Said Taghadouini

Fine-tune TextDiffuser to generate handwritten text

3

I wonder if it's reasonable to try and fine-tune TextDiffuser on handwriting generation instead of digital text. Has anyone tried this so far?

Comparison with llama.cpp and GGML

1

I wonder how this compares to llama.cpp for example in terms of performance in the same settings?

Questions regarding generation and recognition task

2

Hi there, thank you for your work! I have trained the model with a similar setup on IAM dataset. I have a few questions for you: 1- is it normal...

Add example which implements Vision Transformer(ViT) image classification

1

I have been working on implementing a ViT model using ggml here : [vit.cpp](https://github.com/staghado/vit.cpp). It is still WIP but most of the work is done. It is highly inspired by...

Training benchmarks reproduction

2

The training benchmark link no longer works : https://huggingface.co/blog/huggingface-and-optimum-amd How can one test training throughput on AMD these days? Also, can you provide details about the experiments in the figure...

This PR fixes an issue where each rank calls `torch.empty(..., device="cuda")`, which defaults to cuda:0. This causes every worker to create a CUDA context on GPU 0 (~600 MiB per...