Upload language classifier cnn
Browse files- .gitattributes +1 -0
- .gitignore +2 -0
- description.json +888 -0
- model/llm_eval_examples/0.csv +1 -0
- model/llm_eval_examples/2.csv +1 -0
- model/llm_eval_examples/4.csv +1 -0
- model/llm_eval_examples/6.csv +1 -0
- model/llm_eval_examples/8.csv +1 -0
- model/model_hyperparameters.json +0 -0
- model/model_weights +3 -0
- model/training_progress.json +1461 -0
- model/training_set_metadata.json +0 -0
- training_statistics.json +129 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
model/model_weights filter=lfs diff=lfs merge=lfs -text
|
.gitignore
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model/logs/*
|
| 2 |
+
model/training_checkpoints/*
|
description.json
ADDED
|
@@ -0,0 +1,888 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"command": "/usr/local/bin/ludwig experiment --config /src/language-classifier.yaml --dataset /data/data.json --output_directory /results",
|
| 3 |
+
"compute": {
|
| 4 |
+
"num_nodes": 1
|
| 5 |
+
},
|
| 6 |
+
"config": {
|
| 7 |
+
"backend": null,
|
| 8 |
+
"combiner": {
|
| 9 |
+
"activation": "relu",
|
| 10 |
+
"bias_initializer": "zeros",
|
| 11 |
+
"dropout": 0.0,
|
| 12 |
+
"fc_layers": null,
|
| 13 |
+
"flatten_inputs": false,
|
| 14 |
+
"norm": null,
|
| 15 |
+
"norm_params": null,
|
| 16 |
+
"num_fc_layers": 0,
|
| 17 |
+
"output_size": 256,
|
| 18 |
+
"residual": false,
|
| 19 |
+
"type": "concat",
|
| 20 |
+
"use_bias": true,
|
| 21 |
+
"weights_initializer": "xavier_uniform"
|
| 22 |
+
},
|
| 23 |
+
"defaults": {
|
| 24 |
+
"audio": {
|
| 25 |
+
"encoder": {
|
| 26 |
+
"activation": "relu",
|
| 27 |
+
"bias_initializer": "zeros",
|
| 28 |
+
"conv_layers": null,
|
| 29 |
+
"dropout": 0.0,
|
| 30 |
+
"embedding_size": 256,
|
| 31 |
+
"embeddings_on_cpu": false,
|
| 32 |
+
"embeddings_trainable": true,
|
| 33 |
+
"fc_layers": null,
|
| 34 |
+
"filter_size": 3,
|
| 35 |
+
"max_sequence_length": null,
|
| 36 |
+
"norm": null,
|
| 37 |
+
"norm_params": null,
|
| 38 |
+
"num_conv_layers": null,
|
| 39 |
+
"num_fc_layers": null,
|
| 40 |
+
"num_filters": 256,
|
| 41 |
+
"output_size": 256,
|
| 42 |
+
"pool_function": "max",
|
| 43 |
+
"pool_size": null,
|
| 44 |
+
"pretrained_embeddings": null,
|
| 45 |
+
"reduce_output": "sum",
|
| 46 |
+
"representation": "dense",
|
| 47 |
+
"should_embed": true,
|
| 48 |
+
"skip": false,
|
| 49 |
+
"type": "parallel_cnn",
|
| 50 |
+
"use_bias": true,
|
| 51 |
+
"vocab": null,
|
| 52 |
+
"weights_initializer": "xavier_uniform"
|
| 53 |
+
},
|
| 54 |
+
"preprocessing": {
|
| 55 |
+
"audio_file_length_limit_in_s": 7.5,
|
| 56 |
+
"computed_fill_value": null,
|
| 57 |
+
"fill_value": null,
|
| 58 |
+
"in_memory": true,
|
| 59 |
+
"missing_value_strategy": "bfill",
|
| 60 |
+
"norm": null,
|
| 61 |
+
"num_fft_points": null,
|
| 62 |
+
"num_filter_bands": 80,
|
| 63 |
+
"padding_value": 0.0,
|
| 64 |
+
"type": "fbank",
|
| 65 |
+
"window_length_in_s": 0.04,
|
| 66 |
+
"window_shift_in_s": 0.02,
|
| 67 |
+
"window_type": "hamming"
|
| 68 |
+
}
|
| 69 |
+
},
|
| 70 |
+
"bag": {
|
| 71 |
+
"encoder": {
|
| 72 |
+
"activation": "relu",
|
| 73 |
+
"bias_initializer": "zeros",
|
| 74 |
+
"dropout": 0.0,
|
| 75 |
+
"embedding_size": 50,
|
| 76 |
+
"embeddings_on_cpu": false,
|
| 77 |
+
"embeddings_trainable": true,
|
| 78 |
+
"fc_layers": null,
|
| 79 |
+
"force_embedding_size": false,
|
| 80 |
+
"norm": null,
|
| 81 |
+
"norm_params": null,
|
| 82 |
+
"num_fc_layers": 0,
|
| 83 |
+
"output_size": 10,
|
| 84 |
+
"pretrained_embeddings": null,
|
| 85 |
+
"representation": "dense",
|
| 86 |
+
"skip": false,
|
| 87 |
+
"type": "embed",
|
| 88 |
+
"use_bias": true,
|
| 89 |
+
"vocab": null,
|
| 90 |
+
"weights_initializer": "xavier_uniform"
|
| 91 |
+
},
|
| 92 |
+
"preprocessing": {
|
| 93 |
+
"computed_fill_value": "<UNK>",
|
| 94 |
+
"fill_value": "<UNK>",
|
| 95 |
+
"lowercase": false,
|
| 96 |
+
"missing_value_strategy": "fill_with_const",
|
| 97 |
+
"most_common": 10000,
|
| 98 |
+
"tokenizer": "space"
|
| 99 |
+
}
|
| 100 |
+
},
|
| 101 |
+
"binary": {
|
| 102 |
+
"decoder": {
|
| 103 |
+
"bias_initializer": "zeros",
|
| 104 |
+
"fc_activation": "relu",
|
| 105 |
+
"fc_bias_initializer": "zeros",
|
| 106 |
+
"fc_dropout": 0.0,
|
| 107 |
+
"fc_layers": null,
|
| 108 |
+
"fc_norm": null,
|
| 109 |
+
"fc_norm_params": null,
|
| 110 |
+
"fc_output_size": 256,
|
| 111 |
+
"fc_use_bias": true,
|
| 112 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 113 |
+
"input_size": null,
|
| 114 |
+
"num_fc_layers": 0,
|
| 115 |
+
"type": "regressor",
|
| 116 |
+
"use_bias": true,
|
| 117 |
+
"weights_initializer": "xavier_uniform"
|
| 118 |
+
},
|
| 119 |
+
"encoder": {
|
| 120 |
+
"skip": false,
|
| 121 |
+
"type": "passthrough"
|
| 122 |
+
},
|
| 123 |
+
"loss": {
|
| 124 |
+
"confidence_penalty": 0,
|
| 125 |
+
"positive_class_weight": null,
|
| 126 |
+
"robust_lambda": 0,
|
| 127 |
+
"type": "binary_weighted_cross_entropy",
|
| 128 |
+
"weight": 1.0
|
| 129 |
+
},
|
| 130 |
+
"preprocessing": {
|
| 131 |
+
"computed_fill_value": null,
|
| 132 |
+
"fallback_true_label": null,
|
| 133 |
+
"fill_value": null,
|
| 134 |
+
"missing_value_strategy": "fill_with_false"
|
| 135 |
+
}
|
| 136 |
+
},
|
| 137 |
+
"category": {
|
| 138 |
+
"decoder": {
|
| 139 |
+
"bias_initializer": "zeros",
|
| 140 |
+
"fc_activation": "relu",
|
| 141 |
+
"fc_bias_initializer": "zeros",
|
| 142 |
+
"fc_dropout": 0.0,
|
| 143 |
+
"fc_layers": null,
|
| 144 |
+
"fc_norm": null,
|
| 145 |
+
"fc_norm_params": null,
|
| 146 |
+
"fc_output_size": 256,
|
| 147 |
+
"fc_use_bias": true,
|
| 148 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 149 |
+
"input_size": null,
|
| 150 |
+
"num_classes": null,
|
| 151 |
+
"num_fc_layers": 0,
|
| 152 |
+
"type": "classifier",
|
| 153 |
+
"use_bias": true,
|
| 154 |
+
"weights_initializer": "xavier_uniform"
|
| 155 |
+
},
|
| 156 |
+
"encoder": {
|
| 157 |
+
"dropout": 0.0,
|
| 158 |
+
"embedding_initializer": null,
|
| 159 |
+
"embedding_size": 50,
|
| 160 |
+
"embeddings_on_cpu": false,
|
| 161 |
+
"embeddings_trainable": true,
|
| 162 |
+
"pretrained_embeddings": null,
|
| 163 |
+
"skip": false,
|
| 164 |
+
"type": "dense",
|
| 165 |
+
"vocab": null
|
| 166 |
+
},
|
| 167 |
+
"loss": {
|
| 168 |
+
"class_similarities": null,
|
| 169 |
+
"class_similarities_temperature": 0,
|
| 170 |
+
"class_weights": null,
|
| 171 |
+
"confidence_penalty": 0,
|
| 172 |
+
"robust_lambda": 0,
|
| 173 |
+
"type": "softmax_cross_entropy",
|
| 174 |
+
"weight": 1.0
|
| 175 |
+
},
|
| 176 |
+
"preprocessing": {
|
| 177 |
+
"cache_encoder_embeddings": false,
|
| 178 |
+
"computed_fill_value": "<UNK>",
|
| 179 |
+
"fill_value": "<UNK>",
|
| 180 |
+
"lowercase": false,
|
| 181 |
+
"missing_value_strategy": "fill_with_const",
|
| 182 |
+
"most_common": 10000
|
| 183 |
+
}
|
| 184 |
+
},
|
| 185 |
+
"date": {
|
| 186 |
+
"encoder": {
|
| 187 |
+
"activation": "relu",
|
| 188 |
+
"bias_initializer": "zeros",
|
| 189 |
+
"dropout": 0.0,
|
| 190 |
+
"embedding_size": 10,
|
| 191 |
+
"embeddings_on_cpu": false,
|
| 192 |
+
"fc_layers": null,
|
| 193 |
+
"norm": null,
|
| 194 |
+
"norm_params": null,
|
| 195 |
+
"num_fc_layers": 0,
|
| 196 |
+
"output_size": 10,
|
| 197 |
+
"skip": false,
|
| 198 |
+
"type": "embed",
|
| 199 |
+
"use_bias": true,
|
| 200 |
+
"weights_initializer": "xavier_uniform"
|
| 201 |
+
},
|
| 202 |
+
"preprocessing": {
|
| 203 |
+
"computed_fill_value": "",
|
| 204 |
+
"datetime_format": null,
|
| 205 |
+
"fill_value": "",
|
| 206 |
+
"missing_value_strategy": "fill_with_const"
|
| 207 |
+
}
|
| 208 |
+
},
|
| 209 |
+
"h3": {
|
| 210 |
+
"encoder": {
|
| 211 |
+
"activation": "relu",
|
| 212 |
+
"bias_initializer": "zeros",
|
| 213 |
+
"dropout": 0.0,
|
| 214 |
+
"embedding_size": 10,
|
| 215 |
+
"embeddings_on_cpu": false,
|
| 216 |
+
"fc_layers": null,
|
| 217 |
+
"norm": null,
|
| 218 |
+
"norm_params": null,
|
| 219 |
+
"num_fc_layers": 0,
|
| 220 |
+
"output_size": 10,
|
| 221 |
+
"reduce_output": "sum",
|
| 222 |
+
"skip": false,
|
| 223 |
+
"type": "embed",
|
| 224 |
+
"use_bias": true,
|
| 225 |
+
"weights_initializer": "xavier_uniform"
|
| 226 |
+
},
|
| 227 |
+
"preprocessing": {
|
| 228 |
+
"computed_fill_value": 576495936675512319,
|
| 229 |
+
"fill_value": 576495936675512319,
|
| 230 |
+
"missing_value_strategy": "fill_with_const"
|
| 231 |
+
}
|
| 232 |
+
},
|
| 233 |
+
"image": {
|
| 234 |
+
"augmentation": [],
|
| 235 |
+
"decoder": {
|
| 236 |
+
"conv_norm": "batch",
|
| 237 |
+
"fc_activation": "relu",
|
| 238 |
+
"fc_bias_initializer": "zeros",
|
| 239 |
+
"fc_dropout": 0.0,
|
| 240 |
+
"fc_layers": null,
|
| 241 |
+
"fc_norm": null,
|
| 242 |
+
"fc_norm_params": null,
|
| 243 |
+
"fc_output_size": 256,
|
| 244 |
+
"fc_use_bias": true,
|
| 245 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 246 |
+
"height": null,
|
| 247 |
+
"input_size": 1024,
