tool! I've search to see if this was possible but couldn't find it. I have many csv files where the values are padded with whitespaces (from...

unified transformer : training loss start raising suddenly

8

I'm reusing the librispeech configuration for the unified transformer, but I'm using our own internal data (about 10k hours), with a 10,000 BPE model as the tokenizer, on 8 A100...

getting spurious "HSIC computation resulted in NANs"

6

Thanks for this great little module! I was able to adapt the code to deal with models suitable for speech recognition (mostly transformers and conformers) and I'm learning a lot...

I'm curious what your thoughts are on implementing something like the fast [Model M Lite](https://domino.watson.ibm.com/library/cyberdig.nsf/1e4115aea78b6e7c85256b360066f0d4/b56e88b4b3e13cd68525807d0050f777!OpenDocument&Highlight=0,speech) into KenLM?

incompatible with recent opencv versions?

3

It's a very nice project you have here. I tried to run the project on ArchLinux with opencv 4 installed and with a virtualenv just for page_dewarp and I get...

Thanks for this project! I'm currently training the small version on openwebtext with 8 x A100 gpus, using torch 2.0 (nigthly). The data is local to the instance and the...

DDP wraps the real module behind model.module, so we need to account for that when calling custom model methods.

examples/rwkv is missing the 'scripts' folder

3

The docker file has: ``` WORKDIR /build COPY ./scripts/ . ``` but tring to build the rwkv-converter, I get: ```Step 3/6 : WORKDIR /build ---> Using cache ---> 42df0f584a8b Step...

can't get proper prediction score from python when --boosting N option used at training time

2

### Describe the bug I've trained a small model using the `--boosting N` option, like for example `--loss_function logistic -b 18 --l1 0.1 --l2 0.0001 --nn 50 --boosting 5` using...