Upload 9 files
Browse files- config.json +109 -0
- optimizer.pt +3 -0
- preprocessor_config.json +10 -0
- pytorch_model.bin +3 -0
- rng_state.pth +3 -0
- scaler.pt +3 -0
- scheduler.pt +3 -0
- trainer_state.json +3676 -0
- training_args.bin +3 -0
config.json
ADDED
|
@@ -0,0 +1,109 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "pburub/Aramaic_pretrained_XLSR",
|
| 3 |
+
"activation_dropout": 0.0,
|
| 4 |
+
"adapter_kernel_size": 3,
|
| 5 |
+
"adapter_stride": 2,
|
| 6 |
+
"add_adapter": false,
|
| 7 |
+
"apply_spec_augment": true,
|
| 8 |
+
"architectures": [
|
| 9 |
+
"Wav2Vec2ForCTC"
|
| 10 |
+
],
|
| 11 |
+
"attention_dropout": 0.1,
|
| 12 |
+
"bos_token_id": 1,
|
| 13 |
+
"classifier_proj_size": 256,
|
| 14 |
+
"codevector_dim": 768,
|
| 15 |
+
"contrastive_logits_temperature": 0.1,
|
| 16 |
+
"conv_bias": true,
|
| 17 |
+
"conv_dim": [
|
| 18 |
+
512,
|
| 19 |
+
512,
|
| 20 |
+
512,
|
| 21 |
+
512,
|
| 22 |
+
512,
|
| 23 |
+
512,
|
| 24 |
+
512
|
| 25 |
+
],
|
| 26 |
+
"conv_kernel": [
|
| 27 |
+
10,
|
| 28 |
+
3,
|
| 29 |
+
3,
|
| 30 |
+
3,
|
| 31 |
+
3,
|
| 32 |
+
2,
|
| 33 |
+
2
|
| 34 |
+
],
|
| 35 |
+
"conv_stride": [
|
| 36 |
+
5,
|
| 37 |
+
2,
|
| 38 |
+
2,
|
| 39 |
+
2,
|
| 40 |
+
2,
|
| 41 |
+
2,
|
| 42 |
+
2
|
| 43 |
+
],
|
| 44 |
+
"ctc_loss_reduction": "mean",
|
| 45 |
+
"ctc_zero_infinity": true,
|
| 46 |
+
"diversity_loss_weight": 0.1,
|
| 47 |
+
"do_stable_layer_norm": true,
|
| 48 |
+
"eos_token_id": 2,
|
| 49 |
+
"feat_extract_activation": "gelu",
|
| 50 |
+
"feat_extract_dropout": 0.0,
|
| 51 |
+
"feat_extract_norm": "layer",
|
| 52 |
+
"feat_proj_dropout": 0.1,
|
| 53 |
+
"feat_quantizer_dropout": 0.0,
|
| 54 |
+
"final_dropout": 0.0,
|
| 55 |
+
"gradient_checkpointing": false,
|
| 56 |
+
"hidden_act": "gelu",
|
| 57 |
+
"hidden_dropout": 0.1,
|
| 58 |
+
"hidden_size": 1024,
|
| 59 |
+
"initializer_range": 0.02,
|
| 60 |
+
"intermediate_size": 4096,
|
| 61 |
+
"layer_norm_eps": 1e-05,
|
| 62 |
+
"layerdrop": 0.1,
|
| 63 |
+
"mask_feature_length": 10,
|
| 64 |
+
"mask_feature_min_masks": 0,
|
| 65 |
+
"mask_feature_prob": 0.004,
|
| 66 |
+
"mask_time_length": 10,
|
| 67 |
+
"mask_time_min_masks": 2,
|
| 68 |
+
"mask_time_prob": 0.075,
|
| 69 |
+
"model_type": "wav2vec2",
|
| 70 |
+
"num_adapter_layers": 3,
|
| 71 |
+
"num_attention_heads": 16,
|
| 72 |
+
"num_codevector_groups": 2,
|
| 73 |
+
"num_codevectors_per_group": 320,
|
| 74 |
+
"num_conv_pos_embedding_groups": 16,
|
| 75 |
+
"num_conv_pos_embeddings": 128,
|
| 76 |
+
"num_feat_extract_layers": 7,
|
| 77 |
+
"num_hidden_layers": 24,
|
| 78 |
+
"num_negatives": 100,
|
| 79 |
+
"output_hidden_size": 1024,
|
| 80 |
+
"pad_token_id": 0,
|
| 81 |
+
"pretrained_model_name_or_path": "pburub/Aramaic_pretrained_XLSR",
|
| 82 |
+
"proj_codevector_dim": 768,
|
| 83 |
+
"tdnn_dilation": [
|
| 84 |
+
1,
|
| 85 |
+
2,
|
| 86 |
+
3,
|
| 87 |
+
1,
|
| 88 |
+
1
|
| 89 |
+
],
|
| 90 |
+
"tdnn_dim": [
|
| 91 |
+
512,
|
| 92 |
+
512,
|
| 93 |
+
512,
|
| 94 |
+
512,
|
| 95 |
+
1500
|
| 96 |
+
],
|
| 97 |
+
"tdnn_kernel": [
|
| 98 |
+
5,
|
| 99 |
+
3,
|
| 100 |
+
3,
|
| 101 |
+
1,
|
| 102 |
+
1
|
| 103 |
+
],
|
| 104 |
+
"torch_dtype": "float32",
|
| 105 |
+
"transformers_version": "4.24.0",
|
| 106 |
+
"use_weighted_layer_sum": false,
|
| 107 |
+
"vocab_size": 48,
|
| 108 |
+
"xvector_output_dim": 512
|
| 109 |
+
}
|
optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b463dd898106b0180cd5e433f6614aa259f2c820aa060db73db91badb140d7ef
|
| 3 |
+
size 2490552961
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_normalize": true,
|
| 3 |
+
"feature_extractor_type": "Wav2Vec2FeatureExtractor",
|
| 4 |
+
"feature_size": 1,
|
| 5 |
+
"padding_side": "right",
|
| 6 |
+
"padding_value": 0.0,
|
| 7 |
+
"processor_class": "Wav2Vec2Processor",
|
| 8 |
+
"return_attention_mask": true,
|
| 9 |
+
"sampling_rate": 16000
|
| 10 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8b99ee47a081a103769022e30f6bd07395ac961331bc51cf8afe227523f4203
|
| 3 |
+
size 1262098669
|
rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0c43b3d0a497a6699c5cd805e472c274dcb5632232ba5bc1779f6f468c008c7
|
| 3 |
+
size 14575
|
scaler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3f1fde7f9b7eec0cf45b3bfcdbe164d4fcb042888d9f30ac54c0e23cae91bc0
|
| 3 |
+
size 557
|
scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e1efc198b70d31200195b5d99199200293c8fa41295dea3d4d268580a032b3c5