|
| 248 |
+
"num_channels": null,
|
| 249 |
+
"num_classes": null,
|
| 250 |
+
"num_fc_layers": 0,
|
| 251 |
+
"type": "unet",
|
| 252 |
+
"width": null
|
| 253 |
+
},
|
| 254 |
+
"encoder": {
|
| 255 |
+
"conv_activation": "relu",
|
| 256 |
+
"conv_dropout": 0.0,
|
| 257 |
+
"conv_layers": null,
|
| 258 |
+
"conv_norm": null,
|
| 259 |
+
"conv_norm_params": null,
|
| 260 |
+
"conv_use_bias": true,
|
| 261 |
+
"dilation": 1,
|
| 262 |
+
"fc_activation": "relu",
|
| 263 |
+
"fc_bias_initializer": "zeros",
|
| 264 |
+
"fc_dropout": 0.0,
|
| 265 |
+
"fc_layers": null,
|
| 266 |
+
"fc_norm": null,
|
| 267 |
+
"fc_norm_params": null,
|
| 268 |
+
"fc_use_bias": true,
|
| 269 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 270 |
+
"groups": 1,
|
| 271 |
+
"height": null,
|
| 272 |
+
"kernel_size": 3,
|
| 273 |
+
"num_channels": null,
|
| 274 |
+
"num_conv_layers": null,
|
| 275 |
+
"num_fc_layers": 1,
|
| 276 |
+
"out_channels": 32,
|
| 277 |
+
"output_size": 128,
|
| 278 |
+
"padding": "valid",
|
| 279 |
+
"padding_mode": "zeros",
|
| 280 |
+
"pool_dilation": 1,
|
| 281 |
+
"pool_function": "max",
|
| 282 |
+
"pool_kernel_size": 2,
|
| 283 |
+
"pool_padding": 0,
|
| 284 |
+
"pool_stride": null,
|
| 285 |
+
"skip": false,
|
| 286 |
+
"stride": 1,
|
| 287 |
+
"type": "stacked_cnn",
|
| 288 |
+
"width": null
|
| 289 |
+
},
|
| 290 |
+
"loss": {
|
| 291 |
+
"class_similarities": null,
|
| 292 |
+
"class_similarities_temperature": 0,
|
| 293 |
+
"class_weights": null,
|
| 294 |
+
"confidence_penalty": 0,
|
| 295 |
+
"robust_lambda": 0,
|
| 296 |
+
"type": "softmax_cross_entropy",
|
| 297 |
+
"weight": 1.0
|
| 298 |
+
},
|
| 299 |
+
"preprocessing": {
|
| 300 |
+
"computed_fill_value": null,
|
| 301 |
+
"fill_value": null,
|
| 302 |
+
"height": null,
|
| 303 |
+
"in_memory": true,
|
| 304 |
+
"infer_image_dimensions": true,
|
| 305 |
+
"infer_image_max_height": 256,
|
| 306 |
+
"infer_image_max_width": 256,
|
| 307 |
+
"infer_image_num_channels": true,
|
| 308 |
+
"infer_image_num_classes": false,
|
| 309 |
+
"infer_image_sample_size": 100,
|
| 310 |
+
"missing_value_strategy": "bfill",
|
| 311 |
+
"num_channels": null,
|
| 312 |
+
"num_classes": null,
|
| 313 |
+
"num_processes": 1,
|
| 314 |
+
"requires_equal_dimensions": false,
|
| 315 |
+
"resize_method": "interpolate",
|
| 316 |
+
"standardize_image": null,
|
| 317 |
+
"width": null
|
| 318 |
+
}
|
| 319 |
+
},
|
| 320 |
+
"number": {
|
| 321 |
+
"decoder": {
|
| 322 |
+
"bias_initializer": "zeros",
|
| 323 |
+
"fc_activation": "relu",
|
| 324 |
+
"fc_bias_initializer": "zeros",
|
| 325 |
+
"fc_dropout": 0.0,
|
| 326 |
+
"fc_layers": null,
|
| 327 |
+
"fc_norm": null,
|
| 328 |
+
"fc_norm_params": null,
|
| 329 |
+
"fc_output_size": 256,
|
| 330 |
+
"fc_use_bias": true,
|
| 331 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 332 |
+
"input_size": null,
|
| 333 |
+
"num_fc_layers": 0,
|
| 334 |
+
"type": "regressor",
|
| 335 |
+
"use_bias": true,
|
| 336 |
+
"weights_initializer": "xavier_uniform"
|
| 337 |
+
},
|
| 338 |
+
"encoder": {
|
| 339 |
+
"skip": false,
|
| 340 |
+
"type": "passthrough"
|
| 341 |
+
},
|
| 342 |
+
"loss": {
|
| 343 |
+
"type": "mean_squared_error",
|
| 344 |
+
"weight": 1.0
|
| 345 |
+
},
|
| 346 |
+
"preprocessing": {
|
| 347 |
+
"computed_fill_value": 0.0,
|
| 348 |
+
"computed_outlier_fill_value": 0.0,
|
| 349 |
+
"fill_value": 0.0,
|
| 350 |
+
"missing_value_strategy": "fill_with_const",
|
| 351 |
+
"normalization": "zscore",
|
| 352 |
+
"outlier_strategy": null,
|
| 353 |
+
"outlier_threshold": 3.0
|
| 354 |
+
}
|
| 355 |
+
},
|
| 356 |
+
"sequence": {
|
| 357 |
+
"decoder": {
|
| 358 |
+
"cell_type": "gru",
|
| 359 |
+
"fc_activation": "relu",
|
| 360 |
+
"fc_bias_initializer": "zeros",
|
| 361 |
+
"fc_dropout": 0.0,
|
| 362 |
+
"fc_layers": null,
|
| 363 |
+
"fc_norm": null,
|
| 364 |
+
"fc_norm_params": null,
|
| 365 |
+
"fc_output_size": 256,
|
| 366 |
+
"fc_use_bias": true,
|
| 367 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 368 |
+
"input_size": 256,
|
| 369 |
+
"max_sequence_length": null,
|
| 370 |
+
"num_fc_layers": 0,
|
| 371 |
+
"num_layers": 1,
|
| 372 |
+
"reduce_input": "sum",
|
| 373 |
+
"type": "generator",
|
| 374 |
+
"vocab_size": null
|
| 375 |
+
},
|
| 376 |
+
"encoder": {
|
| 377 |
+
"dropout": 0.0,
|
| 378 |
+
"embedding_size": 256,
|
| 379 |
+
"embeddings_on_cpu": false,
|
| 380 |
+
"embeddings_trainable": true,
|
| 381 |
+
"max_sequence_length": null,
|
| 382 |
+
"pretrained_embeddings": null,
|
| 383 |
+
"reduce_output": "sum",
|
| 384 |
+
"representation": "dense",
|
| 385 |
+
"skip": false,
|
| 386 |
+
"type": "embed",
|
| 387 |
+
"vocab": null,
|
| 388 |
+
"weights_initializer": "uniform"
|
| 389 |
+
},
|
| 390 |
+
"loss": {
|
| 391 |
+
"class_similarities": null,
|
| 392 |
+
"class_similarities_temperature": 0,
|
| 393 |
+
"class_weights": null,
|
| 394 |
+
"confidence_penalty": 0,
|
| 395 |
+
"robust_lambda": 0,
|
| 396 |
+
"type": "sequence_softmax_cross_entropy",
|
| 397 |
+
"unique": false,
|
| 398 |
+
"weight": 1.0
|
| 399 |
+
},
|
| 400 |
+
"preprocessing": {
|
| 401 |
+
"cache_encoder_embeddings": false,
|
| 402 |
+
"computed_fill_value": "<UNK>",
|
| 403 |
+
"fill_value": "<UNK>",
|
| 404 |
+
"lowercase": false,
|
| 405 |
+
"max_sequence_length": 256,
|
| 406 |
+
"missing_value_strategy": "fill_with_const",
|
| 407 |
+
"most_common": 20000,
|
| 408 |
+
"ngram_size": 2,
|
| 409 |
+
"padding": "right",
|
| 410 |
+
"padding_symbol": "<PAD>",
|
| 411 |
+
"sequence_length": null,
|
| 412 |
+
"tokenizer": "space",
|
| 413 |
+
"unknown_symbol": "<UNK>",
|
| 414 |
+
"vocab_file": null
|
| 415 |
+
}
|
| 416 |
+
},
|
| 417 |
+
"set": {
|
| 418 |
+
"decoder": {
|
| 419 |
+
"bias_initializer": "zeros",
|
| 420 |
+
"fc_activation": "relu",
|
| 421 |
+
"fc_bias_initializer": "zeros",
|
| 422 |
+
"fc_dropout": 0.0,
|
| 423 |
+
"fc_layers": null,
|
| 424 |
+
"fc_norm": null,
|
| 425 |
+
"fc_norm_params": null,
|
| 426 |
+
"fc_output_size": 256,
|
| 427 |
+
"fc_use_bias": true,
|
| 428 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 429 |
+
"input_size": null,
|
| 430 |
+
"num_classes": null,
|
| 431 |
+
"num_fc_layers": 0,
|
| 432 |
+
"type": "classifier",
|
| 433 |
+
"use_bias": true,
|
| 434 |
+
"weights_initializer": "xavier_uniform"
|
| 435 |
+
},
|
| 436 |
+
"encoder": {
|
| 437 |
+
"activation": "relu",
|
| 438 |
+
"bias_initializer": "zeros",
|
| 439 |
+
"dropout": 0.0,
|
| 440 |
+
"embedding_size": 50,
|
| 441 |
+
"embeddings_on_cpu": false,
|
| 442 |
+
"embeddings_trainable": true,
|
| 443 |
+
"fc_layers": null,
|
| 444 |
+
"norm": null,
|
| 445 |
+
"norm_params": null,
|
| 446 |
+
"num_fc_layers": 0,
|
| 447 |
+
"output_size": 10,
|
| 448 |
+
"pretrained_embeddings": null,
|
| 449 |
+
"representation": "dense",
|
| 450 |
+
"skip": false,
|
| 451 |
+
"type": "embed",
|
| 452 |
+
"use_bias": true,
|
| 453 |
+
"vocab": null,
|
| 454 |
+
"weights_initializer": "xavier_uniform"
|
| 455 |
+
},
|
| 456 |
+
"loss": {
|
| 457 |
+
"class_weights": null,
|
| 458 |
+
"type": "sigmoid_cross_entropy",
|
| 459 |
+
"weight": 1.0
|
| 460 |
+
},
|
| 461 |
+
"preprocessing": {
|
| 462 |
+
"computed_fill_value": "<UNK>",
|
| 463 |
+
"fill_value": "<UNK>",
|
| 464 |
+
"lowercase": false,
|
| 465 |
+
"missing_value_strategy": "fill_with_const",
|
| 466 |
+
"most_common": 10000,
|
| 467 |
+
"tokenizer": "space"
|
| 468 |
+
}
|
| 469 |
+
},
|
| 470 |
+
"text": {
|
| 471 |
+
"decoder": {
|
| 472 |
+
"cell_type": "gru",
|
| 473 |
+
"fc_activation": "relu",
|
| 474 |
+
"fc_bias_initializer": "zeros",
|
| 475 |
+
"fc_dropout": 0.0,
|
| 476 |
+
"fc_layers": null,
|
| 477 |
+
"fc_norm": null,
|
| 478 |
+
"fc_norm_params": null,
|
| 479 |
+
"fc_output_size": 256,
|
| 480 |
+
"fc_use_bias": true,
|
| 481 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 482 |
+
"input_size": 256,
|
| 483 |
+
"max_sequence_length": null,
|
| 484 |
+
"num_fc_layers": 0,
|
| 485 |
+
"num_layers": 1,
|
| 486 |
+
"reduce_input": "sum",
|
| 487 |
+
"type": "generator",
|
| 488 |
+
"vocab_size": null
|
| 489 |
+
},
|
| 490 |
+
"encoder": {
|
| 491 |
+
"activation": "relu",
|
| 492 |
+
"bias_initializer": "zeros",
|
| 493 |
+
"conv_layers": null,
|
| 494 |
+
"dropout": 0.0,
|
| 495 |
+
"embedding_size": 256,
|
| 496 |
+
"embeddings_on_cpu": false,
|
| 497 |
+
"embeddings_trainable": true,
|
| 498 |
+
"fc_layers": null,
|
| 499 |
+
"filter_size": 3,
|
| 500 |
+
"max_sequence_length": null,
|
| 501 |
+
"norm": null,
|
| 502 |
+
"norm_params": null,
|
| 503 |
+
"num_conv_layers": null,
|
| 504 |
+
"num_fc_layers": null,
|
| 505 |
+
"num_filters": 256,
|
| 506 |
+
"output_size": 256,
|
| 507 |
+
"pool_function": "max",
|
| 508 |
+
"pool_size": null,
|
| 509 |
+
"pretrained_embeddings": null,
|
| 510 |
+
"reduce_output": "sum",
|
| 511 |
+
"representation": "dense",
|
| 512 |
+
"should_embed": true,
