listing model instances doesn't return previous versions

[waseem-mend]

When requesting info about model instances using Kaggle API, it returns a single result of the latest (current) version. I would expect that the API response will contain all versions.

For example:

Model request:

cURL request: curl https://www.kaggle.com/api/v1/models/tensorflow/ssd-mobilenet-v1/get -u *****:*****

response:

{
  "id": 299,
  "ref": "tensorflow/ssd-mobilenet-v1",
  "title": "ssd_mobilenet_v1",
  "subtitle": "Object detection model trained on the COCO dataset.",
  "author": "TensorFlow",
  "slug": "ssd-mobilenet-v1",
  "isPrivate": false,
  "description": "## Overview\n\nAn object detection model that has been released in the\n[Tensorflow detection model zoo](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf1_detection_zoo.md)\nand trained on the [COCO dataset](http://cocodataset.org).\n\nThis model is available at http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_2018_01_28.tar.gz.\n\nThe model is useful for out-of-the-box inference, but also useful for\ninitializing the models when training on novel datasets.",
  "instances": [
    {
      "hasBaseModelInstanceInformation": false,
      "id": 2478,
      "slug": "fpn-640x640",
      "framework": "tensorFlow2",
      "fineTunable": false,
      "overview": "SSD Mobilenet V1 Object detection model with FPN feature extractor, shared box\npredictor and focal loss, trained on COCO 2017 dataset with trainning images\nscaled to 640x640.",
      "usage": "## Overview\n\nSSD with Mobilenet V1 FPN feature extractor, shared box predictor and focal loss\n(a mobile version of [Retinanet in Lin et al](https://arxiv.org/abs/1708.02002))\ninitialized from Imagenet classification checkpoint.\n\nTrained on [COCO 2017](https://cocodataset.org/) dataset (images scaled to\n640x640 resolution).\n\nModel created using the\n[TensorFlow Object Detection API](https://github.com/tensorflow/models/tree/master/research/object_detection)\n\nAn example detection result is shown below.\n\n![Object detector output](https://www.gstatic.com/aihub/tfhub/detection/od_no_keypoints.png)\n\n#### Example use\n\n```\n# Apply image detector on a single image.\ndetector = hub.load(\u0022${URL}\u0022)\ndetector_output = detector(image_tensor)\nclass_ids = detector_output[\u0022detection_classes\u0022]\n```\n\n### Inputs\n\nA three-channel image of variable size - the model does **NOT** support\nbatching. The input tensor is a `tf.uint8` tensor with shape `[1, height, width,\n3]` with values in `[0, 255]`.\n\n### Outputs\n\nThe output dictionary contains:\n\n*   `num_detections`: a `tf.int` tensor with only one value, the number of\n    detections `[N]`.\n*   `detection_boxes`: a `tf.float32` tensor of shape `[N, 4]` containing\n    bounding box coordinates in the following order: `[ymin, xmin, ymax, xmax]`.\n*   `detection_classes`: a `tf.int` tensor of shape `[N]` containing detection\n    class index from the label file.\n*   `detection_scores`: a `tf.float32` tensor of shape `[N]` containing\n    detection scores.\n*   `raw_detection_boxes`: a `tf.float32` tensor of shape `[1, M, 4]` containing\n    decoded detection boxes without Non-Max suppression. `M` is the number of\n    raw detections.\n*   `raw_detection_scores`: a `tf.float32` tensor of shape `[1, M, 90]` and\n    contains class score logits for raw detection boxes. `M` is the number of\n    raw detections.\n*   `detection_anchor_indices`: a `tf.float32` tensor of shape `[N]` and\n    contains the anchor indices of the detections after NMS.\n*   `detection_multiclass_scores`: a `tf.float32` tensor of shape `[1, N, 91]`\n    and contains class score distribution (including background) for detection\n    boxes in the image including background class.\n\n#### Source\n\nThe model\u0027s checkpoints are\n[publicly available](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md)\nas a part of the\n[TensorFlow Object Detection API](https://github.com/tensorflow/models/tree/master/research/object_detection).\n\n#### Metrics\n\nMetric                    | Value | Outputs\n------------------------- | ----- | -------\nmAP on COCO 2017 test set | 29.1  | Boxes",
      "downloadUrl": "/models/tensorflow/ssd-mobilenet-v1/TensorFlow2/fpn-640x640/1/download",
      "versionId": 3324,
      "versionNumber": 1,
      "trainingData": [
        