|
| 3 |
+
size 627
|
trainer_state.json
ADDED
|
@@ -0,0 +1,3676 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.35683760683760685,
|
| 3 |
+
"best_model_checkpoint": "/content/drive/MyDrive/model_ft/checkpoint-60000",
|
| 4 |
+
"epoch": 49.01959583588487,
|
| 5 |
+
"global_step": 60000,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.08,
|
| 12 |
+
"learning_rate": 8e-08,
|
| 13 |
+
"loss": 17.0818,
|
| 14 |
+
"step": 100
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.16,
|
| 18 |
+
"learning_rate": 1.6333333333333334e-07,
|
| 19 |
+
"loss": 15.2157,
|
| 20 |
+
"step": 200
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.24,
|
| 24 |
+
"learning_rate": 2.466666666666667e-07,
|
| 25 |
+
"loss": 15.0857,
|
| 26 |
+
"step": 300
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.33,
|
| 30 |
+
"learning_rate": 3.3e-07,
|
| 31 |
+
"loss": 15.4181,
|
| 32 |
+
"step": 400
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.41,
|
| 36 |
+
"learning_rate": 4.133333333333334e-07,
|
| 37 |
+
"loss": 15.3673,
|
| 38 |
+
"step": 500
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.49,
|
| 42 |
+
"learning_rate": 4.958333333333334e-07,
|
| 43 |
+
"loss": 15.3838,
|
| 44 |
+
"step": 600
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.57,
|
| 48 |
+
"learning_rate": 5.791666666666667e-07,
|
| 49 |
+
"loss": 15.0048,
|
| 50 |
+
"step": 700
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.65,
|
| 54 |
+
"learning_rate": 6.625000000000001e-07,
|
| 55 |
+
"loss": 13.8472,
|
| 56 |
+
"step": 800
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.73,
|
| 60 |
+
"learning_rate": 7.458333333333334e-07,
|
| 61 |
+
"loss": 14.4646,
|
| 62 |
+
"step": 900
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 0.82,
|
| 66 |
+
"learning_rate": 8.291666666666667e-07,
|
| 67 |
+
"loss": 11.767,
|
| 68 |
+
"step": 1000
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.9,
|
| 72 |
+
"learning_rate": 9.125e-07,
|
| 73 |
+
"loss": 10.0359,
|
| 74 |
+
"step": 1100
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 0.98,
|
| 78 |
+
"learning_rate": 9.958333333333335e-07,
|
| 79 |
+
"loss": 8.9849,
|
| 80 |
+
"step": 1200
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 1.06,
|
| 84 |
+
"learning_rate": 1.0791666666666668e-06,
|
| 85 |
+
"loss": 7.6915,
|
| 86 |
+
"step": 1300
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 1.14,
|
| 90 |
+
"learning_rate": 1.1625e-06,
|
| 91 |
+
"loss": 7.4909,
|
| 92 |
+
"step": 1400
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 1.23,
|
| 96 |
+
"learning_rate": 1.2458333333333334e-06,
|
| 97 |
+
"loss": 6.3382,
|
| 98 |
+
"step": 1500
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 1.31,
|
| 102 |
+
"learning_rate": 1.3291666666666667e-06,
|
| 103 |
+
"loss": 6.2147,
|
| 104 |
+
"step": 1600
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 1.39,
|
| 108 |
+
"learning_rate": 1.4125e-06,
|
| 109 |
+
"loss": 5.7939,
|
| 110 |
+
"step": 1700
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 1.47,
|
| 114 |
+
"learning_rate": 1.4958333333333336e-06,
|
| 115 |
+
"loss": 5.4857,
|
| 116 |
+
"step": 1800
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.55,
|
| 120 |
+
"learning_rate": 1.5791666666666669e-06,
|
| 121 |
+
"loss": 5.3803,
|
| 122 |
+
"step": 1900
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 1.63,
|
| 126 |
+
"learning_rate": 1.6625000000000002e-06,
|
| 127 |
+
"loss": 5.2082,
|
| 128 |
+
"step": 2000
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 1.72,
|
| 132 |
+
"learning_rate": 1.7458333333333335e-06,
|
| 133 |
+
"loss": 4.7779,
|
| 134 |
+
"step": 2100
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"epoch": 1.8,
|
| 138 |
+
"learning_rate": 1.8291666666666668e-06,
|
| 139 |
+
"loss": 4.7639,
|
| 140 |
+
"step": 2200
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 1.88,
|
| 144 |
+
"learning_rate": 1.9125000000000003e-06,
|
| 145 |
+
"loss": 4.516,
|
| 146 |
+
"step": 2300
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"epoch": 1.96,
|
| 150 |
+
"learning_rate": 1.9958333333333334e-06,
|
| 151 |
+
"loss": 4.5625,
|
| 152 |
+
"step": 2400
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 2.04,
|
| 156 |
+
"learning_rate": 2.079166666666667e-06,
|
| 157 |
+
"loss": 4.5228,
|
| 158 |
+
"step": 2500
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 2.12,