|
| 513 |
+
"skip": false,
|
| 514 |
+
"type": "parallel_cnn",
|
| 515 |
+
"use_bias": true,
|
| 516 |
+
"vocab": null,
|
| 517 |
+
"weights_initializer": "xavier_uniform"
|
| 518 |
+
},
|
| 519 |
+
"loss": {
|
| 520 |
+
"class_similarities": null,
|
| 521 |
+
"class_similarities_temperature": 0,
|
| 522 |
+
"class_weights": null,
|
| 523 |
+
"confidence_penalty": 0,
|
| 524 |
+
"robust_lambda": 0,
|
| 525 |
+
"type": "sequence_softmax_cross_entropy",
|
| 526 |
+
"unique": false,
|
| 527 |
+
"weight": 1.0
|
| 528 |
+
},
|
| 529 |
+
"preprocessing": {
|
| 530 |
+
"cache_encoder_embeddings": false,
|
| 531 |
+
"compute_idf": false,
|
| 532 |
+
"computed_fill_value": "<UNK>",
|
| 533 |
+
"fill_value": "<UNK>",
|
| 534 |
+
"lowercase": false,
|
| 535 |
+
"max_sequence_length": 256,
|
| 536 |
+
"missing_value_strategy": "fill_with_const",
|
| 537 |
+
"most_common": 20000,
|
| 538 |
+
"ngram_size": 2,
|
| 539 |
+
"padding": "right",
|
| 540 |
+
"padding_symbol": "<PAD>",
|
| 541 |
+
"pretrained_model_name_or_path": null,
|
| 542 |
+
"prompt": {
|
| 543 |
+
"retrieval": {
|
| 544 |
+
"index_name": null,
|
| 545 |
+
"k": 0,
|
| 546 |
+
"model_name": null,
|
| 547 |
+
"type": null
|
| 548 |
+
},
|
| 549 |
+
"task": null,
|
| 550 |
+
"template": null
|
| 551 |
+
},
|
| 552 |
+
"sequence_length": null,
|
| 553 |
+
"tokenizer": "space_punct",
|
| 554 |
+
"unknown_symbol": "<UNK>",
|
| 555 |
+
"vocab_file": null
|
| 556 |
+
}
|
| 557 |
+
},
|
| 558 |
+
"timeseries": {
|
| 559 |
+
"decoder": {
|
| 560 |
+
"activation": null,
|
| 561 |
+
"bias_initializer": "zeros",
|
| 562 |
+
"clip": null,
|
| 563 |
+
"fc_activation": "relu",
|
| 564 |
+
"fc_bias_initializer": "zeros",
|
| 565 |
+
"fc_dropout": 0.0,
|
| 566 |
+
"fc_layers": null,
|
| 567 |
+
"fc_norm": null,
|
| 568 |
+
"fc_norm_params": null,
|
| 569 |
+
"fc_output_size": 256,
|
| 570 |
+
"fc_use_bias": true,
|
| 571 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 572 |
+
"input_size": null,
|
| 573 |
+
"multiplier": 1.0,
|
| 574 |
+
"num_fc_layers": 0,
|
| 575 |
+
"output_size": null,
|
| 576 |
+
"type": "projector",
|
| 577 |
+
"use_bias": true,
|
| 578 |
+
"weights_initializer": "xavier_uniform"
|
| 579 |
+
},
|
| 580 |
+
"encoder": {
|
| 581 |
+
"activation": "relu",
|
| 582 |
+
"bias_initializer": "zeros",
|
| 583 |
+
"conv_layers": null,
|
| 584 |
+
"dropout": 0.0,
|
| 585 |
+
"embedding_size": 256,
|
| 586 |
+
"embeddings_on_cpu": false,
|
| 587 |
+
"embeddings_trainable": true,
|
| 588 |
+
"fc_layers": null,
|
| 589 |
+
"filter_size": 3,
|
| 590 |
+
"max_sequence_length": null,
|
| 591 |
+
"norm": null,
|
| 592 |
+
"norm_params": null,
|
| 593 |
+
"num_conv_layers": null,
|
| 594 |
+
"num_fc_layers": null,
|
| 595 |
+
"num_filters": 256,
|
| 596 |
+
"output_size": 256,
|
| 597 |
+
"pool_function": "max",
|
| 598 |
+
"pool_size": null,
|
| 599 |
+
"pretrained_embeddings": null,
|
| 600 |
+
"reduce_output": "sum",
|
| 601 |
+
"representation": "dense",
|
| 602 |
+
"should_embed": true,
|
| 603 |
+
"skip": false,
|
| 604 |
+
"type": "parallel_cnn",
|
| 605 |
+
"use_bias": true,
|
| 606 |
+
"vocab": null,
|
| 607 |
+
"weights_initializer": "xavier_uniform"
|
| 608 |
+
},
|
| 609 |
+
"loss": {
|
| 610 |
+
"delta": 1.0,
|
| 611 |
+
"type": "huber",
|
| 612 |
+
"weight": 1.0
|
| 613 |
+
},
|
| 614 |
+
"preprocessing": {
|
| 615 |
+
"computed_fill_value": "",
|
| 616 |
+
"fill_value": "",
|
| 617 |
+
"missing_value_strategy": "fill_with_const",
|
| 618 |
+
"padding": "right",
|
| 619 |
+
"padding_value": 0.0,
|
| 620 |
+
"timeseries_length_limit": 256,
|
| 621 |
+
"tokenizer": "space",
|
| 622 |
+
"window_size": 0
|
| 623 |
+
}
|
| 624 |
+
},
|
| 625 |
+
"vector": {
|
| 626 |
+
"decoder": {
|
| 627 |
+
"activation": null,
|
| 628 |
+
"bias_initializer": "zeros",
|
| 629 |
+
"clip": null,
|
| 630 |
+
"fc_activation": "relu",
|
| 631 |
+
"fc_bias_initializer": "zeros",
|
| 632 |
+
"fc_dropout": 0.0,
|
| 633 |
+
"fc_layers": null,
|
| 634 |
+
"fc_norm": null,
|
| 635 |
+
"fc_norm_params": null,
|
| 636 |
+
"fc_output_size": 256,
|
| 637 |
+
"fc_use_bias": true,
|
| 638 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 639 |
+
"input_size": null,
|
| 640 |
+
"multiplier": 1.0,
|
| 641 |
+
"num_fc_layers": 0,
|
| 642 |
+
"output_size": null,
|
| 643 |
+
"type": "projector",
|
| 644 |
+
"use_bias": true,
|
| 645 |
+
"weights_initializer": "xavier_uniform"
|
| 646 |
+
},
|
| 647 |
+
"encoder": {
|
| 648 |
+
"activation": "relu",
|
| 649 |
+
"bias_initializer": "zeros",
|
| 650 |
+
"dropout": 0.0,
|
| 651 |
+
"fc_layers": null,
|
| 652 |
+
"input_size": null,
|
| 653 |
+
"norm": null,
|
| 654 |
+
"norm_params": null,
|
| 655 |
+
"num_layers": 1,
|
| 656 |
+
"output_size": 256,
|
| 657 |
+
"skip": false,
|
| 658 |
+
"type": "dense",
|
| 659 |
+
"use_bias": true,
|
| 660 |
+
"weights_initializer": "xavier_uniform"
|
| 661 |
+
},
|
| 662 |
+
"loss": {
|
| 663 |
+
"type": "mean_squared_error",
|
| 664 |
+
"weight": 1.0
|
| 665 |
+
},
|
| 666 |
+
"preprocessing": {
|
| 667 |
+
"computed_fill_value": "",
|
| 668 |
+
"fill_value": "",
|
| 669 |
+
"missing_value_strategy": "fill_with_const",
|
| 670 |
+
"vector_size": null
|
| 671 |
+
}
|
| 672 |
+
}
|
| 673 |
+
},
|
| 674 |
+
"hyperopt": null,
|
| 675 |
+
"input_features": [
|
| 676 |
+
{
|
| 677 |
+
"active": true,
|
| 678 |
+
"column": "text",
|
| 679 |
+
"encoder": {
|
| 680 |
+
"activation": "relu",
|
| 681 |
+
"bias_initializer": "zeros",
|
| 682 |
+
"conv_layers": null,
|
| 683 |
+
"dropout": 0.0,
|
| 684 |
+
"embedding_size": 256,
|
| 685 |
+
"embeddings_on_cpu": false,
|
| 686 |
+
"embeddings_trainable": true,
|
| 687 |
+
"fc_layers": null,
|
| 688 |
+
"filter_size": 3,
|
| 689 |
+
"max_sequence_length": null,
|
| 690 |
+
"norm": null,
|
| 691 |
+
"norm_params": null,
|
| 692 |
+
"num_conv_layers": null,
|
| 693 |
+
"num_fc_layers": null,
|
| 694 |
+
"num_filters": 256,
|
| 695 |
+
"output_size": 256,
|
| 696 |
+
"pool_function": "max",
|
| 697 |
+
"pool_size": null,
|
| 698 |
+
"pretrained_embeddings": null,
|
| 699 |
+
"reduce_output": "sum",
|
| 700 |
+
"representation": "dense",
|
| 701 |
+
"should_embed": true,
|
| 702 |
+
"skip": false,
|
| 703 |
+
"type": "parallel_cnn",
|
| 704 |
+
"use_bias": true,
|
| 705 |
+
"vocab": null,
|
| 706 |
+
"weights_initializer": "xavier_uniform"
|
| 707 |
+
},
|
| 708 |
+
"name": "text",
|
| 709 |
+
"preprocessing": {
|
| 710 |
+
"cache_encoder_embeddings": false,
|
| 711 |
+
"compute_idf": false,
|
| 712 |
+
"computed_fill_value": "<UNK>",
|
| 713 |
+
"fill_value": "<UNK>",
|
| 714 |
+
"lowercase": true,
|
| 715 |
+
"max_sequence_length": 512,
|
| 716 |
+
"missing_value_strategy": "fill_with_const",
|
| 717 |
+
"most_common": 20000,
|
| 718 |
+
"ngram_size": 2,
|
| 719 |
+
"padding": "right",
|
| 720 |
+
"padding_symbol": "<PAD>",
|
| 721 |
+
"pretrained_model_name_or_path": null,
|
| 722 |
+
"prompt": {
|
| 723 |
+
"retrieval": {
|
| 724 |
+
"index_name": null,
|
| 725 |
+
"k": 0,
|
| 726 |
+
"model_name": null,
|
| 727 |
+
"type": null
|
| 728 |
+
},
|
| 729 |
+
"task": null,
|
| 730 |
+
"template": null
|
| 731 |
+
},
|
| 732 |
+
"sequence_length": null,
|
| 733 |
+
"tokenizer": "ngram",
|
| 734 |
+
"unknown_symbol": "<UNK>",
|
| 735 |
+
"vocab_file": null
|
| 736 |
+
},
|
| 737 |
+
"proc_column": "text_VQBTaW",
|
| 738 |
+
"tied": null,
|
| 739 |
+
"type": "text"
|
| 740 |
+
}
|
| 741 |
+
],
|
| 742 |
+
"ludwig_version": "0.10.2.dev",
|
| 743 |
+
"model_type": "ecd",
|
| 744 |
+
"output_features": [
|
| 745 |
+
{
|
| 746 |
+
"active": true,
|
| 747 |
+
"calibration": false,
|
| 748 |
+
"column": "language",
|
| 749 |
+
"decoder": {
|
| 750 |
+
"bias_initializer": "zeros",
|
| 751 |
+
"fc_activation": "relu",
|
| 752 |
+
"fc_bias_initializer": "zeros",
|
| 753 |
+
"fc_dropout": 0.0,
|
| 754 |
+
"fc_layers": null,
|
| 755 |
+
"fc_norm": null,
|
| 756 |
+
"fc_norm_params": null,
|
| 757 |
+
"fc_output_size": 256,
|
| 758 |
+
"fc_use_bias": true,
|
| 759 |
+
"fc_weights_initializer": "xavier_uniform",
|
| 760 |
+
"input_size": null,
|
| 761 |
+
"num_classes": null,
|
| 762 |
+
"num_fc_layers": 0,
|
| 763 |
+
"type": "classifier",
|
| 764 |
+
"use_bias": true,
|
| 765 |
+
"weights_initializer": "xavier_uniform"
|
| 766 |
+
},
|
| 767 |
+
"default_validation_metric": "accuracy",
|
| 768 |
+
"dependencies": [],
|
| 769 |
+
"input_size": null,
|
| 770 |
+
"loss": {
|
| 771 |
+
"class_similarities": null,
|
| 772 |
+
"class_similarities_temperature": 0,
|
| 773 |
+
"class_weights": null,
|
| 774 |
+
"confidence_penalty": 0,
|
| 775 |
+
"robust_lambda": 0,
|
| 776 |
+
"type": "softmax_cross_entropy",
|
| 777 |
+
"weight": 1.0
|
| 778 |
+
},
|
| 779 |
+
"name": "language",
|
| 780 |
+
"num_classes": null,
|
| 781 |
+
"preprocessing": {
|
| 782 |
+
"cache_encoder_embeddings": false,
|
| 783 |
+
"computed_fill_value": "<UNK>",
|
| 784 |
+
"fill_value": "<UNK>",
