create children class to PyTorchClassifier to implement the fit method where we initialize the trainer and model objects

This commit is contained in:
Yinon Polak
2023-03-19 14:38:49 +02:00
parent 566346dd87
commit 833aaf8e10
4 changed files with 146 additions and 87 deletions

View File

@@ -88,10 +88,12 @@ def test_extract_data_and_train_model_Standard(mocker, freqai_conf, model, pca,
if 'PyTorchClassifierMultiTarget' in model:
model_save_ext = 'zip'
freqai_conf['freqai']['model_training_parameters'].update({
"max_iters": 1,
"batch_size": 64,
"learning_rate": 3e-4,
"max_n_eval_batches": 1,
"trainer_kwargs": {
"max_iters": 1,
"batch_size": 64,
"max_n_eval_batches": 1,
},
"model_kwargs": {
"hidden_dim": 32,
"dropout_percent": 0.2,