Weight and Biases
- Easy to share the experiemnt with other people!
- can track hardware status
- install with pip
- make config file
config={"epochs":EPOCHS, "batch_size":BATCH_SIZE, "learning_rate":LEARNING_RATE}
# initialize. set entity name from web
wandb.init(project="project_name", entity="entity_name")
# after setting the configuartion dictionary
wandb.init(project="project_name", config=config)
# setting configuration
wandb.config.batch_size = BATCH_SIZE
wandb.config.learning_rate = LEARNING_RATE
- how to use in the model
for e in range(1, EPOCHS+1):
epoch_loss = 0
epoch_acc = 0
for X_batch, y_batch in train_dataset:
X_batch, y_batch = X_batch.to(device), y_batch.to(device).type(torch.cuda.FloatTensor)
...
...
# saving the log
wandb.log({'accuary':train_acc, 'loss':train_loss})