Eric Nguyen comments

Results 30 comments of


                                            Eric Nguyen

Slow dataloading every N batches, where N=num_threads

Interesting, thanks for the thoughtful response. So I am using a cluster at school, and have 28 cpu (or cores?) available. I've been doing 1 worker/cpu, which I thought was...

Progressive Learning with adaptive Regularization

did anybody ever find the code for these?

Progressive learning

Did anyone ever find the parts in the codebase for the progressive learning? It's hard to find :(

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xdd in position 112: invalid continuation byte

@cydonia999 Don't suppose you heard of anyone getting an error like this? Here's the Traceback: Traceback (most recent call last): File "tiny_face_eval.py", line 227, in main() File "tiny_face_eval.py", line 224,...

about argument parse

@greed2411 Thanks for this greed! However, I got an error and thought maybe it's from the way I converted the weight file. Did you get something like this? Traceback (most...

The configuration on coco and imagenet

Hi @silencelzx, I was curious if you were able to train HashGan on CoCo or imagenet? If so, would you be willing to share the trained model parameters?

[data request] Adobe Visual Font Recognition (VFR)

Hi everyone, thanks for all the great work! I had a few questions: 1. Is the VFR_real_u data available as BCF as well? If not, when training the stacked conv...

open in colab button does not work (Tutorial Part 13: Modeling Protein-Ligand Interactions)

same, the "open in Colab" link doesn't work for me either. However, when I just copy and paste the code cell by cell, it fails at the first one. Seems...

Mamba training loss at 7B parameters

Mamba 7B included in [preprint v2](https://www.biorxiv.org/content/10.1101/2024.02.27.582234v2) now

Fixes for training models with bf16 + freshly initialized optimizer via `load_module_only`

How come this PR was never merged? It fixed the finetuning bug I experienced myself, which took about 3-4 days... Specifically, removing the extra condition to check for fp16_enabled is...