pingsutw commented on a change in pull request #132: SUBMARINE-296. [SDK]
Submarine pipeline example
URL: https://github.com/apache/submarine/pull/132#discussion_r381076545
##########
File path: submarine-sdk/pysubmarine/example/deepFM.json
##########
@@ -0,0 +1,31 @@
+{
+ "input": {
+ "train_data": "./data/tr.libsvm",
+ "valid_data": "./data/va.libsvm",
+ "test_data": "./data/te.libsvm",
+ "type": "libsvm"
+ },
+ "output": {
+ "save_model_dir": "./experiment",
+ "metric": "auc"
+ },
+ "training": {
+ "batch_size" : 512,
+ "field_size": 39,
+ "num_epochs": 3,
+ "feature_size": 117581,
+ "embedding_size": 256,
+ "learning_rate": 0.0005,
+ "batch_norm_decay": 0.9,
+ "l2_reg": 0.0001,
+ "deep_layers": "400,400,400",
+ "dropout": "0.3,0.3,0.3",
+ "batch_norm": false,
+ "optimizer": "adam",
+ "log_steps": 10,
+ "num_threads": 4,
+ "num_gpu": 0,
+ "seed": 77,
+ "mode": "local"
Review comment:
> Can we reorg the `training` parameters into two parts: `algorithm` and
`system`. The `algorithm` part should be more friendly to data scientists and
machine learning engineers, and the `system` part would be more advanced.
Agree, it will be more user-friendly for users, I will update it.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
