{"id":17324331,"url":"https://github.com/xplip/pixel","last_synced_at":"2025-08-27T03:18:20.495Z","repository":{"id":45706272,"uuid":"514064958","full_name":"xplip/pixel","owner":"xplip","description":"Research code for pixel-based encoders of language (PIXEL)","archived":false,"fork":false,"pushed_at":"2025-07-15T23:19:09.000Z","size":31042,"stargazers_count":337,"open_issues_count":10,"forks_count":36,"subscribers_count":7,"default_branch":"main","last_synced_at":"2025-07-16T23:48:15.273Z","etag":null,"topics":["deep-learning","deep-neural-networks","language-model","machine-learning","nlp","pytorch"],"latest_commit_sha":null,"homepage":"https://arxiv.org/abs/2207.06991","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/xplip.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2022-07-14T22:30:14.000Z","updated_at":"2025-07-06T19:41:34.000Z","dependencies_parsed_at":"2025-01-13T06:16:13.099Z","dependency_job_id":null,"html_url":"https://github.com/xplip/pixel","commit_stats":{"total_commits":11,"total_committers":3,"mean_commits":"3.6666666666666665","dds":0.2727272727272727,"last_synced_commit":"1b0da2942ac0da13b98fd31dbbd720f9a438d8b7"},"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/xplip/pixel","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xplip%2Fpixel","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xplip%2Fpixel/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xplip%2Fpixel/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xplip%2Fpixel/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/xplip","download_url":"https://codeload.github.com/xplip/pixel/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xplip%2Fpixel/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":272284379,"owners_count":24906877,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-08-27T02:00:09.397Z","response_time":76,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deep-learning","deep-neural-networks","language-model","machine-learning","nlp","pytorch"],"created_at":"2024-10-15T14:10:19.686Z","updated_at":"2025-08-27T03:18:20.466Z","avatar_url":"https://github.com/xplip.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# PIXEL\n\nThis repository contains code for **PIXEL**, the **Pix**el-based **E**ncoder of **L**anguage. PIXEL is a language model that operates on text rendered as images, fully removing the need for a fixed vocabulary. This effectively allows for transfer to any language and script that can be typeset on your computer screen.\n\nWe pretrained a monolingual PIXEL model on the English Wikipedia and BookCorpus (in total around 3.2B words), the same data as BERT, and showed that PIXEL substantially outperforms BERT on syntactic and semantic processing tasks on scripts that are not found in the pretraining data, but PIXEL is slightly weaker than BERT when working with Latin scripts.\n\nFor details about PIXEL, please have a look at our paper [Language Modelling with Pixels](https://arxiv.org/abs/2207.06991). Information on how to cite our work can be found at the bottom.\n\n\u003cp align=\"middle\"\u003e\n  \u003cimg src=\".github/pixel_pretraining.png\"  alt=\"PIXEL pretraining architecture\" width=\"280\"\u003e\n  \u003cimg height=\"250\" hspace=\"20\"/\u003e\n  \u003cimg src=\".github/pixel_finetuning.png\"  alt=\"PIXEL finetuning architecture\" width=\"280\"\u003e\n\u003c/p\u003e\n\nPIXEL consists of three major components: a text renderer, which draws text as an image; an encoder, which encodes the unmasked regions of the rendered image; and a decoder, which reconstructs the masked regions at the pixel level. It is built on [ViT-MAE](https://arxiv.org/abs/2111.06377).\n\nDuring pretraining, the renderer produces images containing the training sentences. Patches of these images are linearly projected to obtain patch embeddings (as opposed to having an embedding matrix like e.g. in BERT), and 25% of the patches are masked out. The encoder, which is a Vision Transformer (ViT), then only processes the unmasked patches. The lightweight decoder with hidden size 512 and 8 transformer layers inserts learnable mask tokens into the encoder's output sequence and learns to reconstruct the raw pixel values at the masked positions.\n\nAfter pretraining, the decoder can be discarded leaving an 86M parameter encoder, upon which task-specific classification heads can be stacked. Alternatively, the decoder can be retained and PIXEL can be used as a pixel-level generative language model (see Figures 3 and 6 in the paper for examples).\n\n## Demo\nCheck out our Gradio demo for text reconstruction with PIXEL at [https://huggingface.co/spaces/Team-PIXEL/PIXEL](https://huggingface.co/spaces/Team-PIXEL/PIXEL)!\n\n## Coming Soon\n- Rendering guide\n- Finetuned robustness models\n- Integration into HuggingFace transformers\n\n## Setup\n\nThis codebase is built on [Transformers](https://github.com/huggingface/transformers) for PyTorch. We also took inspiration from the original [ViT-MAE codebase](https://github.com/facebookresearch/mae). The default font `GoNotoCurrent.ttf` that we used for all experiments is a merged Noto font built with [go-noto-universal](https://github.com/satbyy/go-noto-universal).