|
| 785 |
+
"lowercase": false,
|
| 786 |
+
"missing_value_strategy": "drop_row",
|
| 787 |
+
"most_common": 10000
|
| 788 |
+
},
|
| 789 |
+
"proc_column": "language_YwJjWN",
|
| 790 |
+
"reduce_dependencies": "sum",
|
| 791 |
+
"reduce_input": "sum",
|
| 792 |
+
"top_k": 3,
|
| 793 |
+
"type": "category"
|
| 794 |
+
}
|
| 795 |
+
],
|
| 796 |
+
"preprocessing": {
|
| 797 |
+
"global_max_sequence_length": null,
|
| 798 |
+
"oversample_minority": null,
|
| 799 |
+
"sample_ratio": 1.0,
|
| 800 |
+
"sample_size": null,
|
| 801 |
+
"split": {
|
| 802 |
+
"column": "language",
|
| 803 |
+
"probabilities": [
|
| 804 |
+
0.85,
|
| 805 |
+
0.1,
|
| 806 |
+
0.05
|
| 807 |
+
],
|
| 808 |
+
"type": "stratify"
|
| 809 |
+
},
|
| 810 |
+
"undersample_majority": null
|
| 811 |
+
},
|
| 812 |
+
"trainer": {
|
| 813 |
+
"batch_size": "auto",
|
| 814 |
+
"bucketing_field": null,
|
| 815 |
+
"checkpoints_per_epoch": 0,
|
| 816 |
+
"compile": false,
|
| 817 |
+
"early_stop": 0,
|
| 818 |
+
"effective_batch_size": "auto",
|
| 819 |
+
"enable_gradient_checkpointing": false,
|
| 820 |
+
"enable_profiling": false,
|
| 821 |
+
"epochs": 5,
|
| 822 |
+
"eval_batch_size": null,
|
| 823 |
+
"eval_steps": null,
|
| 824 |
+
"evaluate_training_set": false,
|
| 825 |
+
"gradient_accumulation_steps": "auto",
|
| 826 |
+
"gradient_clipping": {
|
| 827 |
+
"clipglobalnorm": 0.5,
|
| 828 |
+
"clipnorm": null,
|
| 829 |
+
"clipvalue": null
|
| 830 |
+
},
|
| 831 |
+
"increase_batch_size_eval_metric": "loss",
|
| 832 |
+
"increase_batch_size_eval_split": "training",
|
| 833 |
+
"increase_batch_size_on_plateau": 0,
|
| 834 |
+
"increase_batch_size_on_plateau_patience": 5,
|
| 835 |
+
"increase_batch_size_on_plateau_rate": 2.0,
|
| 836 |
+
"learning_rate": 0.001,
|
| 837 |
+
"learning_rate_scaling": "linear",
|
| 838 |
+
"learning_rate_scheduler": {
|
| 839 |
+
"decay": null,
|
| 840 |
+
"decay_rate": 0.96,
|
| 841 |
+
"decay_steps": 10000,
|
| 842 |
+
"eta_min": 0,
|
| 843 |
+
"reduce_eval_metric": "loss",
|
| 844 |
+
"reduce_eval_split": "training",
|
| 845 |
+
"reduce_on_plateau": 0,
|
| 846 |
+
"reduce_on_plateau_patience": 10,
|
| 847 |
+
"reduce_on_plateau_rate": 0.1,
|
| 848 |
+
"staircase": false,
|
| 849 |
+
"t_0": null,
|
| 850 |
+
"t_mult": 1,
|
| 851 |
+
"warmup_evaluations": 0,
|
| 852 |
+
"warmup_fraction": 0.0
|
| 853 |
+
},
|
| 854 |
+
"max_batch_size": 1099511627776,
|
| 855 |
+
"optimizer": {
|
| 856 |
+
"amsgrad": false,
|
| 857 |
+
"betas": [
|
| 858 |
+
0.9,
|
| 859 |
+
0.999
|
| 860 |
+
],
|
| 861 |
+
"eps": 1e-08,
|
| 862 |
+
"type": "adam",
|
| 863 |
+
"weight_decay": 0.0
|
| 864 |
+
},
|
| 865 |
+
"profiler": {
|
| 866 |
+
"active": 3,
|
| 867 |
+
"repeat": 5,
|
| 868 |
+
"skip_first": 0,
|
| 869 |
+
"wait": 1,
|
| 870 |
+
"warmup": 1
|
| 871 |
+
},
|
| 872 |
+
"regularization_lambda": 0.0,
|
| 873 |
+
"regularization_type": "l2",
|
| 874 |
+
"should_shuffle": true,
|
| 875 |
+
"skip_all_evaluation": false,
|
| 876 |
+
"steps_per_checkpoint": 0,
|
| 877 |
+
"train_steps": null,
|
| 878 |
+
"use_mixed_precision": false,
|
| 879 |
+
"validation_field": "language",
|
| 880 |
+
"validation_metric": "accuracy"
|
| 881 |
+
}
|
| 882 |
+
},
|
| 883 |
+
"data_format": "json",
|
| 884 |
+
"dataset": "/data/data.json",
|
| 885 |
+
"ludwig_version": "0.10.2.dev",
|
| 886 |
+
"random_seed": 42,
|
| 887 |
+
"torch_version": "2.0.0+cpu"
|
| 888 |
+
}
|
model/llm_eval_examples/0.csv
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
inputs,targets,outputs
|
model/llm_eval_examples/2.csv
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
inputs,targets,outputs
|
model/llm_eval_examples/4.csv
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
inputs,targets,outputs
|
model/llm_eval_examples/6.csv
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
inputs,targets,outputs
|
model/llm_eval_examples/8.csv
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
inputs,targets,outputs
|
model/model_hyperparameters.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
model/model_weights
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dccb2f0416fadce9c5e4e4d2cc528db6064ec4d6296c67f95ed137975db95c66
|
| 3 |
+
size 17302302
|
model/training_progress.json
ADDED
|
@@ -0,0 +1,1461 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"batch_size": 64,
|
| 3 |
+
"best_eval_metric_checkpoint_number": 6,
|
| 4 |
+
"best_eval_metric_epoch": 4,
|
| 5 |
+
"best_eval_metric_steps": 372,
|
| 6 |
+
"best_eval_metric_value": 0.9799168109893799,
|
| 7 |
+
"best_eval_test_metrics": {
|
| 8 |
+
"combined": {
|
| 9 |
+
"loss": 0.07570453733205795
|
| 10 |
+
},
|
| 11 |
+
"language": {
|
| 12 |
+
"accuracy": 0.9817298650741577,
|
| 13 |
+
"accuracy_micro": 0.9799426794052124,
|
| 14 |
+
"loss": 0.07570453733205795,
|
| 15 |
+
"roc_auc": 0.999114990234375
|
| 16 |
+
}
|
| 17 |
+
},
|
| 18 |
+
"best_eval_train_metrics": {
|
| 19 |
+
"combined": {
|
| 20 |
+
"loss": 0.03008267655968666
|
| 21 |
+
},
|
| 22 |
+
"language": {
|
| 23 |
+
"accuracy": 0.9913396835327148,
|
| 24 |
+
"accuracy_micro": 0.9910488128662109,
|
| 25 |
+
"loss": 0.03008267655968666,
|
| 26 |
+
"roc_auc": 0.9997660517692566
|
| 27 |
+
}
|
| 28 |
+
},
|
| 29 |
+
"best_eval_validation_metrics": {
|
| 30 |
+
"combined": {
|
| 31 |
+
"loss": 0.06920339167118073
|
| 32 |
+
},
|
| 33 |
+
"language": {
|
| 34 |
+
"accuracy": 0.9799168109893799,
|
| 35 |
+
"accuracy_micro": 0.9784792065620422,
|
| 36 |
+
"loss": 0.06920339167118073,
|
| 37 |
+
"roc_auc": 0.998852550983429
|
| 38 |
+
}
|
| 39 |
+
},
|
| 40 |
+
"best_increase_batch_size_eval_metric": Infinity,
|
| 41 |
+
"checkpoint_number": 10,
|
| 42 |
+
"checkpoint_to_epoch": {
|
| 43 |
+
"1": 1,
|
| 44 |
+
"10": 5,
|
| 45 |
+
"2": 1,
|
| 46 |
+
"3": 2,
|
| 47 |
+
"4": 2,
|
| 48 |
+
"5": 3,
|
| 49 |
+
"6": 3,
|
| 50 |
+
"7": 4,
|
| 51 |
+
"8": 4,
|
| 52 |
+
"9": 5
|
| 53 |
+
},
|
| 54 |
+
"checkpoint_to_step": {
|
| 55 |
+
"1": 93,
|
| 56 |
+
"10": 465,
|
| 57 |
+
"2": 93,
|
| 58 |
+
"3": 186,
|
| 59 |
+
"4": 186,
|
| 60 |
+
"5": 279,
|
| 61 |
+
"6": 279,
|
| 62 |
+
"7": 372,
|
| 63 |
+
"8": 372,
|
| 64 |
+
"9": 465
|
| 65 |
+
},
|
| 66 |
+
"cumulative_checkpoint_token_usage": {
|
| 67 |
+
"1": 2356558,
|
| 68 |
+
"10": 11782790,
|
| 69 |
+
"2": 2356558,
|
| 70 |
+
"3": 4713116,
|
| 71 |
+
"4": 4713116,
|
| 72 |
+
"5": 7069674,
|
| 73 |
+
"6": 7069674,
|
| 74 |
+
"7": 9426232,
|
| 75 |
+
"8": 9426232,
|
| 76 |
+
"9": 11782790
|
| 77 |
+
},
|
| 78 |
+
"cumulative_step_token_usage": {
|
| 79 |
+
"0": 25472,
|
| 80 |
+
"1": 50944,
|
| 81 |
+
"10": 280192,
|
| 82 |
+
"100": 2560334,
|
| 83 |
+
"101": 2585806,
|
| 84 |
+
"102": 2611278,
|
| 85 |
+
"103": 2636750,
|
| 86 |
+
"104": 2662222,
|
| 87 |
+
"105": 2687694,
|
| 88 |
+
"106": 2713166,
|
| 89 |
+
"107": 2738638,
|
| 90 |
+
"108": 2764110,
|
| 91 |
+
"109": 2789582,
|
| 92 |
+
"11": 305664,
|
| 93 |
+
"110": 2815054,
|
| 94 |
+
"111": 2840526,
|
| 95 |
+
"112": 2865998,
|
| 96 |
+
"113": 2891470,
|
| 97 |
+
"114": 2916942,
|
| 98 |
+
"115": 2942414,
|
| 99 |
+
"116": 2967886,
|
| 100 |
+
"117": 2993358,
|
| 101 |
+
"118": 3018830,
|
| 102 |
+
"119": 3044302,
|
| 103 |
+
"12": 331136,
|
| 104 |
+
"120": 3069774,
|
| 105 |
+
"121": 3095246,
|
| 106 |
+
"122": 3120718,
|
| 107 |
+
"123": 3146190,
|
| 108 |
+
"124": 3171662,
|
| 109 |
+
"125": 3197134,
|
| 110 |
+
"126": 3222606,
|
| 111 |
+
"127": 3248078,
|
| 112 |
+
"128": 3273550,
|
| 113 |
+
"129": 3299022,
|
| 114 |
+
"13": 356608,
|
| 115 |
+
"130": 3324494,
|
| 116 |
+
"131": 3349966,
|
| 117 |
+
"132": 3375438,
|
| 118 |
+
"133": 3400910,
|
| 119 |
+
"134": 3426382,
|
| 120 |
+
"135": 3451854,
|
| 121 |
+
"136": 3477326,
|
| 122 |
+
"137": 3502798,
|
| 123 |
+
"138": 3528270,
|
| 124 |
+
"139": 3553742,
|
| 125 |
+
"14": 382080,
|
| 126 |
+
"140": 3579214,
|
| 127 |
+
"141": 3604686,
|
| 128 |
+
"142": 3630158,
|
| 129 |
+
"143": 3655630,
|
| 130 |
+
"144": 3681102,
|
| 131 |
+
"145": 3706574,
|
| 132 |
+
"146": 3732046,
|
| 133 |
+
"147": 3757518,
|
| 134 |
+
"148": 3782990,
|
| 135 |
+
"149": 3808462,
|
| 136 |
+
"15": 407552,
|
| 137 |
+
"150": 3833934,
|
| 138 |
+
"151": 3859406,
|
| 139 |
+
"152": 3884878,
|
| 140 |
+
"153": 3910350,
|
| 141 |
+
"154": 3935822,
|
| 142 |
+
"155": 3961294,
|
| 143 |
+
"156": 3986766,
|
| 144 |
+
"157": 4012238,
|
| 145 |
+
"158": 4037710,
|
| 146 |
+
"159": 4063182,
|
| 147 |
+
"16": 433024,
|
| 148 |
+
"160": 4088654,
|
| 149 |
+
"161": 4114126,
|
| 150 |
+
"162": 4139598,
|
| 151 |
+
"163": 4165070,
|
| 152 |
+
"164": 4190542,
|
| 153 |
+
"165": 4216014,
|
| 154 |
+
"166": 4241486,
|
| 155 |
+
"167": 4266958,
|
| 156 |
+
"168": 4292430,
|
| 157 |
+
"169": 4317902,
|
| 158 |
+
"17": 458496,
|
| 159 |
+
"170": 4343374,
|
| 160 |
+
"171": 4368846,
|
| 161 |
+
"172": 4394318,
|
| 162 |
+
"173": 4419790,
|
| 163 |
+
"174": 4445262,
|
| 164 |
+
"175": 4470734,
|
| 165 |
