Opinie ([DEPRECATED] Continuous Training with TensorFlow, PyTorch, XGBoost, and Scikit Learn Models with Kubeflow and AI Platform Pipelines)

Siegfried H. · Sprawdzono około 2 lata temu

Too many errors

Rombout H. · Sprawdzono około 2 lata temu

does not work

Lukas D. · Sprawdzono ponad 2 lata temu

Lukas D. · Sprawdzono ponad 2 lata temu

Naufer N. · Sprawdzono ponad 2 lata temu

Naufer N. · Sprawdzono ponad 2 lata temu

Very buggy, workarounds must be completed in a certain order or the lab will fail.

Blake C. · Sprawdzono ponad 2 lata temu

Not understandable

Devpal S. · Sprawdzono ponad 2 lata temu

Asma B. · Sprawdzono ponad 2 lata temu

There were errors while running the notebook, so I could not create and run the pipeline: 1) None of the 4 Docker images were created successfully. 2) There is some problem with importing Iterable from collections - it seems like a Python version problem. In general this lab is ALMOST USELESS now, since the PIPELINE CANNOT BE CREATED AND EXECUTED, and the notebook can be found and accessed on GCP's official github page without paying.

Michał H. · Sprawdzono ponad 2 lata temu

Lab does not work! Jupyter notebook badly initiated and missing kernel... and most of the dpendencies are outdated which makes it fail.

gerald g. · Sprawdzono ponad 2 lata temu

Lab does not work! Jupyter notebook badly initiated and missing kernel... which makes it fail.

gerald g. · Sprawdzono ponad 2 lata temu

Lab does not work! Jupyter notebook badly initiated and missing kernel... which makes it fail.

gerald g. · Sprawdzono ponad 2 lata temu

Wanming H. · Sprawdzono ponad 2 lata temu

gerald g. · Sprawdzono ponad 2 lata temu

Wanming H. · Sprawdzono ponad 2 lata temu

Shriram G. · Sprawdzono ponad 2 lata temu

Salvatore V. · Sprawdzono ponad 2 lata temu

Notebook failing build

Christer D. · Sprawdzono ponad 2 lata temu

K S. · Sprawdzono ponad 2 lata temu

SOUMOJIT G. · Sprawdzono ponad 2 lata temu

Farhan S. · Sprawdzono ponad 2 lata temu

Rich B. · Sprawdzono ponad 2 lata temu

To make it work, I had to remove the library versions from each Dockerfile. After that, I added num_retries = 3 .set_retry(num_retries) after each .set_display_name, and update kfp: !pip install --upgrade kfp I did this for every artifact in the pipeline because at first, it would fail, but subsequent runs executed correctly. and change in scikit_trainer_image/train.py SGDClassifier(loss='log') -> SGDClassifier(loss='log_loss')

Carlos V. · Sprawdzono ponad 2 lata temu

To make it work, I had to remove the library versions from each Dockerfile. After that, I added num_retries = 3 .set_retry(num_retries, policy=RetryPolicy.ALWAYS) after each .set_display_name. I did this for every artifact in the pipeline because at first, it would fail, but subsequent runs executed correctly.

Carlos V. · Sprawdzono ponad 2 lata temu