\n\nYou can set up this codebase as follows to get started with using PIXEL models:\n\n\u003cdetails\u003e\n  \u003csummary\u003e\u003ci\u003eShow Instructions\u003c/i\u003e\u003c/summary\u003e\n\u0026nbsp;\n\n1. Clone repo and initialize submodules\n```\ngit clone https://github.com/xplip/pixel.git\ncd pixel\ngit submodule update --init --recursive\n```\n\n2. Create a fresh conda environment\n```\nconda create -n pixel-env python=3.9\nconda activate pixel-env\n```\n\n3. Install Python packages\n```bash\nconda install pytorch torchvision cudatoolkit=11.3 -c pytorch\nconda install -c conda-forge pycairo pygobject manimpango\npip install --upgrade pip\npip install -r requirements.txt\npip install ./datasets\npip install -e .\n```\n\n4. (Optional) Install Nvidia Apex\n```bash\ngit clone https://github.com/NVIDIA/apex\ncd apex\npip install -v --disable-pip-version-check --no-cache-dir --global-option=\"--cpp_ext\" --global-option=\"--cuda_ext\" ./\n```\n\n5. Verify Installation on Vietnamese POS tagging\n```bash\n# Create a folder in which we keep the data\nmkdir -p data\n# Get and extract the UD data for parsing and POS tagging\nwget -qO- https://lindat.mff.cuni.cz/repository/xmlui/bitstream/handle/11234/1-4758/ud-treebanks-v2.10.tgz | tar xvz -C data\n\npython scripts/training/run_pos.py \\\n  --model_name_or_path=\"Team-PIXEL/pixel-base-finetuned-pos-ud-vietnamese-vtb\" \\\n  --data_dir=\"data/ud-treebanks-v2.10/UD_Vietnamese-VTB\" \\\n  --remove_unused_columns=False \\\n  --output_dir=\"sanity_check\" \\\n  --do_eval \\\n  --max_seq_length=256 \\\n  --overwrite_cache\n```\n\nIf everything is configured correctly, you should expect to see results similar to the following:\n\n```bash\n***** eval metrics *****\n  eval_accuracy           =     0.8632\n  eval_loss               =     1.2375\n```\n\n\u003c/details\u003e\n\n## Pretraining PIXEL\n\nWe provide instructions for pretraining PIXEL in [PRETRAINING.md](.github/PRETRAINING.md).\n\nYou can find our pretrained PIXEL-base at [https://huggingface.co/Team-PIXEL/pixel-base](https://huggingface.co/Team-PIXEL/pixel-base).\n\n**Note**: This link also gives access to all intermediate training checkpoints from 10k to 1M steps through the commit history. You can select these checkpoints when finetuning PIXEL via `--model_revision=\u003ccommit_id\u003e`\n\nOur pretraining datasets are also on the HuggingFace hub and can be loaded via the [datasets](https://github.com/huggingface/datasets) library:\n- [Team-PIXEL/rendered-bookcorpus](https://huggingface.co/datasets/Team-PIXEL/rendered-bookcorpus)\n- [Team-PIXEL/rendered-wikipedia-english](https://huggingface.co/datasets/Team-PIXEL/rendered-wikipedia-english)\n\n## Finetuning PIXEL\n\nWe provide instructions for finetuning PIXEL in [FINETUNING.md](.github/FINETUNING.md). If you follow our training recipes or simply evaluate using the models we provide via the links below, you can expect similar results as below.\n\n**Note**: The links give access to all 5 random seeds that we averaged results over for each model (one in the main branch, and the others in branches seed2–seed5). You can select different seeds via `--model_revision=\u003cbranch_name\u003e`.\n\n\n#### Universal Dependencies (POS Tagging and Dependency Parsing)\n\n\u003cdetails\u003e\n  \u003csummary\u003e\u003ci\u003eShow Table\u003c/i\u003e\u003c/summary\u003e\n\u0026nbsp;\n\n\n|                                   \t|                                             English-EWT                                            \t|                                             Arabic-PADT                                            \t|                                             Coptic-Scriptorium                                            \t|                                             Hindi-HDTB                                            \t|                                             Japanese-GSD                                            \t|                                             Korean-GSD                                            \t|                                           Tamil-TTB                                           \t|                                             Vietnamese-VTB                                            \t|                                             Chinese-GSD                                            \t|\n|-----------------------------------\t|:--------------------------------------------------------------------------------------------------:\t|:--------------------------------------------------------------------------------------------------:\t|:---------------------------------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------------------------:\t|:---------------------------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------------------------:\t|:---------------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------------------:\t|:--------------------------------------------------------------------------------------------------:\t|\n| **POS