+
"176": 4496206,
|
| 166 |
+
"177": 4521678,
|
| 167 |
+
"178": 4547150,
|
| 168 |
+
"179": 4572622,
|
| 169 |
+
"18": 483968,
|
| 170 |
+
"180": 4598094,
|
| 171 |
+
"181": 4623566,
|
| 172 |
+
"182": 4649038,
|
| 173 |
+
"183": 4674510,
|
| 174 |
+
"184": 4699982,
|
| 175 |
+
"185": 4713116,
|
| 176 |
+
"186": 4738588,
|
| 177 |
+
"187": 4764060,
|
| 178 |
+
"188": 4789532,
|
| 179 |
+
"189": 4815004,
|
| 180 |
+
"19": 509440,
|
| 181 |
+
"190": 4840476,
|
| 182 |
+
"191": 4865948,
|
| 183 |
+
"192": 4891420,
|
| 184 |
+
"193": 4916892,
|
| 185 |
+
"194": 4942364,
|
| 186 |
+
"195": 4967836,
|
| 187 |
+
"196": 4993308,
|
| 188 |
+
"197": 5018780,
|
| 189 |
+
"198": 5044252,
|
| 190 |
+
"199": 5069724,
|
| 191 |
+
"2": 76416,
|
| 192 |
+
"20": 534912,
|
| 193 |
+
"200": 5095196,
|
| 194 |
+
"201": 5120668,
|
| 195 |
+
"202": 5146140,
|
| 196 |
+
"203": 5171612,
|
| 197 |
+
"204": 5197084,
|
| 198 |
+
"205": 5222556,
|
| 199 |
+
"206": 5248028,
|
| 200 |
+
"207": 5273500,
|
| 201 |
+
"208": 5298972,
|
| 202 |
+
"209": 5324444,
|
| 203 |
+
"21": 560384,
|
| 204 |
+
"210": 5349916,
|
| 205 |
+
"211": 5375388,
|
| 206 |
+
"212": 5400860,
|
| 207 |
+
"213": 5426332,
|
| 208 |
+
"214": 5451804,
|
| 209 |
+
"215": 5477276,
|
| 210 |
+
"216": 5502748,
|
| 211 |
+
"217": 5528220,
|
| 212 |
+
"218": 5553692,
|
| 213 |
+
"219": 5579164,
|
| 214 |
+
"22": 585856,
|
| 215 |
+
"220": 5604636,
|
| 216 |
+
"221": 5630108,
|
| 217 |
+
"222": 5655580,
|
| 218 |
+
"223": 5681052,
|
| 219 |
+
"224": 5706524,
|
| 220 |
+
"225": 5731996,
|
| 221 |
+
"226": 5757468,
|
| 222 |
+
"227": 5782940,
|
| 223 |
+
"228": 5808412,
|
| 224 |
+
"229": 5833884,
|
| 225 |
+
"23": 611328,
|
| 226 |
+
"230": 5859356,
|
| 227 |
+
"231": 5884828,
|
| 228 |
+
"232": 5910300,
|
| 229 |
+
"233": 5935772,
|
| 230 |
+
"234": 5961244,
|
| 231 |
+
"235": 5986716,
|
| 232 |
+
"236": 6012188,
|
| 233 |
+
"237": 6037660,
|
| 234 |
+
"238": 6063132,
|
| 235 |
+
"239": 6088604,
|
| 236 |
+
"24": 636800,
|
| 237 |
+
"240": 6114076,
|
| 238 |
+
"241": 6139548,
|
| 239 |
+
"242": 6165020,
|
| 240 |
+
"243": 6190492,
|
| 241 |
+
"244": 6215964,
|
| 242 |
+
"245": 6241436,
|
| 243 |
+
"246": 6266908,
|
| 244 |
+
"247": 6292380,
|
| 245 |
+
"248": 6317852,
|
| 246 |
+
"249": 6343324,
|
| 247 |
+
"25": 662272,
|
| 248 |
+
"250": 6368796,
|
| 249 |
+
"251": 6394268,
|
| 250 |
+
"252": 6419740,
|
| 251 |
+
"253": 6445212,
|
| 252 |
+
"254": 6470684,
|
| 253 |
+
"255": 6496156,
|
| 254 |
+
"256": 6521628,
|
| 255 |
+
"257": 6547100,
|
| 256 |
+
"258": 6572572,
|
| 257 |
+
"259": 6598044,
|
| 258 |
+
"26": 687744,
|
| 259 |
+
"260": 6623516,
|
| 260 |
+
"261": 6648988,
|
| 261 |
+
"262": 6674460,
|
| 262 |
+
"263": 6699932,
|
| 263 |
+
"264": 6725404,
|
| 264 |
+
"265": 6750876,
|
| 265 |
+
"266": 6776348,
|
| 266 |
+
"267": 6801820,
|
| 267 |
+
"268": 6827292,
|
| 268 |
+
"269": 6852764,
|
| 269 |
+
"27": 713216,
|
| 270 |
+
"270": 6878236,
|
| 271 |
+
"271": 6903708,
|
| 272 |
+
"272": 6929180,
|
| 273 |
+
"273": 6954652,
|
| 274 |
+
"274": 6980124,
|
| 275 |
+
"275": 7005596,
|
| 276 |
+
"276": 7031068,
|
| 277 |
+
"277": 7056540,
|
| 278 |
+
"278": 7069674,
|
| 279 |
+
"279": 7095146,
|
| 280 |
+
"28": 738688,
|
| 281 |
+
"280": 7120618,
|
| 282 |
+
"281": 7146090,
|
| 283 |
+
"282": 7171562,
|
| 284 |
+
"283": 7197034,
|
| 285 |
+
"284": 7222506,
|
| 286 |
+
"285": 7247978,
|
| 287 |
+
"286": 7273450,
|
| 288 |
+
"287": 7298922,
|
| 289 |
+
"288": 7324394,
|
| 290 |
+
"289": 7349866,
|
| 291 |
+
"29": 764160,
|
| 292 |
+
"290": 7375338,
|
| 293 |
+
"291": 7400810,
|
| 294 |
+
"292": 7426282,
|
| 295 |
+
"293": 7451754,
|
| 296 |
+
"294": 7477226,
|
| 297 |
+
"295": 7502698,
|
| 298 |
+
"296": 7528170,
|
| 299 |
+
"297": 7553642,
|
| 300 |
+
"298": 7579114,
|
| 301 |
+
"299": 7604586,
|
| 302 |
+
"3": 101888,
|
| 303 |
+
"30": 789632,
|
| 304 |
+
"300": 7630058,
|
| 305 |
+
"301": 7655530,
|
| 306 |
+
"302": 7681002,
|
| 307 |
+
"303": 7706474,
|
| 308 |
+
"304": 7731946,
|
| 309 |
+
"305": 7757418,
|
| 310 |
+
"306": 7782890,
|
| 311 |
+
"307": 7808362,
|
| 312 |
+
"308": 7833834,
|
| 313 |
+
"309": 7859306,
|
| 314 |
+
"31": 815104,
|
| 315 |
+
"310": 7884778,
|
| 316 |
+
"311": 7910250,
|
| 317 |
+
"312": 7935722,
|
| 318 |
+
"313": 7961194,
|
| 319 |
+
"314": 7986666,
|
| 320 |
+
"315": 8012138,
|
| 321 |
+
"316": 8037610,
|
| 322 |
+
"317": 8063082,
|
| 323 |
+
"318": 8088554,
|
| 324 |
+
"319": 8114026,
|
| 325 |
+
"32": 840576,
|
| 326 |
+
"320": 8139498,
|
| 327 |
+
"321": 8164970,
|
| 328 |
+
"322": 8190442,
|
| 329 |
+
"323": 8215914,
|
| 330 |
+
"324": 8241386,
|
| 331 |
+
"325": 8266858,
|
| 332 |
+
"326": 8292330,
|
| 333 |
+
"327": 8317802,
|
| 334 |
+
"328": 8343274,
|
| 335 |
+
"329": 8368746,
|
| 336 |
+
"33": 866048,
|
| 337 |
+
"330": 8394218,
|
| 338 |
+
"331": 8419690,
|
| 339 |
+
"332": 8445162,
|
| 340 |
+
"333": 8470634,
|
| 341 |
+
"334": 8496106,
|
| 342 |
+
"335": 8521578,
|
| 343 |
+
"336": 8547050,
|
| 344 |
+
"337": 8572522,
|
| 345 |
+
"338": 8597994,
|
| 346 |
+
"339": 8623466,
|
| 347 |
+
"34": 891520,
|
| 348 |
+
"340": 8648938,
|
| 349 |
+
"341": 8674410,
|
| 350 |
+
"342": 8699882,
|
| 351 |
+
"343": 8725354,
|
| 352 |
+
"344": 8750826,
|
| 353 |
+
"345": 8776298,
|
| 354 |
+
"346": 8801770,
|
| 355 |
+
"347": 8827242,
|
| 356 |
+
"348": 8852714,
|
| 357 |
+
"349": 8878186,
|
| 358 |
+
"35": 916992,
|
| 359 |
+
"350": 8903658,
|
| 360 |
+
"351": 8929130,
|
| 361 |
+
"352": 8954602,
|
| 362 |
+
"353": 8980074,
|
| 363 |
+
"354": 9005546,
|
| 364 |
+
"355": 9031018,
|
| 365 |
+
"356": 9056490,
|
| 366 |
+
"357": 9081962,
|
| 367 |
+
"358": 9107434,
|
| 368 |
+
"359": 9132906,
|
| 369 |
+
"36": 942464,
|
| 370 |
+
"360": 9158378,
|
| 371 |
+
"361": 9183850,
|
| 372 |
+
"362": 9209322,
|
| 373 |
+
"363": 9234794,
|
| 374 |
+
"364": 9260266,
|
| 375 |
+
"365": 9285738,
|
| 376 |
+
"366": 9311210,
|
| 377 |
+
"367": 9336682,
|
| 378 |
+
"368": 9362154,
|
| 379 |
+
"369": 9387626,
|
| 380 |
+
"37": 967936,
|
| 381 |
+
"370": 9413098,
|
| 382 |
+
"371": 9426232,
|
| 383 |
+
"372": 9451704,
|
| 384 |
+
"373": 9477176,
|
| 385 |
+
"374": 9502648,
|
| 386 |
+
"375": 9528120,
|
| 387 |
+
"376": 9553592,
|
| 388 |
+
"377": 9579064,
|
| 389 |
+
"378": 9604536,
|
| 390 |
+
"379": 9630008,
|
| 391 |
+
"38": 993408,
|
| 392 |
+
"380": 9655480,
|
| 393 |
+
"381": 9680952,
|
| 394 |
+
"382": 9706424,
|
| 395 |
+
"383": 9731896,
|
| 396 |
+
"384": 9757368,
|
| 397 |
+
"385": 9782840,
|
| 398 |
+
"386": 9808312,
|
| 399 |
+
"387": 9833784,
|
| 400 |
+
"388": 9859256,
|
| 401 |
+
"389": 9884728,
|
| 402 |
+
"39": 1018880,
|
| 403 |
+
"390": 9910200,
|
| 404 |
+
"391": 9935672,
|
| 405 |
+
"392": 9961144,
|
| 406 |
+
"393": 9986616,
|
| 407 |
+
"394": 10012088,
|
| 408 |
+
"395": 10037560,
|
| 409 |
+
"396": 10063032,
|
| 410 |
+
"397": 10088504,
|
| 411 |
+
"398": 10113976,
|
| 412 |
+
"399": 10139448,
|
| 413 |
+
"4": 127360,
|
| 414 |
+
"40": 1044352,
|
| 415 |
+
"400": 10164920,
|
| 416 |
+
"401": 10190392,
|
| 417 |
+
"402": 10215864,
|
| 418 |
+
"403": 10241336,
|
| 419 |
+
"404": 10266808,
|
| 420 |
+
"405": 10292280,
|
| 421 |
+
"406": 10317752,
|
| 422 |
+
"407": 10343224,
|
| 423 |
+
"408": 10368696,
|
| 424 |
+
"409": 10394168,
|
| 425 |
+
"41": 1069824,
|
| 426 |
+
"410": 10419640,
|
| 427 |
+
"411": 10445112,
|
| 428 |
+
"412": 10470584,
|
| 429 |
+
"413": 10496056,
|
| 430 |
+
"414": 10521528,
|
| 431 |
+
"415": 10547000,
|
| 432 |
+
"416": 10572472,
|
| 433 |
+
"417": 10597944,
|
| 434 |
+
"418": 10623416,
|
| 435 |
+
"419": 10648888,
|
| 436 |
+
"42": 1095296,
|
| 437 |
+
"420": 10674360,
|
| 438 |
+
"421": 10699832,
|
| 439 |
+
"422": 10725304,
|
| 440 |
+
"423": 10750776,
|
| 441 |
+
"424": 10776248,
|
| 442 |
+
"425": 10801720,
|
| 443 |
+
"426": 10827192,
|
| 444 |
+
"427": 10852664,
|
| 445 |
+
"428": 10878136,
|
| 446 |
+
"429": 10903608,
|
| 447 |
+
"43": 1120768,
|
| 448 |
+
"430": 10929080,
|
| 449 |
+
"431": 10954552,
|
| 450 |
+
"432": 10980024,
|
| 451 |
+
"433": 11005496,
|
| 452 |
+
"434": 11030968,
|
| 453 |
+
"435": 11056440,
|
| 454 |
+
"436": 11081912,
|
| 455 |
+
"437": 11107384,
|
| 456 |
+
"438": 11132856,
|
| 457 |
+
"439": 11158328,
|
| 458 |
+
"44": 1146240,
|
| 459 |
+
"440": 11183800,
|
| 460 |
+
"441": 11209272,
|
| 461 |
+
"442": 11234744,
|
| 462 |
+
"443": 11260216,
|
| 463 |
+
"444": 11285688,
|
| 464 |
+
"445": 11311160,
|
| 465 |
+
"446": 11336632,
|
| 466 |
+
"447": 11362104,
|
| 467 |
+
"448": 11387576,
|
| 468 |
+
"449": 11413048,
|
| 469 |
+
"45": 1171712,
|
| 470 |
+
"450": 11438520,
|
| 471 |
+
"451": 11463992,
|
| 472 |
+
"452": 11489464,
|
| 473 |
+
"453": 11514936,
|
| 474 |
+
"454": 11540408,
|
| 475 |
+
"455": 11565880,
|
| 476 |
+
"456": 11591352,
|
| 477 |
+
"457": 11616824,
|
| 478 |
+
"458": 11642296,
|
| 479 |
+
"459": 11667768,
|
| 480 |
+
"46": 1197184,
|
