Kang Min Yoo

For SNIPS, the classification accuracy (bert-base-uncased) using randomly sampled 1% of the training set is as follows. ```json { "acc": { "mean": 0.9297142857142857, "std": 0.02000145767282727, "raw": [ 0.9042857142857142, 0.9285714285714286, 0.9542857142857143,...

Thanks for the reply. The SNIPS results of 92% accuracy at 1% data level (around 20 examples per class) are definitely plausible. As an indirect evidence, you can check out...

@karthiksharma98 Hi, the prediction code has been temporarily removed during the overhaul, but you can access the old code from the past commit (https://github.com/kaniblu/pytorch-bilstmcrf/commit/386ddf9da5a5c2ae1ff2783f7e04694296636b98).

You are correct. I have made some fixes since then, but I haven't push it to the repo yet. I will hopefully do so soon.

Nice find. I am currently working on a major update that can replicate results in [arXiv:1607.06450](https://arxiv.org/abs/1607.06450) and work faster in Multi-GPU environment. This minor issue should be fixed by then.