|
| 162 |
+
"learning_rate": 2.1625e-06,
|
| 163 |
+
"loss": 4.2219,
|
| 164 |
+
"step": 2600
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 2.21,
|
| 168 |
+
"learning_rate": 2.2458333333333336e-06,
|
| 169 |
+
"loss": 4.1294,
|
| 170 |
+
"step": 2700
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 2.29,
|
| 174 |
+
"learning_rate": 2.3291666666666667e-06,
|
| 175 |
+
"loss": 4.2024,
|
| 176 |
+
"step": 2800
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 2.37,
|
| 180 |
+
"learning_rate": 2.4125e-06,
|
| 181 |
+
"loss": 3.9283,
|
| 182 |
+
"step": 2900
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 2.45,
|
| 186 |
+
"learning_rate": 2.4958333333333333e-06,
|
| 187 |
+
"loss": 3.8443,
|
| 188 |
+
"step": 3000
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.53,
|
| 192 |
+
"learning_rate": 2.5791666666666672e-06,
|
| 193 |
+
"loss": 3.8079,
|
| 194 |
+
"step": 3100
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 2.61,
|
| 198 |
+
"learning_rate": 2.6625e-06,
|
| 199 |
+
"loss": 3.6979,
|
| 200 |
+
"step": 3200
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 2.7,
|
| 204 |
+
"learning_rate": 2.745833333333334e-06,
|
| 205 |
+
"loss": 3.7727,
|
| 206 |
+
"step": 3300
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 2.78,
|
| 210 |
+
"learning_rate": 2.8291666666666665e-06,
|
| 211 |
+
"loss": 3.5884,
|
| 212 |
+
"step": 3400
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 2.86,
|
| 216 |
+
"learning_rate": 2.9125000000000005e-06,
|
| 217 |
+
"loss": 3.6409,
|
| 218 |
+
"step": 3500
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 2.94,
|
| 222 |
+
"learning_rate": 2.995833333333333e-06,
|
| 223 |
+
"loss": 3.5832,
|
| 224 |
+
"step": 3600
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 3.02,
|
| 228 |
+
"learning_rate": 3.079166666666667e-06,
|
| 229 |
+
"loss": 3.5055,
|
| 230 |
+
"step": 3700
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"epoch": 3.1,
|
| 234 |
+
"learning_rate": 3.1625000000000002e-06,
|
| 235 |
+
"loss": 3.5141,
|
| 236 |
+
"step": 3800
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 3.19,
|
| 240 |
+
"learning_rate": 3.2458333333333337e-06,
|
| 241 |
+
"loss": 3.4218,
|
| 242 |
+
"step": 3900
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 3.27,
|
| 246 |
+
"learning_rate": 3.3291666666666673e-06,
|
| 247 |
+
"loss": 3.4604,
|
| 248 |
+
"step": 4000
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 3.35,
|
| 252 |
+
"learning_rate": 3.4125000000000004e-06,
|
| 253 |
+
"loss": 3.4134,
|
| 254 |
+
"step": 4100
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 3.43,
|
| 258 |
+
"learning_rate": 3.495833333333334e-06,
|
| 259 |
+
"loss": 3.3811,
|
| 260 |
+
"step": 4200
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 3.51,
|
| 264 |
+
"learning_rate": 3.579166666666667e-06,
|
| 265 |
+
"loss": 3.4506,
|
| 266 |
+
"step": 4300
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 3.59,
|
| 270 |
+
"learning_rate": 3.6625000000000005e-06,
|
| 271 |
+
"loss": 3.3525,
|
| 272 |
+
"step": 4400
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 3.68,
|
| 276 |
+
"learning_rate": 3.7458333333333336e-06,
|
| 277 |
+
"loss": 3.3783,
|
| 278 |
+
"step": 4500
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"epoch": 3.76,
|
| 282 |
+
"learning_rate": 3.829166666666667e-06,
|
| 283 |
+
"loss": 3.3088,
|
| 284 |
+
"step": 4600
|
| 285 |
+
},
|
| 286 |
+
{
|
| 287 |
+
"epoch": 3.84,
|
| 288 |
+
"learning_rate": 3.9125e-06,
|
| 289 |
+
"loss": 3.3303,
|
| 290 |
+
"step": 4700
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 3.92,
|
| 294 |
+
"learning_rate": 3.995833333333334e-06,
|
| 295 |
+
"loss": 3.3213,
|
| 296 |
+
"step": 4800
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 4.0,
|
| 300 |
+
"learning_rate": 4.0791666666666664e-06,
|
| 301 |
+
"loss": 3.3047,
|
| 302 |
+
"step": 4900
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 4.08,
|
| 306 |
+
"learning_rate": 4.1625e-06,
|
| 307 |
+
"loss": 3.2773,
|
| 308 |
+
"step": 5000
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"epoch": 4.17,
|
| 312 |
+
"learning_rate": 4.2458333333333335e-06,
|
| 313 |
+
"loss": 3.2719,
|
| 314 |
+
"step": 5100
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 4.25,
|
| 318 |
+
"learning_rate": 4.3291666666666674e-06,
|
| 319 |
+
"loss": 3.2904,
|
| 320 |
+
"step": 5200
