{"id":15677143,"url":"https://github.com/sayakpaul/bit-jax2tf","last_synced_at":"2025-05-07T00:44:50.185Z","repository":{"id":106647724,"uuid":"399688436","full_name":"sayakpaul/BiT-jax2tf","owner":"sayakpaul","description":"This repository hosts the code to port NumPy model weights of BiT-ResNets to TensorFlow SavedModel format.","archived":false,"fork":false,"pushed_at":"2021-12-21T04:26:49.000Z","size":15,"stargazers_count":14,"open_issues_count":0,"forks_count":2,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-03-31T04:41:12.330Z","etag":null,"topics":["bit-resnet","computer-vision","jax","tensorflow"],"latest_commit_sha":null,"homepage":"https://tfhub.dev/sayakpaul/collections/bit-resnet/1","language":"Jupyter Notebook","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/sayakpaul.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":"2021-08-25T04:25:00.000Z","updated_at":"2024-06-24T16:08:16.000Z","dependencies_parsed_at":null,"dependency_job_id":"f7dbaf54-d7b1-4fe7-9dbe-54d292a3c80c","html_url":"https://github.com/sayakpaul/BiT-jax2tf","commit_stats":{"total_commits":7,"total_committers":1,"mean_commits":7.0,"dds":0.0,"last_synced_commit":"dbee8fad3b2e13f6e3ca8a33ae2c8dab73e5a6df"},"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sayakpaul%2FBiT-jax2tf","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sayakpaul%2FBiT-jax2tf/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sayakpaul%2FBiT-jax2tf/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sayakpaul%2FBiT-jax2tf/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/sayakpaul","download_url":"https://codeload.github.com/sayakpaul/BiT-jax2tf/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":252793562,"owners_count":21805053,"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","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":["bit-resnet","computer-vision","jax","tensorflow"],"created_at":"2024-10-03T16:08:38.197Z","updated_at":"2025-05-07T00:44:50.171Z","avatar_url":"https://github.com/sayakpaul.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# BiT-jax2tf\nThis repository hosts the code to port NumPy model weights of BiT-ResNets [1] to TensorFlow SavedModel format. These models\nare results of [2]. The original model weights come from [3]. \n\nHuge thanks to [Willi Gierke](https://ch.linkedin.com/in/willi-gierke) (of Google) for helping with the porting. \n\nThe TensorFlow SavedModels are available on TensorFlow Hub as a collection: https://tfhub.dev/sayakpaul/collections/bit-resnet/1. A total of 8 models are available:\n\n|  Model \u003cbr\u003eName \t| Input\u003cbr\u003eResolution \t|                                 Classifier                                 \t|                            Feature\u003cbr\u003eExtractor                            \t|\n|:---------------:\t|:-------------------:\t|:--------------------------------------------------------------------------:\t|:--------------------------------------------------------------------------:\t|\n| BiT-ResNet152x2 \t|         384         \t|  [Link](https://tfhub.dev/sayakpaul/bit_resnet152x2_384_classification/1)  \t|   [Link](https://tfhub.dev/sayakpaul/bit_r152x2_384_feature_extraction/1)  \t|\n| BiT-ResNet152x2 \t|         224         \t|  [Link](https://tfhub.dev/sayakpaul/bit_resnet152x2_224_classification/1)  \t|   [Link](https://tfhub.dev/sayakpaul/bit_r152x2_224_feature_extraction/1)  \t|\n|  BiT-ResNet50x1 \t|         224         \t| [Link](https://tfhub.dev/sayakpaul/distill_bit_r50x1_224_classification/1) \t| [Link](https://tfhub.dev/sayakpaul/distill_bit_r50x1_224_classification/1) \t|\n|  BiT-ResNet50x1 \t|         160         \t| [Link](https://tfhub.dev/sayakpaul/distill_bit_r50x1_160_classification/1) \t| [Link](https://tfhub.dev/sayakpaul/distill_bit_r50x1_160_classification/1) \t|\n\nYou could use the `convert_jax_weights_tf.ipynb` notebook to understand how model porting works between JAX and TensorFlow. There\nis also an experimental tool called `jax2tf` from the JAX team that you can find [here](https://github.com/google/jax/blob/main/jax/experimental/jax2tf/README.md).\n\n## References\n\n[1] [Big Transfer (BiT): General Visual Representation Learning by Kolesnikov et al.](https://arxiv.org/abs/1912.11370)\n\n[2] [Knowledge distillation: A good teacher is patient and consistent by Beyer et al.](https://arxiv.org/abs/2106.05237)\n\n[3] [BiT GitHub](https://github.com/google-research/big_transfer)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsayakpaul%2Fbit-jax2tf","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsayakpaul%2Fbit-jax2tf","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsayakpaul%2Fbit-jax2tf/lists"}