Tagging**\u003cbr /\u003e*Accuracy*   \t|   96.7\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-english-ewt)   \t|   95.7\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-arabic-padt)   \t|   96.0\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-coptic-scriptorium)   \t|   96.3\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-hindi-hdtb)   \t|   97.2\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-japanese-gsd)   \t|   94.2\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-korean-gsd)   \t|  81.0\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-tamil-ttb) \t|   85.7\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-vietnamese-vtb)   \t|   92.8\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-pos-ud-chinese-gsd)   \t|\n| **Dependency Parsing**\u003cbr /\u003e*LAS* \t| 88.7\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-english-ewt) \t| 77.3\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-arabic-padt) \t| 83.5\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-coptic-scriptorium) \t| 89.2\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-hindi-hdtb) \t| 90.7\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-japanese-gsd) \t| 78.5\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-korean-gsd) \t| 52.6\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-tamil-ttb) \t| 50.5\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-vietnamese-vtb) \t| 73.7\u003cbr /\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-parsing-ud-chinese-gsd) \t|\n\n\u003c/details\u003e\n\n#### MasakhaNER\n\n\u003cdetails\u003e\n  \u003csummary\u003e\u003ci\u003eShow Table\u003c/i\u003e\u003c/summary\u003e\n\u0026nbsp;\n\n|            \t|                                 ConLL-2003\u003c/br\u003e English                                 \t|                                          Amharic                                          \t|                                           Hausa                                           \t|                                            Igbo                                           \t|                                        Kinyarwanda                                        \t|                                          Luganda                                          \t|                                            Luo                                            \t|                                     Naija \u003c/br\u003e Pidgin                                    \t|                                          Swahili                                          \t|                                           Wolof                                           \t|                                           Yorùbá                                          \t|\n|------------\t|:---------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|:-----------------------------------------------------------------------------------------:\t|\n| *F1 Score* \t| 89.5\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-conll2003-en) \t| 47.7\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-amh) \t| 82.4\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-hau) \t| 79.9\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-ibo) \t| 64.2\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-kin) \t| 76.5\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-lug) \t| 66.6\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-luo) \t| 78.7\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-pcm) \t| 79.8\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-swa) \t| 59.7\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-wol) \t| 70.7\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-masakhaner-yor) \t|\n\n\u003c/details\u003e\n\n#### GLUE Validation Sets\n\n\u003cdetails\u003e\n  \u003csummary\u003e\u003ci\u003eShow Table\u003c/i\u003e\u003c/summary\u003e\n\u0026nbsp;\n\n|                                  MNLI-M/MM\u003c/br\u003e *Acc*                                  \t|                                  QQP\u003c/br\u003e *F1*                                 \t|                                 QNLI\u003c/br\u003e *Acc*                                 \t|                                 SST-2\u003c/br\u003e *Acc*                                \t|                           COLA\u003c/br\u003e *Matthew's Corr.