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 4.33,
|
| 324 |
+
"learning_rate": 4.4125000000000005e-06,
|
| 325 |
+
"loss": 3.2518,
|
| 326 |
+
"step": 5300
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 4.41,
|
| 330 |
+
"learning_rate": 4.495833333333334e-06,
|
| 331 |
+
"loss": 3.2756,
|
| 332 |
+
"step": 5400
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 4.49,
|
| 336 |
+
"learning_rate": 4.5791666666666676e-06,
|
| 337 |
+
"loss": 3.3096,
|
| 338 |
+
"step": 5500
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 4.57,
|
| 342 |
+
"learning_rate": 4.662500000000001e-06,
|
| 343 |
+
"loss": 3.2517,
|
| 344 |
+
"step": 5600
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 4.66,
|
| 348 |
+
"learning_rate": 4.745833333333334e-06,
|
| 349 |
+
"loss": 3.2298,
|
| 350 |
+
"step": 5700
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 4.74,
|
| 354 |
+
"learning_rate": 4.829166666666667e-06,
|
| 355 |
+
"loss": 3.2291,
|
| 356 |
+
"step": 5800
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 4.82,
|
| 360 |
+
"learning_rate": 4.912500000000001e-06,
|
| 361 |
+
"loss": 3.2146,
|
| 362 |
+
"step": 5900
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 4.9,
|
| 366 |
+
"learning_rate": 4.995833333333334e-06,
|
| 367 |
+
"loss": 3.1858,
|
| 368 |
+
"step": 6000
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 4.98,
|
| 372 |
+
"learning_rate": 5e-06,
|
| 373 |
+
"loss": 3.1731,
|
| 374 |
+
"step": 6100
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"epoch": 5.07,
|
| 378 |
+
"learning_rate": 5e-06,
|
| 379 |
+
"loss": 3.1668,
|
| 380 |
+
"step": 6200
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"epoch": 5.15,
|
| 384 |
+
"learning_rate": 5e-06,
|
| 385 |
+
"loss": 3.1453,
|
| 386 |
+
"step": 6300
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 5.23,
|
| 390 |
+
"learning_rate": 5e-06,
|
| 391 |
+
"loss": 3.0692,
|
| 392 |
+
"step": 6400
|
| 393 |
+
},
|
| 394 |
+
{
|
| 395 |
+
"epoch": 5.31,
|
| 396 |
+
"learning_rate": 5e-06,
|
| 397 |
+
"loss": 3.0427,
|
| 398 |
+
"step": 6500
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 5.39,
|
| 402 |
+
"learning_rate": 5e-06,
|
| 403 |
+
"loss": 2.9999,
|
| 404 |
+
"step": 6600
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 5.47,
|
| 408 |
+
"learning_rate": 5e-06,
|
| 409 |
+
"loss": 2.9585,
|
| 410 |
+
"step": 6700
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 5.56,
|
| 414 |
+
"learning_rate": 5e-06,
|
| 415 |
+
"loss": 2.8819,
|
| 416 |
+
"step": 6800
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 5.64,
|
| 420 |
+
"learning_rate": 5e-06,
|
| 421 |
+
"loss": 2.8528,
|
| 422 |
+
"step": 6900
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 5.72,
|
| 426 |
+
"learning_rate": 5e-06,
|
| 427 |
+
"loss": 2.8175,
|
| 428 |
+
"step": 7000
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 5.8,
|
| 432 |
+
"learning_rate": 5e-06,
|
| 433 |
+
"loss": 2.735,
|
| 434 |
+
"step": 7100
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"epoch": 5.88,
|
| 438 |
+
"learning_rate": 5e-06,
|
| 439 |
+
"loss": 2.6801,
|
| 440 |
+
"step": 7200
|
| 441 |
+
},
|
| 442 |
+
{
|
| 443 |
+
"epoch": 5.96,
|
| 444 |
+
"learning_rate": 5e-06,
|
| 445 |
+
"loss": 2.6716,
|
| 446 |
+
"step": 7300
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 6.05,
|
| 450 |
+
"learning_rate": 5e-06,
|
| 451 |
+
"loss": 2.5997,
|
| 452 |
+
"step": 7400
|
| 453 |
+
},
|
| 454 |
+
{
|
| 455 |
+
"epoch": 6.13,
|
| 456 |
+
"learning_rate": 5e-06,
|
| 457 |
+
"loss": 2.5214,
|
| 458 |
+
"step": 7500
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"epoch": 6.21,
|
| 462 |
+
"learning_rate": 5e-06,
|
| 463 |
+
"loss": 2.5189,
|
| 464 |
+
"step": 7600
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 6.29,
|
| 468 |
+
"learning_rate": 5e-06,
|
| 469 |
+
"loss": 2.4448,
|
| 470 |
+
"step": 7700
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 6.37,
|
| 474 |
+
"learning_rate": 5e-06,
|
| 475 |
+
"loss": 2.3884,
|
| 476 |
+
"step": 7800
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 6.45,
|
| 480 |
+
"learning_rate": 5e-06,
|
| 481 |
+
"loss": 2.3592,
|
| 482 |
+
"step": 7900
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 6.54,
|
| 486 |
+
"learning_rate": 5e-06,
|
| 487 |
+
"loss": 2.2958,
|
| 488 |
+
"step": 8000
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 6.62,
|
| 492 |
+
"learning_rate": 5e-06,
