Alejandro Vaca comments

Results 33 comments of


                                            Alejandro Vaca

has an _output_shapes attribute inconsistent with the GraphDef for output #3: Dimension 0 in both shapes must be equal, but are 0 and 1. Shapes are [0] and [1].

Tensorflow always causes problems, it would be great to migrate this to PyTorch.

BART Pretraining Script

I'm also highly interested in this, is there any update?

DeBERTa-V3 Pre-training code ?

I'm also looking forward to v3 pretrain code. Is it planned to be available anytime soon? @BigBird01 :) Thanks in advance!

DeBERTa-V3 Pre-training code ?

@BigBird01 hey, any updates on this? 😄

GPU Memory Benchmark

I'm gonna add here the trials I do myself with the corresponding memory usage: dim = 1024, seq_len=8960, depth=12, heads=16, batch_size=1: 8501MiB

StopIteration error when using word_segmentation

For making it easier, I put the full code here: `symspell.load_dictionary('CREA_total.TXT', term_index=0, count_index=1, separator='\t', encoding='latin-1') result = symspell.word_segmentation('holaadiós') `

Getting black images as ouput on custom data

I'm also getting black images as output after training on custom data. However, the first patch is in color. For my training, I used train_patch_size of 60 (my LR images...

LLaMA FastTokenizer does not add `eos_token_id` at the end.

The same is happening with Falcon...

LLaMA FastTokenizer does not add `eos_token_id` at the end.

That it doesn't generate (token id 11) when calling generate, therefore it never stops generating. I have tried by setting `eos_token_id` to 193, which corresponds to `\n`, but I don't...