*                           \t|                            STS-B\u003c/br\u003e *Spearman's ρ*                            \t|                                  MRPC\u003c/br\u003e *F1*                                 \t|                                 RTE\u003c/br\u003e *Acc*                                 \t|                                 WNLI\u003c/br\u003e *Acc*                                 \t|    Avg    \t|\n|:--------------------------------------------------------------------------------------:\t|:------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------:\t|:------------------------------------------------------------------------------:\t|:-------------------------------------------------------------------------------:\t|:---------:\t|\n| 78.1 / 78.9\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-mnli) \t| 84.5\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-qqp) \t| 87.8\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-qnli) \t| 89.6\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-sst2) \t| 38.4\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-cola) \t| 81.1\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-stsb) \t| 88.2\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-mrpc) \t| 60.5\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-rte) \t| 53.8\u003c/br\u003e [Models](https://huggingface.co/Team-PIXEL/pixel-base-finetuned-wnli) \t| 74.1\u003c/br\u003e \t|\n\n\u003c/details\u003e\n\n#### Question Answering (TyDiQA-GoldP, SQuAD, KorQuAD 1.0, JaQuAD)\n\n\u003cdetails\u003e\n  \u003csummary\u003e\u003ci\u003eShow Table\u003c/i\u003e\u003c/summary\u003e\n\u0026nbsp;\nNotes:\n\n1) To obtain per-language predictions and scores for TyDiQA-GoldP, follow the instructions from [https://github.com/google-research-datasets/tydiqa/tree/master/gold_passage_baseline](https://github.com/google-research-datasets/tydiqa/tree/master/gold_passage_baseline)\n2) To reproduce our scores for KorQuAD, use the official KorQuAD evaluation script available [here](https://korquad.github.io/KorQuad%201.0/)\n\n\u003ctable\u003e\n\u003cthead\u003e\n  \u003ctr\u003e\n    \u003cth rowspan=\"2\"\u003c/th\u003e\n    \u003cth colspan=\"10\"\u003eTyDiQA-GoldP\u003c/th\u003e\n    \u003cth\u003eSQuADv1\u003c/br\u003e\u003c/th\u003e\n    \u003cth\u003eKorQuADv1\u003cbr\u003e\u003c/th\u003e\n    \u003cth\u003eJaQuAD\u003cbr\u003e\u003c/th\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003eEnglish\u003c/td\u003e\n    \u003ctd\u003eArabic\u003c/td\u003e\n    \u003ctd\u003eBengali\u003c/td\u003e\n    \u003ctd\u003eFinnish\u003c/td\u003e\n    \u003ctd\u003eIndonesian\u003c/td\u003e\n    \u003ctd\u003eKorean\u003c/td\u003e\n    \u003ctd\u003eRussian\u003c/td\u003e\n    \u003ctd\u003eSwahili\u003c/td\u003e\n    \u003ctd\u003eTelugu\u003c/td\u003e\n    \u003ctd\u003eAvg\u003c/td\u003e\n    \u003ctd\u003eEnglish\u003c/td\u003e\n    \u003ctd\u003eKorean\u003c/td\u003e\n    \u003ctd\u003eJapanese\u003c/td\u003e\n  \u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n  \u003ctr\u003e\n    \u003ctd\u003e\u003ci\u003eF1 Score\u003c/i\u003e\u003c/td\u003e\n    \u003ctd\u003e59.6\u003c/td\u003e\n    \u003ctd\u003e57.3\u003c/td\u003e\n    \u003ctd\u003e36.3\u003c/td\u003e\n    \u003ctd\u003e57.1\u003c/td\u003e\n    \u003ctd\u003e63.6\u003c/td\u003e\n    \u003ctd\u003e26.1\u003c/td\u003e\n    \u003ctd\u003e50.5\u003c/td\u003e\n    \u003ctd\u003e65.9\u003c/td\u003e\n    \u003ctd\u003e61.7\u003c/td\u003e\n    \u003ctd\u003e52.3\u003c/td\u003e\n    \u003ctd\u003e81.4\u003c/td\u003e\n    \u003ctd\u003e78.0\u003c/td\u003e\n    \u003ctd\u003e34.1\u003c/td\u003e\n\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003eURL\u003c/td\u003e\n    \u003ctd colspan=\"10\" align=\"center\" \u003e\u003ca href=\"https://huggingface.co/Team-PIXEL/pixel-base-finetuned-tydiqa-goldp\"\u003eModels\u003c/a\u003e\u003c/td\u003e\n    \u003ctd\u003e\u003ca href=\"https://huggingface.co/Team-PIXEL/pixel-base-finetuned-squadv1\"\u003eModels\u003c/a\u003e\u003c/td\u003e\n    \u003ctd\u003e\u003ca href=\"https://huggingface.co/Team-PIXEL/pixel-base-finetuned-korquadv1\"\u003eModels\u003c/a\u003e\u003c/td\u003e\n    \u003ctd\u003e\u003ca href=\"https://huggingface.co/Team-PIXEL/pixel-base-finetuned-jaquad\"\u003eModels\u003c/a\u003e\u003c/td\u003e\n\n  \u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\n\u003c/details\u003e\n\n### Citation \u0026 Contact\n\n```bibtex\n@inproceedings{rust-etal-2023-pixel,\n  title={Language Modelling with Pixels},\n  author={Phillip Rust and Jonas F. Lotz and Emanuele Bugliarello and Elizabeth Salesky and Miryam de Lhoneux and Desmond Elliott},\n  booktitle={The Eleventh International Conference on Learning Representations},\n  year={2023},\n  url={https://openreview.net/forum?id=FkSp8VW8RjH}\n}\n```\n\nFeel free to open an issue here or send an email to ask questions about PIXEL or report problems with the code! We emphasize that this is experimental research code.\n\n**Contact person:**\nPhillip Rust (p.rust@di.ku.dk)\n\nIf you find this repo useful, we would also be happy about a ⭐️ :).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fxplip%2Fpixel","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fxplip%2Fpixel","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fxplip%2Fpixel/lists"}