|
| 493 |
+
"loss": 2.2593,
|
| 494 |
+
"step": 8100
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 6.7,
|
| 498 |
+
"learning_rate": 5e-06,
|
| 499 |
+
"loss": 2.2337,
|
| 500 |
+
"step": 8200
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 6.78,
|
| 504 |
+
"learning_rate": 5e-06,
|
| 505 |
+
"loss": 2.2003,
|
| 506 |
+
"step": 8300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 6.86,
|
| 510 |
+
"learning_rate": 5e-06,
|
| 511 |
+
"loss": 2.1803,
|
| 512 |
+
"step": 8400
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"epoch": 6.94,
|
| 516 |
+
"learning_rate": 5e-06,
|
| 517 |
+
"loss": 2.1439,
|
| 518 |
+
"step": 8500
|
| 519 |
+
},
|
| 520 |
+
{
|
| 521 |
+
"epoch": 7.03,
|
| 522 |
+
"learning_rate": 5e-06,
|
| 523 |
+
"loss": 2.1152,
|
| 524 |
+
"step": 8600
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 7.11,
|
| 528 |
+
"learning_rate": 5e-06,
|
| 529 |
+
"loss": 2.0329,
|
| 530 |
+
"step": 8700
|
| 531 |
+
},
|
| 532 |
+
{
|
| 533 |
+
"epoch": 7.19,
|
| 534 |
+
"learning_rate": 5e-06,
|
| 535 |
+
"loss": 2.0435,
|
| 536 |
+
"step": 8800
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"epoch": 7.27,
|
| 540 |
+
"learning_rate": 5e-06,
|
| 541 |
+
"loss": 2.0207,
|
| 542 |
+
"step": 8900
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"epoch": 7.35,
|
| 546 |
+
"learning_rate": 5e-06,
|
| 547 |
+
"loss": 1.9882,
|
| 548 |
+
"step": 9000
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 7.43,
|
| 552 |
+
"learning_rate": 5e-06,
|
| 553 |
+
"loss": 2.0032,
|
| 554 |
+
"step": 9100
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 7.52,
|
| 558 |
+
"learning_rate": 5e-06,
|
| 559 |
+
"loss": 1.9383,
|
| 560 |
+
"step": 9200
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 7.6,
|
| 564 |
+
"learning_rate": 5e-06,
|
| 565 |
+
"loss": 1.9368,
|
| 566 |
+
"step": 9300
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 7.68,
|
| 570 |
+
"learning_rate": 5e-06,
|
| 571 |
+
"loss": 1.8718,
|
| 572 |
+
"step": 9400
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 7.76,
|
| 576 |
+
"learning_rate": 5e-06,
|
| 577 |
+
"loss": 1.838,
|
| 578 |
+
"step": 9500
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 7.84,
|
| 582 |
+
"learning_rate": 5e-06,
|
| 583 |
+
"loss": 1.8233,
|
| 584 |
+
"step": 9600
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 7.92,
|
| 588 |
+
"learning_rate": 5e-06,
|
| 589 |
+
"loss": 1.8229,
|
| 590 |
+
"step": 9700
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 8.01,
|
| 594 |
+
"learning_rate": 5e-06,
|
| 595 |
+
"loss": 1.8271,
|
| 596 |
+
"step": 9800
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 8.09,
|
| 600 |
+
"learning_rate": 5e-06,
|
| 601 |
+
"loss": 1.8069,
|
| 602 |
+
"step": 9900
|
| 603 |
+
},
|
| 604 |
+
{
|
| 605 |
+
"epoch": 8.17,
|
| 606 |
+
"learning_rate": 5e-06,
|
| 607 |
+
"loss": 1.7105,
|
| 608 |
+
"step": 10000
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 8.17,
|
| 612 |
+
"eval_cer": 0.2669406109474337,
|
| 613 |
+
"eval_loss": 1.2349891662597656,
|
| 614 |
+
"eval_runtime": 8.338,
|
| 615 |
+
"eval_samples_per_second": 35.98,
|
| 616 |
+
"eval_steps_per_second": 17.99,
|
| 617 |
+
"eval_wer": 0.8188034188034188,
|
| 618 |
+
"step": 10000
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 8.25,
|
| 622 |
+
"learning_rate": 5e-06,
|
| 623 |
+
"loss": 1.7295,
|
| 624 |
+
"step": 10100
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 8.33,
|
| 628 |
+
"learning_rate": 5e-06,
|
| 629 |
+
"loss": 1.6777,
|
| 630 |
+
"step": 10200
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 8.41,
|
| 634 |
+
"learning_rate": 5e-06,
|
| 635 |
+
"loss": 1.6999,
|
| 636 |
+
"step": 10300
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"epoch": 8.5,
|
| 640 |
+
"learning_rate": 5e-06,
|
| 641 |
+
"loss": 1.7222,
|
| 642 |
+
"step": 10400
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 8.58,
|
| 646 |
+
"learning_rate": 5e-06,
|
| 647 |
+
"loss": 1.6903,
|
| 648 |
+
"step": 10500
|
| 649 |
+
},
|
| 650 |
+
{
|
| 651 |
+
"epoch": 8.66,
|
| 652 |
+
"learning_rate": 5e-06,
|
| 653 |
+
"loss": 1.6686,
|
| 654 |
+
"step": 10600
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"epoch": 8.74,
|
| 658 |
+
"learning_rate": 5e-06,
|
| 659 |
+
"loss": 1.675,
|
| 660 |
+
"step": 10700
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 8.82,
|
| 664 |
+
