hlky
hlky
Initial ComfyUI plugin implementation Uses basic workflows for text-to-image, image-to-image, inpaint/outpaint      * Image count in UI set to value from plugin, image count != batch...
# What does this PR do? In Diffusion models such as [`stabilityai/stable-diffusion-xl-base-1.0`](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0) generating images at resolutions far below the base training resolution gives bad results:  Excellent work in [`Extraltodeus/Stable-Diffusion-temperature-settings`](https://github.com/Extraltodeus/Stable-Diffusion-temperature-settings)...
Following [documentation](https://huggingface.co/docs/datasets/repository_structure#define-your-splits-and-subsets-in-yaml) I had defined different configs for [`Dataception`](https://huggingface.co/datasets/bigdata-pw/Dataception), a dataset of datasets: ```yaml configs: - config_name: dataception data_files: - path: dataception.parquet split: train default: true - config_name: dataset_5423 data_files:...
Currently estimated row count assumes file sizes are consistent across the entire set, from what I've seen this results in wildly inaccurate estimates for WebDataset. Typically WebDatasets are created with...
[`is_image_url`](https://github.com/huggingface/dataset-viewer/blob/946b0788fa426007161f2077a70b5ae64b211cf8/libs/libcommon/src/libcommon/utils.py#L131-L134) relies on a filename and extension being present, however, in some cases an image URL does not contain a filename. Example [dataset](https://huggingface.co/datasets/bigdata-pw/SteamScreenshots) and example [URL](https://steamuserimages-a.akamaihd.net/ugc/910172100453203507/062F4787060B2E4E93EFC4631E96183B027A860B/). This could be improved...
# What does this PR do? This PR adds support for Florence-2. Compared to the existing [remote code](https://huggingface.co/microsoft/Florence-2-base/blob/main/modeling_florence2.py) the main difference is removal of `MySequential`, removal of `einops.rearrange` and `trunc_normal_`...
# What does this PR do? In the [original code](https://github.com/aigc-apps/VideoX-Fun/blob/main/videox_fun/models/z_image_transformer2d_control.py) this is not a typical ControlNet, it is integrated into the `transformer` and relies on operations performed in the `transformer`'s...
# What does this PR do? Adds support for FlashPack FlashPack could be used as weight format only (see: `load_flashpack_checkpoint`) - keeping only weight format code would be a cleaner...