ibm/watson-machine-learning-samples

Watson Machine Learning sample models, notebooks and apps.

watson-machine-learning-samples's Issues

Use scikit-learn to predict diabetes progression with Watson Machine Learning REST API is throwing error

cloud notebook, https://github.com/IBM/watson-machine-learning-samples/blob/master/cloud/notebooks/rest_api/curl/deployments/scikit/Use%20scikit-learn%20to%20predict%20diabetes%20progression.ipynb throwing error:

metadata = {
    client.repository.ModelMetaNames.NAME: "CARS4U - Satisfaction Prediction Model",
    client.repository.ModelMetaNames.TYPE: 'tensorflow_2.7',
    client.repository.ModelMetaNames.SOFTWARE_SPEC_UID : software_spec_id
}
print(metadata)    
stored_model_details = client.repository.store_model(tar_filename, meta_props=metadata)

is throwing below error:

Failure during creating model. (POST https://us-south.ml.cloud.ibm.com/ml/v4/models?version=2023-09-06&space_id=807063b4-90d7-4dae-882e-804446280c35)
Status code: 400, body: {
  "trace": "2a5cdea2f1102ba3d08363616f5e3d96",
  "errors": [{
    "code": "invalid_request_entity",
    "message": "Invalid request entity: Unsupported software specification 'runtime-22.1-py3.9'.",
    "more_info": "https://cloud.ibm.com/apidocs/machine-learning"
  }],
  "status_code": "400"
}

"Use Keras to transfer image style.ipynb" training fails

In "Use Keras to transfer image style.ipynb", the training fails because the data and result buckets are swapped.
The job setup should swap their usage of buckets[0] and buckets[1] to fix the problem

hard to find watsonx.ai/genai samples

The watsonxai-python-sdk page links to this repo for examples, but all examples are in folders called cpdx.x or cloud. It took me quiet some searching to find a simple "hello world" example to use wml to consume a LLM through watsonx.ai

It would be easier, if the folder structure would be simplified or if there were probler links on the documentation page

"mllib_3.0" update to "mllib_3.2"

"mllib_3.0" is now depreciated -- kindly update to Spark "mllib_3.2"

Broken link to DO notebooks

The README link inside cloud/notebooks/python_sdk/deployments/decision_optimization
this repository
is broken, it should probably be updated to
https://github.com/IBMDecisionOptimization/DO-Samples/tree/watson_studio_cloud/jupyter/watsonx.ai%20and%20Cloud%20Pak%20for%20Data%20as%20a%20Service

Expose WML training source code explicitly in the notebook

This is a request to simplify and enhance the examples which used an opaque zipped .py source to run in WML training backend.
The enhancement idea would be to extract that code from the zipss, and place it in a %writefile cell in the notebook, so that the code can be inspected and modified.
The code would then be zipped from the notebook and provided to WML as usual.

PyTorch Deployment Example - unsupported versions

Following Jupyter Notebook contains unsupported runtime versions:
cloud/notebooks/python_sdk/deployments/pytorch/Use Pytorch to recognize hand-written digits.ipynb

runtime-22.1-py3.9 is unsupported at the moment (runtime-22.2-py3.10 works)
pytorch-onnx_1.10 is unsupported at the moment (pytorch-onnx_1.12 works)

link wml and ws to wx.ai

is it possible to create a space/project in wx.ai and link a wml and ws instance to that wx.ai space/project?

Training runs get stucked in the running state when run the "Keras and hyperparameter optimization (HPO) to recognize hand-written digits" experiment

I'm trying to run the "Keras and hyperparameter optimization (HPO) to recognize hand-written digits" experiment, but my trainings are never completed