      ],
      "url": "https://www.kaggle.com/models/tensorflow/ssd-mobilenet-v1/TensorFlow2/fpn-640x640",
      "licenseName": "Apache 2.0",
      "modelInstanceType": "unspecified",
      "baseModelInstanceInformation": null,
      "externalBaseModelUrl": ""
    },
    {
      "hasBaseModelInstanceInformation": false,
      "id": 2475,
      "slug": "default",
      "framework": "tfLite",
      "fineTunable": false,
      "overview": "TF Lite deployment of tensorflow/ssd_mobilenet_v1/1.",
      "usage": "## Description\nPre-trained model optimized to work with TensorFlow Lite for Object detection.\n\nSee\n[Object detection overview](https://www.tensorflow.org/lite/examples/object_detection/overview)\npage for documentation and examples.",
      "downloadUrl": "/models/tensorflow/ssd-mobilenet-v1/TfLite/default/1/download",
      "versionId": 3320,
      "versionNumber": 1,
      "trainingData": [
        
      ],
      "url": "https://www.kaggle.com/models/tensorflow/ssd-mobilenet-v1/TfLite/default",
      "licenseName": "Apache 2.0",
      "modelInstanceType": "unspecified",
      "baseModelInstanceInformation": null,
      "externalBaseModelUrl": ""
    },
    {
      "hasBaseModelInstanceInformation": false,
      "id": 2477,
      "slug": "metadata",
      "framework": "tfLite",
      "fineTunable": false,
      "overview": "TF Lite deployment of tensorflow/ssd_mobilenet_v1/2.",
      "usage": "## Description\n\nPre-trained model optimized to work with TensorFlow Lite for Object detection.\nThis model contains both TFLite model metadata and the label file.\n[TFLite metadata](https://www.tensorflow.org/lite/convert/metadata) is a rich\nmodel description including both human and machine readable information.\n\nSee\n[Object detection overview](https://www.tensorflow.org/lite/examples/object_detection/overview)\npage for documentation and examples.\n\n### Release Notes\n\nv2 - updated the models with better metadata information",
      "downloadUrl": "/models/tensorflow/ssd-mobilenet-v1/TfLite/metadata/2/download",
      "versionId": 3323,
      "versionNumber": 2,
      "trainingData": [
        
      ],
      "url": "https://www.kaggle.com/models/tensorflow/ssd-mobilenet-v1/TfLite/metadata",
      "licenseName": "Apache 2.0",
      "modelInstanceType": "unspecified",
      "baseModelInstanceInformation": null,
      "externalBaseModelUrl": ""
    },
    {
      "hasBaseModelInstanceInformation": false,
      "id": 2476,
      "slug": "default",
      "framework": "tfJs",
      "fineTunable": false,
      "overview": "TF.js deployment of tensorflow/ssd_mobilenet_v1/1.",
      "usage": "## Origin\n\nThis model is published on NPM as `@tensorflow-models/coco-ssd`.\n\nSee the [NPM documentation](https://www.npmjs.com/package/@tensorflow-models/coco-ssd)\nfor how to load and use the model in JavaScript.",
      "downloadUrl": "/models/tensorflow/ssd-mobilenet-v1/TfJs/default/1/download",
      "versionId": 3321,
      "versionNumber": 1,
      "trainingData": [
        