"learning_rate": 5e-06,
|
| 665 |
+
"loss": 1.63,
|
| 666 |
+
"step": 10800
|
| 667 |
+
},
|
| 668 |
+
{
|
| 669 |
+
"epoch": 8.9,
|
| 670 |
+
"learning_rate": 5e-06,
|
| 671 |
+
"loss": 1.6014,
|
| 672 |
+
"step": 10900
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 8.99,
|
| 676 |
+
"learning_rate": 5e-06,
|
| 677 |
+
"loss": 1.5622,
|
| 678 |
+
"step": 11000
|
| 679 |
+
},
|
| 680 |
+
{
|
| 681 |
+
"epoch": 9.07,
|
| 682 |
+
"learning_rate": 5e-06,
|
| 683 |
+
"loss": 1.6179,
|
| 684 |
+
"step": 11100
|
| 685 |
+
},
|
| 686 |
+
{
|
| 687 |
+
"epoch": 9.15,
|
| 688 |
+
"learning_rate": 5e-06,
|
| 689 |
+
"loss": 1.5449,
|
| 690 |
+
"step": 11200
|
| 691 |
+
},
|
| 692 |
+
{
|
| 693 |
+
"epoch": 9.23,
|
| 694 |
+
"learning_rate": 5e-06,
|
| 695 |
+
"loss": 1.5691,
|
| 696 |
+
"step": 11300
|
| 697 |
+
},
|
| 698 |
+
{
|
| 699 |
+
"epoch": 9.31,
|
| 700 |
+
"learning_rate": 5e-06,
|
| 701 |
+
"loss": 1.5252,
|
| 702 |
+
"step": 11400
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 9.4,
|
| 706 |
+
"learning_rate": 5e-06,
|
| 707 |
+
"loss": 1.5056,
|
| 708 |
+
"step": 11500
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 9.48,
|
| 712 |
+
"learning_rate": 5e-06,
|
| 713 |
+
"loss": 1.5231,
|
| 714 |
+
"step": 11600
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"epoch": 9.56,
|
| 718 |
+
"learning_rate": 5e-06,
|
| 719 |
+
"loss": 1.4915,
|
| 720 |
+
"step": 11700
|
| 721 |
+
},
|
| 722 |
+
{
|
| 723 |
+
"epoch": 9.64,
|
| 724 |
+
"learning_rate": 5e-06,
|
| 725 |
+
"loss": 1.5377,
|
| 726 |
+
"step": 11800
|
| 727 |
+
},
|
| 728 |
+
{
|
| 729 |
+
"epoch": 9.72,
|
| 730 |
+
"learning_rate": 5e-06,
|
| 731 |
+
"loss": 1.4515,
|
| 732 |
+
"step": 11900
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"epoch": 9.8,
|
| 736 |
+
"learning_rate": 5e-06,
|
| 737 |
+
"loss": 1.4831,
|
| 738 |
+
"step": 12000
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"epoch": 9.89,
|
| 742 |
+
"learning_rate": 5e-06,
|
| 743 |
+
"loss": 1.4679,
|
| 744 |
+
"step": 12100
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"epoch": 9.97,
|
| 748 |
+
"learning_rate": 5e-06,
|
| 749 |
+
"loss": 1.4193,
|
| 750 |
+
"step": 12200
|
| 751 |
+
},
|
| 752 |
+
{
|
| 753 |
+
"epoch": 10.05,
|
| 754 |
+
"learning_rate": 5e-06,
|
| 755 |
+
"loss": 1.4637,
|
| 756 |
+
"step": 12300
|
| 757 |
+
},
|
| 758 |
+
{
|
| 759 |
+
"epoch": 10.13,
|
| 760 |
+
"learning_rate": 5e-06,
|
| 761 |
+
"loss": 1.4281,
|
| 762 |
+
"step": 12400
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"epoch": 10.21,
|
| 766 |
+
"learning_rate": 5e-06,
|
| 767 |
+
"loss": 1.4774,
|
| 768 |
+
"step": 12500
|
| 769 |
+
},
|
| 770 |
+
{
|
| 771 |
+
"epoch": 10.29,
|
| 772 |
+
"learning_rate": 5e-06,
|
| 773 |
+
"loss": 1.3995,
|
| 774 |
+
"step": 12600
|
| 775 |
+
},
|
| 776 |
+
{
|
| 777 |
+
"epoch": 10.38,
|
| 778 |
+
"learning_rate": 5e-06,
|
| 779 |
+
"loss": 1.3925,
|
| 780 |
+
"step": 12700
|
| 781 |
+
},
|
| 782 |
+
{
|
| 783 |
+
"epoch": 10.46,
|
| 784 |
+
"learning_rate": 5e-06,
|
| 785 |
+
"loss": 1.3734,
|
| 786 |
+
"step": 12800
|
| 787 |
+
},
|
| 788 |
+
{
|
| 789 |
+
"epoch": 10.54,
|
| 790 |
+
"learning_rate": 5e-06,
|
| 791 |
+
"loss": 1.4023,
|
| 792 |
+
"step": 12900
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"epoch": 10.62,
|
| 796 |
+
"learning_rate": 5e-06,
|
| 797 |
+
"loss": 1.3599,
|
| 798 |
+
"step": 13000
|
| 799 |
+
},
|
| 800 |
+
{
|
| 801 |
+
"epoch": 10.7,
|
| 802 |
+
"learning_rate": 5e-06,
|
| 803 |
+
"loss": 1.3517,
|
| 804 |
+
"step": 13100
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"epoch": 10.78,
|
| 808 |
+
"learning_rate": 5e-06,
|
| 809 |
+
"loss": 1.3845,
|
| 810 |
+
"step": 13200
|
| 811 |
+
},
|
| 812 |
+
{
|
| 813 |
+
"epoch": 10.87,
|
| 814 |
+
"learning_rate": 5e-06,
|
| 815 |
+
"loss": 1.3561,
|
| 816 |
+
"step": 13300
|
| 817 |
+
},
|
| 818 |
+
{
|
| 819 |
+
"epoch": 10.95,
|
| 820 |
+
"learning_rate": 5e-06,
|
| 821 |
+
"loss": 1.3288,
|
| 822 |
+
"step": 13400
|
| 823 |
+
},
|
| 824 |
+
{
|
| 825 |
+
"epoch": 11.03,
|
| 826 |
+
"learning_rate": 5e-06,
|
| 827 |
+
"loss": 1.329,
|
| 828 |
+
"step": 13500
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"epoch": 11.11,
|
| 832 |
+
"learning_rate": 5e-06,
|
| 833 |
+
"loss": 1.2921,
|
| 834 |
+
"step": 13600
|
| 835 |
+
},
|
| 836 |
+
{
|
| 837 |
+
"epoch": 11.19,
|
| 838 |
+
"learning_rate": 5e-06,
|
| 839 |
+
"loss": 1.2805,
