Huggingface Examples

Description: Some example programs that use Huggingface datasets, transformers, and tokenizer libraries.

Fine-tuning GPT-2 for Magic the Gathering

Status: Complete (Validated)

Fine-tuning BERT on IMDB

Status: Complete (Validating)

Image2Image with Stable Diffusion

Status: Complete (Validated)

InPainting with Stable Diffusion

Status: Complete (Validated)

Huggingface with Tensorflow 2.0

Status: Complete (Validated)

Text2image with Stable Diffusion

Status: Complete (Validated)

Additional Sources:

• Huggingface Transformers course

Notes:

• If you can't get an example to run on docker, chances are I ran it on my Dell desktop bare metal (Intel i7-10700 CPU @ 2.90GHz, RAM 16GB, GPU Nvidia GeForce RTX 2060 SUPER 8GB VRAM). There is a slight bit of resource overhead when running an example on Docker in Windows 10 vs running the program bare metal on the OS.
• I dont want to keep updating the Text2Image_StableDiffusion/ or the MiscStyles_StableDiffusion/ folders when there are stable diffusion updates. Please refer to the following below for any additional major updates with Stable Diffusion.
  • Stable Diffusion XL
    • Pipeline documentation
    • Stable Diffusion XL 1.0 (base) model card
    • Stable Diffusion XL 1.0 (refiner) model card
    • Stable Diffusion XL is very large (and as a result, resource hungry). If it fails to initialize on a device like the DarkStar GPU server, make sure the primary GPU card is available to run it (see this relevant GitHub issue thread).