Registering Models at Covariant!
Instructions and Conventions for Getting Your Models to Production!
Created on August 30|Last edited on August 30
Comment
Team PrioritiesSummary of Models at CovariantSteps to Log and Register Your Models1. Create a Registered Model2. Log Model Checkpoints as Artifacts to W&B3. Link Your Best Model Version to the Registered ModelAlias Conventions at Covariant4. Model Card Documentation Requirements5. Retrieving Models from the Model Registry6. What happens to my model after I register it?
Team Priorities
-
Summary of Models at Covariant
The Covariant team is currently working on three primary projects:
- Object Detection
- PM: Carey
- MLE: Vish
- Depth Estimation
- PM: Tim
- MLE: Ken
- Image Super-resolution
- PM: Gourab
- MLE: Ken
Steps to Log and Register Your Models
1. Create a Registered Model
Go the Model Registry and click "Create Registered Model". Registered Models will house pointers to your model artifact versions which you log in your projects. Think of this as a bookmark folder for the best or most interesting model versions.
2. Log Model Checkpoints as Artifacts to W&B
At Singco we use Pytorch Lightning. Logging model artifacts is really easy with the W&B integration. See this code snippet below along with some reference links:
import wandbfrom pytorch_lightning.callbacks import ModelCheckpointfrom pytorch_lightning.loggers import WandbLoggerfrom pytorch_lightning import Trainerfrom pytorch_lightning import LightningModulewandb_logger = WandbLogger(project='MNIST', # group runs in "MNIST" projectlog_model='all') # log all new checkpoints during trainingcheckpoint_callback = ModelCheckpoint(monitor='val_accuracy', mode='max')trainer = Trainer(logger=wandb_logger, # W&B integrationcallbacks=[log_predictions_callback, # logging of sample predictionscheckpoint_callback], # our model checkpoint callbackmax_epochs=5) # number of epochstrainer.fit(model, training_loader, validation_loader). # model is a lightning module
3. Link Your Best Model Version to the Registered Model
After training a model, you can see all the artifact checkpoints logged under the artifacts tab of your project
Alias Conventions at Covariant
- staging should be assigned to your best model checkpoint after training
- production should be assigned...
4. Model Card Documentation Requirements
5. Retrieving Models from the Model Registry
6. What happens to my model after I register it?
Add a comment