Jong Wook Kim

I confirm that the problem goes away if I don't use `biblatex` and use the old-style `\bibliography`.

@pkgw I've made a minimal example that reproduces this error: [here](https://github.com/jongwook/tectonic-biblatex-issue/blob/jibberish/main.pdf) The same package has another tag [`long_url_error`](https://github.com/jongwook/tectonic-biblatex-issue/blob/long-url-error/library.bib#L15) where `bibtex` fails to process a citation with `url` field longer than...

@mnrvwl Indeed it works! It's long been since I completed the dissertation in question, but it's a wonder of open source. Thank you so much for the news!

@NielsMayer Thanks for the PR! Would it work for you if I make this `write_tsv` instead? CSV format is not standardized and `csv.reader` and `pandas.read_csv` often create headache parsing quotes....

Thanks for accommodating the TSV suggestion! I merged a refactored version of #333 and edited this PR accordingly. The issue about CSV is that, although CSV is more widely used...

I'm hesitant to add this because the incremental utility of this compared to the probabilities returned by `word_timestamps=True` is quite niche, versus the added complexity & latency due to the...

Hi, we used full-batch linear regression using L-BFGS. For Imagenet with 1M+ images in the training split it was quite slow and requires huge memory especially considering the hyperparameter sweep...

I'd use the CLIP image encoder to get the image embeddings for each image, and then either: 1. simply take the average of all those images and use that as...

> Can we feed in an input matrix with custom positive and negative examples (where positives are not necessarily diagonal only)? It's possible, as long as you feed the data...

It appears the code is still assuming the diagonal entries are the correct labels (from `torch.diagonal`). You could consider flattening the logits and use binary cross entropy for every positive/negative...