| 481 |
+
"460": 11693240,
|
| 482 |
+
"461": 11718712,
|
| 483 |
+
"462": 11744184,
|
| 484 |
+
"463": 11769656,
|
| 485 |
+
"464": 11782790,
|
| 486 |
+
"47": 1222656,
|
| 487 |
+
"48": 1248128,
|
| 488 |
+
"49": 1273600,
|
| 489 |
+
"5": 152832,
|
| 490 |
+
"50": 1299072,
|
| 491 |
+
"51": 1324544,
|
| 492 |
+
"52": 1350016,
|
| 493 |
+
"53": 1375488,
|
| 494 |
+
"54": 1400960,
|
| 495 |
+
"55": 1426432,
|
| 496 |
+
"56": 1451904,
|
| 497 |
+
"57": 1477376,
|
| 498 |
+
"58": 1502848,
|
| 499 |
+
"59": 1528320,
|
| 500 |
+
"6": 178304,
|
| 501 |
+
"60": 1553792,
|
| 502 |
+
"61": 1579264,
|
| 503 |
+
"62": 1604736,
|
| 504 |
+
"63": 1630208,
|
| 505 |
+
"64": 1655680,
|
| 506 |
+
"65": 1681152,
|
| 507 |
+
"66": 1706624,
|
| 508 |
+
"67": 1732096,
|
| 509 |
+
"68": 1757568,
|
| 510 |
+
"69": 1783040,
|
| 511 |
+
"7": 203776,
|
| 512 |
+
"70": 1808512,
|
| 513 |
+
"71": 1833984,
|
| 514 |
+
"72": 1859456,
|
| 515 |
+
"73": 1884928,
|
| 516 |
+
"74": 1910400,
|
| 517 |
+
"75": 1935872,
|
| 518 |
+
"76": 1961344,
|
| 519 |
+
"77": 1986816,
|
| 520 |
+
"78": 2012288,
|
| 521 |
+
"79": 2037760,
|
| 522 |
+
"8": 229248,
|
| 523 |
+
"80": 2063232,
|
| 524 |
+
"81": 2088704,
|
| 525 |
+
"82": 2114176,
|
| 526 |
+
"83": 2139648,
|
| 527 |
+
"84": 2165120,
|
| 528 |
+
"85": 2190592,
|
| 529 |
+
"86": 2216064,
|
| 530 |
+
"87": 2241536,
|
| 531 |
+
"88": 2267008,
|
| 532 |
+
"89": 2292480,
|
| 533 |
+
"9": 254720,
|
| 534 |
+
"90": 2317952,
|
| 535 |
+
"91": 2343424,
|
| 536 |
+
"92": 2356558,
|
| 537 |
+
"93": 2382030,
|
| 538 |
+
"94": 2407502,
|
| 539 |
+
"95": 2432974,
|
| 540 |
+
"96": 2458446,
|
| 541 |
+
"97": 2483918,
|
| 542 |
+
"98": 2509390,
|
| 543 |
+
"99": 2534862
|
| 544 |
+
},
|
| 545 |
+
"epoch": 5,
|
| 546 |
+
"incremental_checkpoint_token_usage": {
|
| 547 |
+
"1": 2356558,
|
| 548 |
+
"10": 0,
|
| 549 |
+
"2": 0,
|
| 550 |
+
"3": 2356558,
|
| 551 |
+
"4": 0,
|
| 552 |
+
"5": 2356558,
|
| 553 |
+
"6": 0,
|
| 554 |
+
"7": 2356558,
|
| 555 |
+
"8": 0,
|
| 556 |
+
"9": 2356558
|
| 557 |
+
},
|
| 558 |
+
"incremental_step_token_usage": {
|
| 559 |
+
"0": 25472,
|
| 560 |
+
"1": 25472,
|
| 561 |
+
"10": 25472,
|
| 562 |
+
"100": 25472,
|
| 563 |
+
"101": 25472,
|
| 564 |
+
"102": 25472,
|
| 565 |
+
"103": 25472,
|
| 566 |
+
"104": 25472,
|
| 567 |
+
"105": 25472,
|
| 568 |
+
"106": 25472,
|
| 569 |
+
"107": 25472,
|
| 570 |
+
"108": 25472,
|
| 571 |
+
"109": 25472,
|
| 572 |
+
"11": 25472,
|
| 573 |
+
"110": 25472,
|
| 574 |
+
"111": 25472,
|
| 575 |
+
"112": 25472,
|
| 576 |
+
"113": 25472,
|
| 577 |
+
"114": 25472,
|
| 578 |
+
"115": 25472,
|
| 579 |
+
"116": 25472,
|
| 580 |
+
"117": 25472,
|
| 581 |
+
"118": 25472,
|
| 582 |
+
"119": 25472,
|
| 583 |
+
"12": 25472,
|
| 584 |
+
"120": 25472,
|
| 585 |
+
"121": 25472,
|
| 586 |
+
"122": 25472,
|
| 587 |
+
"123": 25472,
|
| 588 |
+
"124": 25472,
|
| 589 |
+
"125": 25472,
|
| 590 |
+
"126": 25472,
|
| 591 |
+
"127": 25472,
|
| 592 |
+
"128": 25472,
|
| 593 |
+
"129": 25472,
|
| 594 |
+
"13": 25472,
|
| 595 |
+
"130": 25472,
|
| 596 |
+
"131": 25472,
|
| 597 |
+
"132": 25472,
|
| 598 |
+
"133": 25472,
|
| 599 |
+
"134": 25472,
|
| 600 |
+
"135": 25472,
|
| 601 |
+
"136": 25472,
|
| 602 |
+
"137": 25472,
|
| 603 |
+
"138": 25472,
|
| 604 |
+
"139": 25472,
|
| 605 |
+
"14": 25472,
|
| 606 |
+
"140": 25472,
|
| 607 |
+
"141": 25472,
|
| 608 |
+
"142": 25472,
|
| 609 |
+
"143": 25472,
|
| 610 |
+
"144": 25472,
|
| 611 |
+
"145": 25472,
|
| 612 |
+
"146": 25472,
|
| 613 |
+
"147": 25472,
|
| 614 |
+
"148": 25472,
|
| 615 |
+
"149": 25472,
|
| 616 |
+
"15": 25472,
|
| 617 |
+
"150": 25472,
|
| 618 |
+
"151": 25472,
|
| 619 |
+
"152": 25472,
|
| 620 |
+
"153": 25472,
|
| 621 |
+
"154": 25472,
|
| 622 |
+
"155": 25472,
|
| 623 |
+
"156": 25472,
|
| 624 |
+
"157": 25472,
|
| 625 |
+
"158": 25472,
|
| 626 |
+
"159": 25472,
|
| 627 |
+
"16": 25472,
|
| 628 |
+
"160": 25472,
|
| 629 |
+
"161": 25472,
|
| 630 |
+
"162": 25472,
|
| 631 |
+
"163": 25472,
|
| 632 |
+
"164": 25472,
|
| 633 |
+
"165": 25472,
|
| 634 |
+
"166": 25472,
|
| 635 |
+
"167": 25472,
|
| 636 |
+
"168": 25472,
|
| 637 |
+
"169": 25472,
|
| 638 |
+
"17": 25472,
|
| 639 |
+
"170": 25472,
|
| 640 |
+
"171": 25472,
|
| 641 |
+
"172": 25472,
|
| 642 |
+
"173": 25472,
|
| 643 |
+
"174": 25472,
|
| 644 |
+
"175": 25472,
|
| 645 |
+
"176": 25472,
|
| 646 |
+
"177": 25472,
|
| 647 |
+
"178": 25472,
|
| 648 |
+
"179": 25472,
|
| 649 |
+
"18": 25472,
|
| 650 |
+
"180": 25472,
|
| 651 |
+
"181": 25472,
|
| 652 |
+
"182": 25472,
|
| 653 |
+
"183": 25472,
|
| 654 |
+
"184": 25472,
|
| 655 |
+
"185": 13134,
|
| 656 |
+
"186": 25472,
|
| 657 |
+
"187": 25472,
|
| 658 |
+
"188": 25472,
|
| 659 |
+
"189": 25472,
|
| 660 |
+
"19": 25472,
|
| 661 |
+
"190": 25472,
|
| 662 |
+
"191": 25472,
|
| 663 |
+
"192": 25472,
|
| 664 |
+
"193": 25472,
|
| 665 |
+
"194": 25472,
|
| 666 |
+
"195": 25472,
|
| 667 |
+
"196": 25472,
|
| 668 |
+
"197": 25472,
|
| 669 |
+
"198": 25472,
|
| 670 |
+
"199": 25472,
|
| 671 |
+
"2": 25472,
|
| 672 |
+
"20": 25472,
|
| 673 |
+
"200": 25472,
|
| 674 |
+
"201": 25472,
|
| 675 |
+
"202": 25472,
|
| 676 |
+
"203": 25472,
|
| 677 |
+
"204": 25472,
|
| 678 |
+
"205": 25472,
|
| 679 |
+
"206": 25472,
|
| 680 |
+
"207": 25472,
|
| 681 |
+
"208": 25472,
|
| 682 |
+
"209": 25472,
|
| 683 |
+
"21": 25472,
|
| 684 |
+
"210": 25472,
|
| 685 |
+
"211": 25472,
|
| 686 |
+
"212": 25472,
|
| 687 |
+
"213": 25472,
|
| 688 |
+
"214": 25472,
|
| 689 |
+
"215": 25472,
|
| 690 |
+
"216": 25472,
|
| 691 |
+
"217": 25472,
|
| 692 |
+
"218": 25472,
|
| 693 |
+
"219": 25472,
|
| 694 |
+
"22": 25472,
|
| 695 |
+
"220": 25472,
|
| 696 |
+
"221": 25472,
|
| 697 |
+
"222": 25472,
|
| 698 |
+
"223": 25472,
|
| 699 |
+
"224": 25472,
|
| 700 |
+
"225": 25472,
|
| 701 |
+
"226": 25472,
|
| 702 |
+
"227": 25472,
|
| 703 |
+
"228": 25472,
|
| 704 |
+
"229": 25472,
|
| 705 |
+
"23": 25472,
|
| 706 |
+
"230": 25472,
|
| 707 |
+
"231": 25472,
|
| 708 |
+
"232": 25472,
|
| 709 |
+
"233": 25472,
|
| 710 |
+
"234": 25472,
|
| 711 |
+
"235": 25472,
|
| 712 |
+
"236": 25472,
|
| 713 |
+
"237": 25472,
|
| 714 |
+
"238": 25472,
|
| 715 |
+
"239": 25472,
|
| 716 |
+
"24": 25472,
|
| 717 |
+
"240": 25472,
|
| 718 |
+
"241": 25472,
|
| 719 |
+
"242": 25472,
|
| 720 |
+
"243": 25472,
|
| 721 |
+
"244": 25472,
|
| 722 |
+
"245": 25472,
|
| 723 |
+
"246": 25472,
|
| 724 |
+
"247": 25472,
|
| 725 |
+
"248": 25472,
|
| 726 |
+
"249": 25472,
|
| 727 |
+
"25": 25472,
|
| 728 |
+
"250": 25472,
|
| 729 |
+
"251": 25472,
|
| 730 |
+
"252": 25472,
|
| 731 |
+
"253": 25472,
|
| 732 |
+
"254": 25472,
|
| 733 |
+
"255": 25472,
|
| 734 |
+
"256": 25472,
|
| 735 |
+
"257": 25472,
|
| 736 |
+
"258": 25472,
|
| 737 |
+
"259": 25472,
|
| 738 |
+
"26": 25472,
|
| 739 |
+
"260": 25472,
|
| 740 |
+
"261": 25472,
|
| 741 |
+
"262": 25472,
|
| 742 |
+
"263": 25472,
|
| 743 |
+
"264": 25472,
|
| 744 |
+
"265": 25472,
|
| 745 |
+
"266": 25472,
|
| 746 |
+
"267": 25472,
|
| 747 |
+
"268": 25472,
|
| 748 |
+
"269": 25472,
|
| 749 |
+
"27": 25472,
|
| 750 |
+
"270": 25472,
|
| 751 |
+
"271": 25472,
|
| 752 |
+
"272": 25472,
|
| 753 |
+
"273": 25472,
|
| 754 |
+
"274": 25472,
|
| 755 |
+
"275": 25472,
|
| 756 |
+
"276": 25472,
|
| 757 |
+
"277": 25472,
|
| 758 |
+
"278": 13134,
|
| 759 |
+
"279": 25472,
|
| 760 |
+
"28": 25472,
|
| 761 |
+
"280": 25472,
|
| 762 |
+
"281": 25472,
|
| 763 |
+
"282": 25472,
|
| 764 |
+
"283": 25472,
|
| 765 |
+
"284": 25472,
|
| 766 |
+
"285": 25472,
|
| 767 |
+
"286": 25472,
|
| 768 |
+
"287": 25472,
|
| 769 |
+
"288": 25472,
|
| 770 |
+
"289": 25472,
|
| 771 |
+
"29": 25472,
|
| 772 |
+
"290": 25472,
|
| 773 |
+
"291": 25472,
|
| 774 |
+
"292": 25472,
|
| 775 |
+
"293": 25472,
|
| 776 |
+
"294": 25472,
|
| 777 |
+
"295": 25472,
|
| 778 |
+
"296": 25472,
|
| 779 |
+
"297": 25472,
|
| 780 |
+
"298": 25472,
|
| 781 |
+
"299": 25472,
|
| 782 |
+
"3": 25472,
|
| 783 |
+
"30": 25472,
|
| 784 |
+
"300": 25472,
|
| 785 |
+
"301": 25472,
|
| 786 |
+
"302": 25472,
|
| 787 |
+
"303": 25472,
|
| 788 |
+
"304": 25472,
|
| 789 |
+
"305": 25472,
|
| 790 |
+
"306": 25472,
|
| 791 |
+
"307": 25472,
|
| 792 |
+
"308": 25472,
|
| 793 |
+
"309": 25472,
|
| 794 |
+
"31": 25472,
|
| 795 |
+
"310": 25472,
|
| 796 |
+
"311": 25472,
|
| 797 |
+
"312": 25472,
|
| 798 |
+
"313": 25472,
|
| 799 |
+
"314": 25472,
|
| 800 |
+
"315": 25472,
|
| 801 |
+
"316": 25472,
|
| 802 |
+
"317": 25472,
|
| 803 |
+
"318": 25472,
|
| 804 |
+
"319": 25472,
|
| 805 |
+
"32": 25472,
|
| 806 |
+
"320": 25472,
|
| 807 |
+
"321": 25472,
|
| 808 |
+
"322": 25472,
|
| 809 |
+
"323": 25472,
|
| 810 |
+
"324": 25472,
|
| 811 |
+
"325": 25472,
|
| 812 |
+
"326": 25472,
|
| 813 |
+
"327": 25472,
|
| 814 |
+
"328": 25472,
|
| 815 |
+
"329": 25472,
|
| 816 |
+
"33": 25472,
|
| 817 |
+
"330": 25472,
|
| 818 |
+
"331": 25472,
|
| 819 |
+
"332": 25472,
|
| 820 |
+