|
| 840 |
+
"step": 13700
|
| 841 |
+
},
|
| 842 |
+
{
|
| 843 |
+
"epoch": 11.27,
|
| 844 |
+
"learning_rate": 5e-06,
|
| 845 |
+
"loss": 1.3231,
|
| 846 |
+
"step": 13800
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"epoch": 11.36,
|
| 850 |
+
"learning_rate": 5e-06,
|
| 851 |
+
"loss": 1.3033,
|
| 852 |
+
"step": 13900
|
| 853 |
+
},
|
| 854 |
+
{
|
| 855 |
+
"epoch": 11.44,
|
| 856 |
+
"learning_rate": 5e-06,
|
| 857 |
+
"loss": 1.3202,
|
| 858 |
+
"step": 14000
|
| 859 |
+
},
|
| 860 |
+
{
|
| 861 |
+
"epoch": 11.52,
|
| 862 |
+
"learning_rate": 5e-06,
|
| 863 |
+
"loss": 1.289,
|
| 864 |
+
"step": 14100
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 11.6,
|
| 868 |
+
"learning_rate": 5e-06,
|
| 869 |
+
"loss": 1.2955,
|
| 870 |
+
"step": 14200
|
| 871 |
+
},
|
| 872 |
+
{
|
| 873 |
+
"epoch": 11.68,
|
| 874 |
+
"learning_rate": 5e-06,
|
| 875 |
+
"loss": 1.2818,
|
| 876 |
+
"step": 14300
|
| 877 |
+
},
|
| 878 |
+
{
|
| 879 |
+
"epoch": 11.76,
|
| 880 |
+
"learning_rate": 5e-06,
|
| 881 |
+
"loss": 1.2486,
|
| 882 |
+
"step": 14400
|
| 883 |
+
},
|
| 884 |
+
{
|
| 885 |
+
"epoch": 11.85,
|
| 886 |
+
"learning_rate": 5e-06,
|
| 887 |
+
"loss": 1.2817,
|
| 888 |
+
"step": 14500
|
| 889 |
+
},
|
| 890 |
+
{
|
| 891 |
+
"epoch": 11.93,
|
| 892 |
+
"learning_rate": 5e-06,
|
| 893 |
+
"loss": 1.2654,
|
| 894 |
+
"step": 14600
|
| 895 |
+
},
|
| 896 |
+
{
|
| 897 |
+
"epoch": 12.01,
|
| 898 |
+
"learning_rate": 5e-06,
|
| 899 |
+
"loss": 1.312,
|
| 900 |
+
"step": 14700
|
| 901 |
+
},
|
| 902 |
+
{
|
| 903 |
+
"epoch": 12.09,
|
| 904 |
+
"learning_rate": 5e-06,
|
| 905 |
+
"loss": 1.2222,
|
| 906 |
+
"step": 14800
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"epoch": 12.17,
|
| 910 |
+
"learning_rate": 5e-06,
|
| 911 |
+
"loss": 1.2101,
|
| 912 |
+
"step": 14900
|
| 913 |
+
},
|
| 914 |
+
{
|
| 915 |
+
"epoch": 12.25,
|
| 916 |
+
"learning_rate": 5e-06,
|
| 917 |
+
"loss": 1.2481,
|
| 918 |
+
"step": 15000
|
| 919 |
+
},
|
| 920 |
+
{
|
| 921 |
+
"epoch": 12.34,
|
| 922 |
+
"learning_rate": 5e-06,
|
| 923 |
+
"loss": 1.2697,
|
| 924 |
+
"step": 15100
|
| 925 |
+
},
|
| 926 |
+
{
|
| 927 |
+
"epoch": 12.42,
|
| 928 |
+
"learning_rate": 5e-06,
|
| 929 |
+
"loss": 1.2074,
|
| 930 |
+
"step": 15200
|
| 931 |
+
},
|
| 932 |
+
{
|
| 933 |
+
"epoch": 12.5,
|
| 934 |
+
"learning_rate": 5e-06,
|
| 935 |
+
"loss": 1.2031,
|
| 936 |
+
"step": 15300
|
| 937 |
+
},
|
| 938 |
+
{
|
| 939 |
+
"epoch": 12.58,
|
| 940 |
+
"learning_rate": 5e-06,
|
| 941 |
+
"loss": 1.2189,
|
| 942 |
+
"step": 15400
|
| 943 |
+
},
|
| 944 |
+
{
|
| 945 |
+
"epoch": 12.66,
|
| 946 |
+
"learning_rate": 5e-06,
|
| 947 |
+
"loss": 1.2007,
|
| 948 |
+
"step": 15500
|
| 949 |
+
},
|
| 950 |
+
{
|
| 951 |
+
"epoch": 12.74,
|
| 952 |
+
"learning_rate": 5e-06,
|
| 953 |
+
"loss": 1.2208,
|
| 954 |
+
"step": 15600
|
| 955 |
+
},
|
| 956 |
+
{
|
| 957 |
+
"epoch": 12.83,
|
| 958 |
+
"learning_rate": 5e-06,
|
| 959 |
+
"loss": 1.2302,
|
| 960 |
+
"step": 15700
|
| 961 |
+
},
|
| 962 |
+
{
|
| 963 |
+
"epoch": 12.91,
|
| 964 |
+
"learning_rate": 5e-06,
|
| 965 |
+
"loss": 1.1985,
|
| 966 |
+
"step": 15800
|
| 967 |
+
},
|
| 968 |
+
{
|
| 969 |
+
"epoch": 12.99,
|
| 970 |
+
"learning_rate": 5e-06,
|
| 971 |
+
"loss": 1.197,
|
| 972 |
+
"step": 15900
|
| 973 |
+
},
|
| 974 |
+
{
|
| 975 |
+
"epoch": 13.07,
|
| 976 |
+
"learning_rate": 5e-06,
|
| 977 |
+
"loss": 1.2294,
|
| 978 |
+
"step": 16000
|
| 979 |
+
},
|
| 980 |
+
{
|
| 981 |
+
"epoch": 13.15,
|
| 982 |
+
"learning_rate": 5e-06,
|
| 983 |
+
"loss": 1.1481,
|
| 984 |
+
"step": 16100
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"epoch": 13.24,
|
| 988 |
+
"learning_rate": 5e-06,
|
| 989 |
+
"loss": 1.1518,
|
| 990 |
+
"step": 16200
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"epoch": 13.32,
|
| 994 |
+
"learning_rate": 5e-06,
|
| 995 |
+
"loss": 1.2009,
|
| 996 |
+
"step": 16300
|
| 997 |
+
},
|
| 998 |
+
{
|
| 999 |
+
"epoch": 13.4,
|
| 1000 |
+
"learning_rate": 5e-06,
|
| 1001 |
+
"loss": 1.1421,
|
| 1002 |
+
"step": 16400
|
| 1003 |
+
},
|
| 1004 |
+
{
|
| 1005 |
+
"epoch": 13.48,
|
| 1006 |
+
"learning_rate": 5e-06,
|
| 1007 |
+
"loss": 1.1966,
|
| 1008 |
+
"step": 16500
|
| 1009 |
+
},
|
| 1010 |
+
{
|
| 1011 |
+
"epoch": 13.56,
|
| 1012 |
+
"learning_rate": 5e-06,
|
| 1013 |
+
"loss": 1.1785,
|
| 1014 |
+
"step": 16600
|
| 1015 |
+
},
|
| 1016 |
+
{
|