      ],
      "url": "https://www.kaggle.com/models/tensorflow/ssd-mobilenet-v1/TfJs/default",
      "licenseName": "Apache 2.0",
      "modelInstanceType": "unspecified",
      "baseModelInstanceInformation": null,
      "externalBaseModelUrl": ""
    }
  ],
  "tags": [
    {
      "nameNullable": "image",
      "descriptionNullable": null,
      "fullPathNullable": "data type \u003e image",
      "ref": "image",
      "name": "image",
      "hasName": true,
      "description": "",
      "hasDescription": false,
      "fullPath": "data type \u003e image",
      "hasFullPath": true,
      "competitionCount": 379,
      "datasetCount": 7233,
      "scriptCount": 5590,
      "totalCount": 13202
    },
    {
      "nameNullable": "object detection",
      "descriptionNullable": "",
      "fullPathNullable": "task \u003e object-detection",
      "ref": "object detection",
      "name": "object detection",
      "hasName": true,
      "description": "",
      "hasDescription": true,
      "fullPath": "task \u003e object-detection",
      "hasFullPath": true,
      "competitionCount": 5,
      "datasetCount": 480,
      "scriptCount": 231,
      "totalCount": 716
    }
  ],
  "publishTime": "2020-10-06T00:00:00Z",
  "provenanceSources": "https://github.com/tensorflow/tfhub.dev/tree/master/assets/docs/tensorflow",
  "url": "https://www.kaggle.com/models/tensorflow/ssd-mobilenet-v1"
}

Model instance request:

cURL request:
curl https://www.kaggle.com/api/v1/models/tensorflow/ssd-mobilenet-v1/tfLite/metadata/get -u *****:****

response:

{
  "hasBaseModelInstanceInformation": false,
  "id": 2477,
  "slug": "metadata",
  "framework": "tfLite",
  "fineTunable": false,
  "overview": "TF Lite deployment of tensorflow/ssd_mobilenet_v1/2.",
  "usage": "## Description\n\nPre-trained model optimized to work with TensorFlow Lite for Object detection.\nThis model contains both TFLite model metadata and the label file.\n[TFLite metadata](https://www.tensorflow.org/lite/convert/metadata) is a rich\nmodel description including both human and machine readable information.\n\nSee\n[Object detection overview](https://www.tensorflow.org/lite/examples/object_detection/overview)\npage for documentation and examples.\n\n### Release Notes\n\nv2 - updated the models with better metadata information",
  "downloadUrl": "/models/tensorflow/ssd-mobilenet-v1/TfLite/metadata/2/download",
  "versionId": 3323,
  "versionNumber": 2,
  "trainingData": [
    
  ],
  "url": "https://www.kaggle.com/models/tensorflow/ssd-mobilenet-v1/TfLite/metadata",
  "licenseName": "Apache 2.0",
  "modelInstanceType": "unspecified",
  "baseModelInstanceInformation": null,
  "externalBaseModelUrl": ""
}

As you can see, the response contains only the latest version "versionNumber":2

[stevemessick]

I don't think we officially support using curl to download anything. It certainly is not part of the Kaggle API (this GitHub project).

However, listing model instance versions seems like a reasonable feature. I added a feature request to our internal issue tracker.

http://b/346568526

[waseem-mend]

@stevemessick since Kaggle API uses the REST API, I assumed it's the right project to open issue for.

[stevemessick]

@waseem-mend This isn't the best place to make that request. You might get more traction by posting in the product feedback forum:
https://www.kaggle.com/discussions/product-feedback

The reason is that this GitHub project is for Python code that interacts with the Kaggle server, which is also what your curl commands do. In order to get a different response, the Kaggle site has to be changed.

In the issue I mentioned above, I suggested we add support for two new CLI commands:

kaggle models instances list ...
kaggle models instances versions list ...

That will require some work here, but it also requires work on the Kaggle site. Posting in the product feedback forum increases the visibility of the request. (Not too many people read this issue tracker, but many Kagglers read the PF forum.)