"333": 25472,
|
| 821 |
+
"334": 25472,
|
| 822 |
+
"335": 25472,
|
| 823 |
+
"336": 25472,
|
| 824 |
+
"337": 25472,
|
| 825 |
+
"338": 25472,
|
| 826 |
+
"339": 25472,
|
| 827 |
+
"34": 25472,
|
| 828 |
+
"340": 25472,
|
| 829 |
+
"341": 25472,
|
| 830 |
+
"342": 25472,
|
| 831 |
+
"343": 25472,
|
| 832 |
+
"344": 25472,
|
| 833 |
+
"345": 25472,
|
| 834 |
+
"346": 25472,
|
| 835 |
+
"347": 25472,
|
| 836 |
+
"348": 25472,
|
| 837 |
+
"349": 25472,
|
| 838 |
+
"35": 25472,
|
| 839 |
+
"350": 25472,
|
| 840 |
+
"351": 25472,
|
| 841 |
+
"352": 25472,
|
| 842 |
+
"353": 25472,
|
| 843 |
+
"354": 25472,
|
| 844 |
+
"355": 25472,
|
| 845 |
+
"356": 25472,
|
| 846 |
+
"357": 25472,
|
| 847 |
+
"358": 25472,
|
| 848 |
+
"359": 25472,
|
| 849 |
+
"36": 25472,
|
| 850 |
+
"360": 25472,
|
| 851 |
+
"361": 25472,
|
| 852 |
+
"362": 25472,
|
| 853 |
+
"363": 25472,
|
| 854 |
+
"364": 25472,
|
| 855 |
+
"365": 25472,
|
| 856 |
+
"366": 25472,
|
| 857 |
+
"367": 25472,
|
| 858 |
+
"368": 25472,
|
| 859 |
+
"369": 25472,
|
| 860 |
+
"37": 25472,
|
| 861 |
+
"370": 25472,
|
| 862 |
+
"371": 13134,
|
| 863 |
+
"372": 25472,
|
| 864 |
+
"373": 25472,
|
| 865 |
+
"374": 25472,
|
| 866 |
+
"375": 25472,
|
| 867 |
+
"376": 25472,
|
| 868 |
+
"377": 25472,
|
| 869 |
+
"378": 25472,
|
| 870 |
+
"379": 25472,
|
| 871 |
+
"38": 25472,
|
| 872 |
+
"380": 25472,
|
| 873 |
+
"381": 25472,
|
| 874 |
+
"382": 25472,
|
| 875 |
+
"383": 25472,
|
| 876 |
+
"384": 25472,
|
| 877 |
+
"385": 25472,
|
| 878 |
+
"386": 25472,
|
| 879 |
+
"387": 25472,
|
| 880 |
+
"388": 25472,
|
| 881 |
+
"389": 25472,
|
| 882 |
+
"39": 25472,
|
| 883 |
+
"390": 25472,
|
| 884 |
+
"391": 25472,
|
| 885 |
+
"392": 25472,
|
| 886 |
+
"393": 25472,
|
| 887 |
+
"394": 25472,
|
| 888 |
+
"395": 25472,
|
| 889 |
+
"396": 25472,
|
| 890 |
+
"397": 25472,
|
| 891 |
+
"398": 25472,
|
| 892 |
+
"399": 25472,
|
| 893 |
+
"4": 25472,
|
| 894 |
+
"40": 25472,
|
| 895 |
+
"400": 25472,
|
| 896 |
+
"401": 25472,
|
| 897 |
+
"402": 25472,
|
| 898 |
+
"403": 25472,
|
| 899 |
+
"404": 25472,
|
| 900 |
+
"405": 25472,
|
| 901 |
+
"406": 25472,
|
| 902 |
+
"407": 25472,
|
| 903 |
+
"408": 25472,
|
| 904 |
+
"409": 25472,
|
| 905 |
+
"41": 25472,
|
| 906 |
+
"410": 25472,
|
| 907 |
+
"411": 25472,
|
| 908 |
+
"412": 25472,
|
| 909 |
+
"413": 25472,
|
| 910 |
+
"414": 25472,
|
| 911 |
+
"415": 25472,
|
| 912 |
+
"416": 25472,
|
| 913 |
+
"417": 25472,
|
| 914 |
+
"418": 25472,
|
| 915 |
+
"419": 25472,
|
| 916 |
+
"42": 25472,
|
| 917 |
+
"420": 25472,
|
| 918 |
+
"421": 25472,
|
| 919 |
+
"422": 25472,
|
| 920 |
+
"423": 25472,
|
| 921 |
+
"424": 25472,
|
| 922 |
+
"425": 25472,
|
| 923 |
+
"426": 25472,
|
| 924 |
+
"427": 25472,
|
| 925 |
+
"428": 25472,
|
| 926 |
+
"429": 25472,
|
| 927 |
+
"43": 25472,
|
| 928 |
+
"430": 25472,
|
| 929 |
+
"431": 25472,
|
| 930 |
+
"432": 25472,
|
| 931 |
+
"433": 25472,
|
| 932 |
+
"434": 25472,
|
| 933 |
+
"435": 25472,
|
| 934 |
+
"436": 25472,
|
| 935 |
+
"437": 25472,
|
| 936 |
+
"438": 25472,
|
| 937 |
+
"439": 25472,
|
| 938 |
+
"44": 25472,
|
| 939 |
+
"440": 25472,
|
| 940 |
+
"441": 25472,
|
| 941 |
+
"442": 25472,
|
| 942 |
+
"443": 25472,
|
| 943 |
+
"444": 25472,
|
| 944 |
+
"445": 25472,
|
| 945 |
+
"446": 25472,
|
| 946 |
+
"447": 25472,
|
| 947 |
+
"448": 25472,
|
| 948 |
+
"449": 25472,
|
| 949 |
+
"45": 25472,
|
| 950 |
+
"450": 25472,
|
| 951 |
+
"451": 25472,
|
| 952 |
+
"452": 25472,
|
| 953 |
+
"453": 25472,
|
| 954 |
+
"454": 25472,
|
| 955 |
+
"455": 25472,
|
| 956 |
+
"456": 25472,
|
| 957 |
+
"457": 25472,
|
| 958 |
+
"458": 25472,
|
| 959 |
+
"459": 25472,
|
| 960 |
+
"46": 25472,
|
| 961 |
+
"460": 25472,
|
| 962 |
+
"461": 25472,
|
| 963 |
+
"462": 25472,
|
| 964 |
+
"463": 25472,
|
| 965 |
+
"464": 13134,
|
| 966 |
+
"47": 25472,
|
| 967 |
+
"48": 25472,
|
| 968 |
+
"49": 25472,
|
| 969 |
+
"5": 25472,
|
| 970 |
+
"50": 25472,
|
| 971 |
+
"51": 25472,
|
| 972 |
+
"52": 25472,
|
| 973 |
+
"53": 25472,
|
| 974 |
+
"54": 25472,
|
| 975 |
+
"55": 25472,
|
| 976 |
+
"56": 25472,
|
| 977 |
+
"57": 25472,
|
| 978 |
+
"58": 25472,
|
| 979 |
+
"59": 25472,
|
| 980 |
+
"6": 25472,
|
| 981 |
+
"60": 25472,
|
| 982 |
+
"61": 25472,
|
| 983 |
+
"62": 25472,
|
| 984 |
+
"63": 25472,
|
| 985 |
+
"64": 25472,
|
| 986 |
+
"65": 25472,
|
| 987 |
+
"66": 25472,
|
| 988 |
+
"67": 25472,
|
| 989 |
+
"68": 25472,
|
| 990 |
+
"69": 25472,
|
| 991 |
+
"7": 25472,
|
| 992 |
+
"70": 25472,
|
| 993 |
+
"71": 25472,
|
| 994 |
+
"72": 25472,
|
| 995 |
+
"73": 25472,
|
| 996 |
+
"74": 25472,
|
| 997 |
+
"75": 25472,
|
| 998 |
+
"76": 25472,
|
| 999 |
+
"77": 25472,
|
| 1000 |
+
"78": 25472,
|
| 1001 |
+
"79": 25472,
|
| 1002 |
+
"8": 25472,
|
| 1003 |
+
"80": 25472,
|
| 1004 |
+
"81": 25472,
|
| 1005 |
+
"82": 25472,
|
| 1006 |
+
"83": 25472,
|
| 1007 |
+
"84": 25472,
|
| 1008 |
+
"85": 25472,
|
| 1009 |
+
"86": 25472,
|
| 1010 |
+
"87": 25472,
|
| 1011 |
+
"88": 25472,
|
| 1012 |
+
"89": 25472,
|
| 1013 |
+
"9": 25472,
|
| 1014 |
+
"90": 25472,
|
| 1015 |
+
"91": 25472,
|
| 1016 |
+
"92": 13134,
|
| 1017 |
+
"93": 25472,
|
| 1018 |
+
"94": 25472,
|
| 1019 |
+
"95": 25472,
|
| 1020 |
+
"96": 25472,
|
| 1021 |
+
"97": 25472,
|
| 1022 |
+
"98": 25472,
|
| 1023 |
+
"99": 25472
|
| 1024 |
+
},
|
| 1025 |
+
"last_improvement_steps": 93,
|
| 1026 |
+
"last_increase_batch_size": 0,
|
| 1027 |
+
"last_increase_batch_size_eval_metric_improvement": 0,
|
| 1028 |
+
"last_increase_batch_size_steps": 0,
|
| 1029 |
+
"last_learning_rate_reduction": 0,
|
| 1030 |
+
"last_learning_rate_reduction_steps": 0,
|
| 1031 |
+
"learning_rate": 0.001,
|
| 1032 |
+
"llm_eval_examples": {},
|
| 1033 |
+
"num_increases_batch_size": 0,
|
| 1034 |
+
"num_reductions_learning_rate": 0,
|
| 1035 |
+
"steps": 465,
|
| 1036 |
+
"test_metrics": {
|
| 1037 |
+
"combined": {
|
| 1038 |
+
"loss": [
|
| 1039 |
+
[
|
| 1040 |
+
1,
|
| 1041 |
+
93,
|
| 1042 |
+
0.23742298781871796
|
| 1043 |
+
],
|
| 1044 |
+
[
|
| 1045 |
+
2,
|
| 1046 |
+
186,
|
| 1047 |
+
0.21031303703784943
|
| 1048 |
+
],
|
| 1049 |
+
[
|
| 1050 |
+
3,
|
| 1051 |
+
279,
|
| 1052 |
+
0.12192762643098831
|
| 1053 |
+
],
|
| 1054 |
+
[
|
| 1055 |
+
4,
|
| 1056 |
+
372,
|
| 1057 |
+
0.07570453733205795
|
| 1058 |
+
],
|
| 1059 |
+
[
|
| 1060 |
+
5,
|
| 1061 |
+
465,
|
| 1062 |
+
0.05710742995142937
|
| 1063 |
+
]
|
| 1064 |
+
]
|
| 1065 |
+
},
|
| 1066 |
+
"language": {
|
| 1067 |
+
"accuracy": [
|
| 1068 |
+
[
|
| 1069 |
+
1,
|
| 1070 |
+
93,
|
| 1071 |
+
0.924312949180603
|
| 1072 |
+
],
|
| 1073 |
+
[
|
| 1074 |
+
2,
|
| 1075 |
+
186,
|
| 1076 |
+
0.9528379440307617
|
| 1077 |
+
],
|
| 1078 |
+
[
|
| 1079 |
+
3,
|
| 1080 |
+
279,
|
| 1081 |
+
0.961493730545044
|
| 1082 |
+
],
|
| 1083 |
+
[
|
| 1084 |
+
4,
|
| 1085 |
+
372,
|
| 1086 |
+
0.9817298650741577
|
| 1087 |
+
],
|
| 1088 |
+
[
|
| 1089 |
+
5,
|
| 1090 |
+
465,
|
| 1091 |
+
0.9708698987960815
|
| 1092 |
+
]
|
| 1093 |
+
],
|
| 1094 |
+
"accuracy_micro": [
|
| 1095 |
+
[
|
| 1096 |
+
1,
|
| 1097 |
+
93,
|
| 1098 |
+
0.9312320947647095
|
| 1099 |
+
],
|
| 1100 |
+
[
|
| 1101 |
+
2,
|
| 1102 |
+
186,
|
| 1103 |
+
0.9484240412712097
|
| 1104 |
+
],
|
| 1105 |
+
[
|
| 1106 |
+
3,
|
| 1107 |
+
279,
|
| 1108 |
+
0.9598853588104248
|
| 1109 |
+
],
|
| 1110 |
+
[
|
| 1111 |
+
4,
|
| 1112 |
+
372,
|
| 1113 |
+
0.9799426794052124
|
| 1114 |
+
],
|
| 1115 |
+
[
|
| 1116 |
+
5,
|
| 1117 |
+
465,
|
| 1118 |
+
0.9742120504379272
|
| 1119 |
+
]
|
| 1120 |
+
],
|
| 1121 |
+
"loss": [
|
| 1122 |
+
[
|
| 1123 |
+
1,
|
| 1124 |
+
93,
|
| 1125 |
+
0.23742298781871796
|
| 1126 |
+
],
|
| 1127 |
+
[
|
| 1128 |
+
2,
|
| 1129 |
+
186,
|
| 1130 |
+
0.21031303703784943
|
| 1131 |
+
],
|
| 1132 |
+
[
|
| 1133 |
+
3,
|
| 1134 |
+
279,
|
| 1135 |
+
0.12192762643098831
|
| 1136 |
+
],
|
| 1137 |
+
[
|
| 1138 |
+
4,
|
| 1139 |
+
372,
|
| 1140 |
+
0.07570453733205795
|
| 1141 |
+
],
|
| 1142 |
+
[
|
| 1143 |
+
5,
|
| 1144 |
+
465,
|
| 1145 |
+
0.05710742995142937
|
| 1146 |
+
]
|
| 1147 |
+
],
|
| 1148 |
+
"roc_auc": [
|
| 1149 |
+
[
|
| 1150 |
+
1,
|
| 1151 |
+
93,
|
| 1152 |
+
0.9985259175300598
|
| 1153 |
+
],
|
| 1154 |
+
[
|
| 1155 |
+
2,
|
| 1156 |
+
186,
|
| 1157 |
+
0.9976196885108948
|
| 1158 |
+
],
|
| 1159 |
+
[
|
| 1160 |
+
3,
|
| 1161 |
+
279,
|
| 1162 |
+
0.9982259273529053
|
| 1163 |
+
],
|
| 1164 |
+
[
|
| 1165 |
+
4,
|
| 1166 |
+
372,
|
| 1167 |
+
0.999114990234375
|
| 1168 |
+
],
|
| 1169 |
+
[
|
| 1170 |
+
5,
|
| 1171 |
+
465,
|
| 1172 |
+
0.9991633296012878
|
| 1173 |
+
]
|
| 1174 |
+
]
|
| 1175 |
+
}
|
| 1176 |
+
},
|
| 1177 |
+
"total_tokens_used": 11782790,
|
| 1178 |
+
"train_metrics": {
|
| 1179 |
+
"combined": {
|
| 1180 |
+
"loss": [
|
| 1181 |
+
[
|
| 1182 |
+
1,
|
| 1183 |
+
93,
|
| 1184 |
+
2.6028196811676025
|
| 1185 |
+
],
|
| 1186 |
+
[
|
| 1187 |
+
2,
|
| 1188 |
+
186,
|
| 1189 |
+
0.05241914466023445
|
| 1190 |
+
],
|
| 1191 |
+
[
