Upload train_config.json with huggingface_hub
Browse files- train_config.json +123 -0
train_config.json
ADDED
|
@@ -0,0 +1,123 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"skip_load_state": 1,
|
| 3 |
+
"N_max": 100,
|
| 4 |
+
"batch_size": 256,
|
| 5 |
+
"beta_1": 0.9,
|
| 6 |
+
"beta_2": 0.999,
|
| 7 |
+
"comb_dropout": 0.7193785398638215,
|
| 8 |
+
"comb_fc1_neurons": 256,
|
| 9 |
+
"comb_fc2_neurons": 32,
|
| 10 |
+
"data_aug_h_flip": 1,
|
| 11 |
+
"data_aug_rot": 1,
|
| 12 |
+
"data_aug_v_flip": 1,
|
| 13 |
+
"epochs": 50,
|
| 14 |
+
"generate_embeddings": 0,
|
| 15 |
+
"image_model_dir": "models/MaxViT_v11_N100_cuda/gallant-sweep-16/",
|
| 16 |
+
"image_model_config": {
|
| 17 |
+
"N_max": 100,
|
| 18 |
+
"batch_size": 16,
|
| 19 |
+
"beta_1": 0.9,
|
| 20 |
+
"beta_2": 0.999,
|
| 21 |
+
"data_aug_h_flip": 1,
|
| 22 |
+
"data_aug_rot": 1,
|
| 23 |
+
"data_aug_v_flip": 1,
|
| 24 |
+
"dropout": 0.4595202732711804,
|
| 25 |
+
"epochs": 30,
|
| 26 |
+
"fc1_neurons": 128,
|
| 27 |
+
"fc2_neurons": 16,
|
| 28 |
+
"generate_embeddings": 0,
|
| 29 |
+
"learning_rate": 5e-06,
|
| 30 |
+
"loss": "binary_crossentropy",
|
| 31 |
+
"model_kind": "maxvit_tiny_rw_224.sw_in1k",
|
| 32 |
+
"model_name": "MaxViT",
|
| 33 |
+
"patience": 4,
|
| 34 |
+
"pretrained": 0,
|
| 35 |
+
"random_seed": 2,
|
| 36 |
+
"testing": 0,
|
| 37 |
+
"train_data_version": "v11",
|
| 38 |
+
"warmup_epochs": 4
|
| 39 |
+
},
|
| 40 |
+
"learning_rate": 5e-06,
|
| 41 |
+
"loss": "binary_crossentropy",
|
| 42 |
+
"meta_model_dir": "models/um_nn_v11_N100_cuda/jumping-sweep-20/",
|
| 43 |
+
"meta_model_config": {
|
| 44 |
+
"N_max": 100,
|
| 45 |
+
"batch_size": 256,
|
| 46 |
+
"beta_1": 0.9,
|
| 47 |
+
"beta_2": 0.999,
|
| 48 |
+
"epochs": 50,
|
| 49 |
+
"generate_embeddings": 0,
|
| 50 |
+
"learning_rate": 0.0005,
|
| 51 |
+
"loss": "binary_crossentropy",
|
| 52 |
+
"meta_dropout": 0.7182685711620065,
|
| 53 |
+
"meta_fc1_neurons": 128,
|
| 54 |
+
"meta_fc2_neurons": 8,
|
| 55 |
+
"metadata_cols": [
|
| 56 |
+
"sgscore1",
|
| 57 |
+
"distpsnr1",
|
| 58 |
+
"sgscore2",
|
| 59 |
+
"distpsnr2",
|
| 60 |
+
"fwhm",
|
| 61 |
+
"magpsf",
|
| 62 |
+
"sigmapsf",
|
| 63 |
+
"chipsf",
|
| 64 |
+
"ra",
|
| 65 |
+
"dec",
|
| 66 |
+
"diffmaglim",
|
| 67 |
+
"ndethist",
|
| 68 |
+
"nmtchps",
|
| 69 |
+
"age",
|
| 70 |
+
"days_since_peak",
|
| 71 |
+
"days_to_peak",
|
| 72 |
+
"peakmag_so_far",
|
| 73 |
+
"new_drb",
|
| 74 |
+
"ncovhist",
|
| 75 |
+
"nnotdet",
|
| 76 |
+
"chinr",
|
| 77 |
+
"sharpnr",
|
| 78 |
+
"scorr",
|
| 79 |
+
"sky",
|
| 80 |
+
"maxmag_so_far"
|
| 81 |
+
],
|
| 82 |
+
"model_name": "um_nn",
|
| 83 |
+
"patience": 4,
|
| 84 |
+
"random_seed": 2,
|
| 85 |
+
"testing": 0,
|
| 86 |
+
"train_data_version": "v11",
|
| 87 |
+
"warmup_epochs": 6
|
| 88 |
+
},
|
| 89 |
+
"metadata_cols": [
|
| 90 |
+
"sgscore1",
|
| 91 |
+
"distpsnr1",
|
| 92 |
+
"sgscore2",
|
| 93 |
+
"distpsnr2",
|
| 94 |
+
"fwhm",
|
| 95 |
+
"magpsf",
|
| 96 |
+
"sigmapsf",
|
| 97 |
+
"chipsf",
|
| 98 |
+
"ra",
|
| 99 |
+
"dec",
|
| 100 |
+
"diffmaglim",
|
| 101 |
+
"ndethist",
|
| 102 |
+
"nmtchps",
|
| 103 |
+
"age",
|
| 104 |
+
"days_since_peak",
|
| 105 |
+
"days_to_peak",
|
| 106 |
+
"peakmag_so_far",
|
| 107 |
+
"new_drb",
|
| 108 |
+
"ncovhist",
|
| 109 |
+
"nnotdet",
|
| 110 |
+
"chinr",
|
| 111 |
+
"sharpnr",
|
| 112 |
+
"scorr",
|
| 113 |
+
"sky",
|
| 114 |
+
"maxmag_so_far"
|
| 115 |
+
],
|
| 116 |
+
"model_name": "frozen_fusion",
|
| 117 |
+
"patience": 4,
|
| 118 |
+
"random_seed": 43,
|
| 119 |
+
"testing": 0,
|
| 120 |
+
"train_data_version": "v11",
|
| 121 |
+
"use_test_split": 1,
|
| 122 |
+
"warmup_epochs": 2
|
| 123 |
+
}
|