|
| 1192 |
+
3,
|
| 1193 |
+
279,
|
| 1194 |
+
0.1079530417919159
|
| 1195 |
+
],
|
| 1196 |
+
[
|
| 1197 |
+
4,
|
| 1198 |
+
372,
|
| 1199 |
+
0.03008267655968666
|
| 1200 |
+
],
|
| 1201 |
+
[
|
| 1202 |
+
5,
|
| 1203 |
+
465,
|
| 1204 |
+
0.020476186648011208
|
| 1205 |
+
]
|
| 1206 |
+
]
|
| 1207 |
+
},
|
| 1208 |
+
"language": {
|
| 1209 |
+
"accuracy": [
|
| 1210 |
+
[
|
| 1211 |
+
1,
|
| 1212 |
+
93,
|
| 1213 |
+
0.709525465965271
|
| 1214 |
+
],
|
| 1215 |
+
[
|
| 1216 |
+
2,
|
| 1217 |
+
186,
|
| 1218 |
+
0.9832712411880493
|
| 1219 |
+
],
|
| 1220 |
+
[
|
| 1221 |
+
3,
|
| 1222 |
+
279,
|
| 1223 |
+
0.9843869209289551
|
| 1224 |
+
],
|
| 1225 |
+
[
|
| 1226 |
+
4,
|
| 1227 |
+
372,
|
| 1228 |
+
0.9913396835327148
|
| 1229 |
+
],
|
| 1230 |
+
[
|
| 1231 |
+
5,
|
| 1232 |
+
465,
|
| 1233 |
+
0.9936301708221436
|
| 1234 |
+
]
|
| 1235 |
+
],
|
| 1236 |
+
"accuracy_micro": [
|
| 1237 |
+
[
|
| 1238 |
+
1,
|
| 1239 |
+
93,
|
| 1240 |
+
0.7127174735069275
|
| 1241 |
+
],
|
| 1242 |
+
[
|
| 1243 |
+
2,
|
| 1244 |
+
186,
|
| 1245 |
+
0.9831109642982483
|
| 1246 |
+
],
|
| 1247 |
+
[
|
| 1248 |
+
3,
|
| 1249 |
+
279,
|
| 1250 |
+
0.9853065609931946
|
| 1251 |
+
],
|
| 1252 |
+
[
|
| 1253 |
+
4,
|
| 1254 |
+
372,
|
| 1255 |
+
0.9910488128662109
|
| 1256 |
+
],
|
| 1257 |
+
[
|
| 1258 |
+
5,
|
| 1259 |
+
465,
|
| 1260 |
+
0.9930754899978638
|
| 1261 |
+
]
|
| 1262 |
+
],
|
| 1263 |
+
"loss": [
|
| 1264 |
+
[
|
| 1265 |
+
1,
|
| 1266 |
+
93,
|
| 1267 |
+
2.6028196811676025
|
| 1268 |
+
],
|
| 1269 |
+
[
|
| 1270 |
+
2,
|
| 1271 |
+
186,
|
| 1272 |
+
0.05241914466023445
|
| 1273 |
+
],
|
| 1274 |
+
[
|
| 1275 |
+
3,
|
| 1276 |
+
279,
|
| 1277 |
+
0.1079530417919159
|
| 1278 |
+
],
|
| 1279 |
+
[
|
| 1280 |
+
4,
|
| 1281 |
+
372,
|
| 1282 |
+
0.03008267655968666
|
| 1283 |
+
],
|
| 1284 |
+
[
|
| 1285 |
+
5,
|
| 1286 |
+
465,
|
| 1287 |
+
0.020476186648011208
|
| 1288 |
+
]
|
| 1289 |
+
],
|
| 1290 |
+
"roc_auc": [
|
| 1291 |
+
[
|
| 1292 |
+
1,
|
| 1293 |
+
93,
|
| 1294 |
+
0.8537933826446533
|
| 1295 |
+
],
|
| 1296 |
+
[
|
| 1297 |
+
2,
|
| 1298 |
+
186,
|
| 1299 |
+
0.9993903636932373
|
| 1300 |
+
],
|
| 1301 |
+
[
|
| 1302 |
+
3,
|
| 1303 |
+
279,
|
| 1304 |
+
0.9988884925842285
|
| 1305 |
+
],
|
| 1306 |
+
[
|
| 1307 |
+
4,
|
| 1308 |
+
372,
|
| 1309 |
+
0.9997660517692566
|
| 1310 |
+
],
|
| 1311 |
+
[
|
| 1312 |
+
5,
|
| 1313 |
+
465,
|
| 1314 |
+
0.9998530745506287
|
| 1315 |
+
]
|
| 1316 |
+
]
|
| 1317 |
+
}
|
| 1318 |
+
},
|
| 1319 |
+
"tune_checkpoint_num": 0,
|
| 1320 |
+
"validation_metrics": {
|
| 1321 |
+
"combined": {
|
| 1322 |
+
"loss": [
|
| 1323 |
+
[
|
| 1324 |
+
1,
|
| 1325 |
+
93,
|
| 1326 |
+
0.2647719979286194
|
| 1327 |
+
],
|
| 1328 |
+
[
|
| 1329 |
+
2,
|
| 1330 |
+
186,
|
| 1331 |
+
0.27919644117355347
|
| 1332 |
+
],
|
| 1333 |
+
[
|
| 1334 |
+
3,
|
| 1335 |
+
279,
|
| 1336 |
+
0.1389603614807129
|
| 1337 |
+
],
|
| 1338 |
+
[
|
| 1339 |
+
4,
|
| 1340 |
+
372,
|
| 1341 |
+
0.06920339167118073
|
| 1342 |
+
],
|
| 1343 |
+
[
|
| 1344 |
+
5,
|
| 1345 |
+
465,
|
| 1346 |
+
0.07529747486114502
|
| 1347 |
+
]
|
| 1348 |
+
]
|
| 1349 |
+
},
|
| 1350 |
+
"language": {
|
| 1351 |
+
"accuracy": [
|
| 1352 |
+
[
|
| 1353 |
+
1,
|
| 1354 |
+
93,
|
| 1355 |
+
0.911348819732666
|
| 1356 |
+
],
|
| 1357 |
+
[
|
| 1358 |
+
2,
|
| 1359 |
+
186,
|
| 1360 |
+
0.9450980424880981
|
| 1361 |
+
],
|
| 1362 |
+
[
|
| 1363 |
+
3,
|
| 1364 |
+
279,
|
| 1365 |
+
0.9551990628242493
|
| 1366 |
+
],
|
| 1367 |
+
[
|
| 1368 |
+
4,
|
| 1369 |
+
372,
|
| 1370 |
+
0.9799168109893799
|
| 1371 |
+
],
|
| 1372 |
+
[
|
| 1373 |
+
5,
|
| 1374 |
+
465,
|
| 1375 |
+
0.9636363983154297
|
| 1376 |
+
]
|
| 1377 |
+
],
|
| 1378 |
+
"accuracy_micro": [
|
| 1379 |
+
[
|
| 1380 |
+
1,
|
| 1381 |
+
93,
|
| 1382 |
+
0.9225251078605652
|
| 1383 |
+
],
|
| 1384 |
+
[
|
| 1385 |
+
2,
|
| 1386 |
+
186,
|
| 1387 |
+
0.9397417306900024
|
| 1388 |
+
],
|
| 1389 |
+
[
|
| 1390 |
+
3,
|
| 1391 |
+
279,
|
| 1392 |
+
0.955523669719696
|
| 1393 |
+
],
|
| 1394 |
+
[
|
| 1395 |
+
4,
|
| 1396 |
+
372,
|
| 1397 |
+
0.9784792065620422
|
| 1398 |
+
],
|
| 1399 |
+
[
|
| 1400 |
+
5,
|
| 1401 |
+
465,
|
| 1402 |
+
0.9684361815452576
|
| 1403 |
+
]
|
| 1404 |
+
],
|
| 1405 |
+
"loss": [
|
| 1406 |
+
[
|
| 1407 |
+
1,
|
| 1408 |
+
93,
|
| 1409 |
+
0.2647719979286194
|
| 1410 |
+
],
|
| 1411 |
+
[
|
| 1412 |
+
2,
|
| 1413 |
+
186,
|
| 1414 |
+
0.27919644117355347
|
| 1415 |
+
],
|
| 1416 |
+
[
|
| 1417 |
+
3,
|
| 1418 |
+
279,
|
| 1419 |
+
0.1389603614807129
|
| 1420 |
+
],
|
| 1421 |
+
[
|
| 1422 |
+
4,
|
| 1423 |
+
372,
|
| 1424 |
+
0.06920339167118073
|
| 1425 |
+
],
|
| 1426 |
+
[
|
| 1427 |
+
5,
|
| 1428 |
+
465,
|
| 1429 |
+
0.07529747486114502
|
| 1430 |
+
]
|
| 1431 |
+
],
|
| 1432 |
+
"roc_auc": [
|
| 1433 |
+
[
|
| 1434 |
+
1,
|
| 1435 |
+
93,
|
| 1436 |
+
0.9981002807617188
|
| 1437 |
+
],
|
| 1438 |
+
[
|
| 1439 |
+
2,
|
| 1440 |
+
186,
|
| 1441 |
+
0.99806147813797
|
| 1442 |
+
],
|
| 1443 |
+
[
|
| 1444 |
+
3,
|
| 1445 |
+
279,
|
| 1446 |
+
0.9984046816825867
|
| 1447 |
+
],
|
| 1448 |
+
[
|
| 1449 |
+
4,
|
| 1450 |
+
372,
|
| 1451 |
+
0.998852550983429
|
| 1452 |
+
],
|
| 1453 |
+
[
|
| 1454 |
+
5,
|
| 1455 |
+
465,
|
| 1456 |
+
0.9990896582603455
|
| 1457 |
+
]
|
| 1458 |
+
]
|
| 1459 |
+
}
|
| 1460 |
+
}
|
| 1461 |
+
}
|
model/training_set_metadata.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
training_statistics.json
ADDED
|
@@ -0,0 +1,129 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"evaluation_frequency": {
|
| 3 |
+
"frequency": 1,
|
| 4 |
+
"period": "epoch"
|
| 5 |
+
},
|
| 6 |
+
"test": {
|
| 7 |
+
"combined": {
|
| 8 |
+
"loss": [
|
| 9 |
+
0.23742298781871796,
|
| 10 |
+
0.21031303703784943,
|
| 11 |
+
0.12192762643098831,
|
| 12 |
+
0.07570453733205795,
|
| 13 |
+
0.05710742995142937
|
| 14 |
+
]
|
| 15 |
+
},
|
| 16 |
+
"language": {
|
| 17 |
+
"accuracy": [
|
| 18 |
+
0.924312949180603,
|
| 19 |
+
0.9528379440307617,
|
| 20 |
+
0.961493730545044,
|
| 21 |
+
0.9817298650741577,
|
| 22 |
+
0.9708698987960815
|
| 23 |
+
],
|
| 24 |
+
"accuracy_micro": [
|
| 25 |
+
0.9312320947647095,
|
| 26 |
+
0.9484240412712097,
|
| 27 |
+
0.9598853588104248,
|
| 28 |
+
0.9799426794052124,
|
| 29 |
+
0.9742120504379272
|
| 30 |
+
],
|
| 31 |
+
"loss": [
|
| 32 |
+
0.23742298781871796,
|
| 33 |
+
0.21031303703784943,
|
| 34 |
+
0.12192762643098831,
|
| 35 |
+
0.07570453733205795,
|
| 36 |
+
0.05710742995142937
|
| 37 |
+
],
|
| 38 |
+
"roc_auc": [
|
| 39 |
+
0.9985259175300598,
|
| 40 |
+
0.9976196885108948,
|
| 41 |
+
0.9982259273529053,
|
| 42 |
+
0.999114990234375,
|
| 43 |
+
0.9991633296012878
|
| 44 |
+
]
|
| 45 |
+
}
|
| 46 |
+
},
|
| 47 |
+
"training": {
|
| 48 |
+
"combined": {
|
| 49 |
+
"loss": [
|
| 50 |
+
2.6028196811676025,
|
| 51 |
+
0.05241914466023445,
|
| 52 |
+
0.1079530417919159,
|
| 53 |
+
0.03008267655968666,
|
| 54 |
+
0.020476186648011208
|
| 55 |
+
]
|
| 56 |
+
},
|
| 57 |
+
"language": {
|
| 58 |
+
"accuracy": [
|
| 59 |
+
0.709525465965271,
|
| 60 |
+
0.9832712411880493,
|
| 61 |
+
0.9843869209289551,
|
| 62 |
+
0.9913396835327148,
|
| 63 |
+
0.9936301708221436
|
| 64 |
+
],
|
| 65 |
+
"accuracy_micro": [
|
| 66 |
+
0.7127174735069275,
|
| 67 |
+
0.9831109642982483,
|
| 68 |
+
0.9853065609931946,
|
| 69 |
+
0.9910488128662109,
|
| 70 |
+
0.9930754899978638
|
| 71 |
+
],
|
| 72 |
+
"loss": [
|
| 73 |
+
2.6028196811676025,
|
| 74 |
+
0.05241914466023445,
|
| 75 |
+
0.1079530417919159,
|
| 76 |
+
0.03008267655968666,
|
| 77 |
+
0.020476186648011208
|
| 78 |
+
],
|
| 79 |
+
"roc_auc": [
|
| 80 |
+
0.8537933826446533,
|
| 81 |
+
0.9993903636932373,
|
| 82 |
+
0.9988884925842285,
|
| 83 |
+
0.9997660517692566,
|
| 84 |
+
0.9998530745506287
|
| 85 |
+
]
|
| 86 |
+
}
|
| 87 |
+
},
|
| 88 |
+
"validation": {
|
| 89 |
+
"combined": {
|
| 90 |
+
"loss": [
|
| 91 |
+
0.2647719979286194,
|
| 92 |
+
0.27919644117355347,
|
| 93 |
+
0.1389603614807129,
|
| 94 |
+
0.06920339167118073,
|
| 95 |
+
0.07529747486114502
|
| 96 |
+
]
|
| 97 |
+
},
|
| 98 |
+
"language": {
|
| 99 |
+
"accuracy": [
|
| 100 |
+
0.911348819732666,
|
| 101 |
+
0.9450980424880981,
|
| 102 |
+
0.9551990628242493,
|
| 103 |
+
0.9799168109893799,
|
| 104 |
+
0.9636363983154297
|
| 105 |
+
],
|
| 106 |
+
"accuracy_micro": [
|
| 107 |
+
0.9225251078605652,
|
| 108 |
+
0.9397417306900024,
|
| 109 |
+
0.955523669719696,
|
| 110 |
+
0.9784792065620422,
|
| 111 |
+
0.9684361815452576
|
| 112 |
+
],
|
| 113 |
+
"loss": [
|
| 114 |
+
0.2647719979286194,
|
| 115 |
+
0.27919644117355347,
|
| 116 |
+
0.1389603614807129,
|
| 117 |
+
0.06920339167118073,
|
| 118 |
+
0.07529747486114502
|
| 119 |
+
],
|
| 120 |
+
"roc_auc": [
|
| 121 |
+
0.9981002807617188,
|
| 122 |
+
0.99806147813797,
|
| 123 |
+
0.9984046816825867,
|
| 124 |
+
0.998852550983429,
|
| 125 |
+
0.9990896582603455
|
| 126 |
+
]
|
| 127 |
+
}
|
| 128 |
+
}
|
| 129 |
+
}
|