| 1017 |
+
"epoch": 13.64,
|
| 1018 |
+
"learning_rate": 5e-06,
|
| 1019 |
+
"loss": 1.1549,
|
| 1020 |
+
"step": 16700
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 13.73,
|
| 1024 |
+
"learning_rate": 5e-06,
|
| 1025 |
+
"loss": 1.1844,
|
| 1026 |
+
"step": 16800
|
| 1027 |
+
},
|
| 1028 |
+
{
|
| 1029 |
+
"epoch": 13.81,
|
| 1030 |
+
"learning_rate": 5e-06,
|
| 1031 |
+
"loss": 1.1471,
|
| 1032 |
+
"step": 16900
|
| 1033 |
+
},
|
| 1034 |
+
{
|
| 1035 |
+
"epoch": 13.89,
|
| 1036 |
+
"learning_rate": 5e-06,
|
| 1037 |
+
"loss": 1.1748,
|
| 1038 |
+
"step": 17000
|
| 1039 |
+
},
|
| 1040 |
+
{
|
| 1041 |
+
"epoch": 13.97,
|
| 1042 |
+
"learning_rate": 5e-06,
|
| 1043 |
+
"loss": 1.1463,
|
| 1044 |
+
"step": 17100
|
| 1045 |
+
},
|
| 1046 |
+
{
|
| 1047 |
+
"epoch": 14.05,
|
| 1048 |
+
"learning_rate": 5e-06,
|
| 1049 |
+
"loss": 1.154,
|
| 1050 |
+
"step": 17200
|
| 1051 |
+
},
|
| 1052 |
+
{
|
| 1053 |
+
"epoch": 14.13,
|
| 1054 |
+
"learning_rate": 5e-06,
|
| 1055 |
+
"loss": 1.1551,
|
| 1056 |
+
"step": 17300
|
| 1057 |
+
},
|
| 1058 |
+
{
|
| 1059 |
+
"epoch": 14.22,
|
| 1060 |
+
"learning_rate": 5e-06,
|
| 1061 |
+
"loss": 1.1271,
|
| 1062 |
+
"step": 17400
|
| 1063 |
+
},
|
| 1064 |
+
{
|
| 1065 |
+
"epoch": 14.3,
|
| 1066 |
+
"learning_rate": 5e-06,
|
| 1067 |
+
"loss": 1.1019,
|
| 1068 |
+
"step": 17500
|
| 1069 |
+
},
|
| 1070 |
+
{
|
| 1071 |
+
"epoch": 14.38,
|
| 1072 |
+
"learning_rate": 5e-06,
|
| 1073 |
+
"loss": 1.1212,
|
| 1074 |
+
"step": 17600
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"epoch": 14.46,
|
| 1078 |
+
"learning_rate": 5e-06,
|
| 1079 |
+
"loss": 1.0808,
|
| 1080 |
+
"step": 17700
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"epoch": 14.54,
|
| 1084 |
+
"learning_rate": 5e-06,
|
| 1085 |
+
"loss": 1.171,
|
| 1086 |
+
"step": 17800
|
| 1087 |
+
},
|
| 1088 |
+
{
|
| 1089 |
+
"epoch": 14.62,
|
| 1090 |
+
"learning_rate": 5e-06,
|
| 1091 |
+
"loss": 1.157,
|
| 1092 |
+
"step": 17900
|
| 1093 |
+
},
|
| 1094 |
+
{
|
| 1095 |
+
"epoch": 14.71,
|
| 1096 |
+
"learning_rate": 5e-06,
|
| 1097 |
+
"loss": 1.0964,
|
| 1098 |
+
"step": 18000
|
| 1099 |
+
},
|
| 1100 |
+
{
|
| 1101 |
+
"epoch": 14.79,
|
| 1102 |
+
"learning_rate": 5e-06,
|
| 1103 |
+
"loss": 1.1006,
|
| 1104 |
+
"step": 18100
|
| 1105 |
+
},
|
| 1106 |
+
{
|
| 1107 |
+
"epoch": 14.87,
|
| 1108 |
+
"learning_rate": 5e-06,
|
| 1109 |
+
"loss": 1.1505,
|
| 1110 |
+
"step": 18200
|
| 1111 |
+
},
|
| 1112 |
+
{
|
| 1113 |
+
"epoch": 14.95,
|
| 1114 |
+
"learning_rate": 5e-06,
|
| 1115 |
+
"loss": 1.1402,
|
| 1116 |
+
"step": 18300
|
| 1117 |
+
},
|
| 1118 |
+
{
|
| 1119 |
+
"epoch": 15.03,
|
| 1120 |
+
"learning_rate": 5e-06,
|
| 1121 |
+
"loss": 1.0922,
|
| 1122 |
+
"step": 18400
|
| 1123 |
+
},
|
| 1124 |
+
{
|
| 1125 |
+
"epoch": 15.11,
|
| 1126 |
+
"learning_rate": 5e-06,
|
| 1127 |
+
"loss": 1.1458,
|
| 1128 |
+
"step": 18500
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"epoch": 15.2,
|
| 1132 |
+
"learning_rate": 5e-06,
|
| 1133 |
+
"loss": 1.0835,
|
| 1134 |
+
"step": 18600
|
| 1135 |
+
},
|
| 1136 |
+
{
|
| 1137 |
+
"epoch": 15.28,
|
| 1138 |
+
"learning_rate": 5e-06,
|
| 1139 |
+
"loss": 1.0964,
|
| 1140 |
+
"step": 18700
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"epoch": 15.36,
|
| 1144 |
+
"learning_rate": 5e-06,
|
| 1145 |
+
"loss": 1.0833,
|
| 1146 |
+
"step": 18800
|
| 1147 |
+
},
|
| 1148 |
+
{
|
| 1149 |
+
"epoch": 15.44,
|
| 1150 |
+
"learning_rate": 5e-06,
|
| 1151 |
+
"loss": 1.0897,
|
| 1152 |
+
"step": 18900
|
| 1153 |
+
},
|
| 1154 |
+
{
|
| 1155 |
+
"epoch": 15.52,
|
| 1156 |
+
"learning_rate": 5e-06,
|
| 1157 |
+
"loss": 1.0544,
|
| 1158 |
+
"step": 19000
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"epoch": 15.6,
|
| 1162 |
+
"learning_rate": 5e-06,
|
| 1163 |
+
"loss": 1.0939,
|
| 1164 |
+
"step": 19100
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"epoch": 15.69,
|
| 1168 |
+
"learning_rate": 5e-06,
|
| 1169 |
+
"loss": 1.0333,
|
| 1170 |
+
"step": 19200
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 15.77,
|
| 1174 |
+
"learning_rate": 5e-06,
|
| 1175 |
+
"loss": 1.1071,
|
| 1176 |
+
"step": 19300
|
| 1177 |
+
},
|
| 1178 |
+
{
|
| 1179 |
+
"epoch": 15.85,
|
| 1180 |
+
"learning_rate": 5e-06,
|
| 1181 |
+
"loss": 1.0865,
|
| 1182 |
+
"step": 19400
|
| 1183 |
+
},
|
| 1184 |
+
{
|
| 1185 |
+
"epoch": 15.93,
|
| 1186 |
+
"learning_rate": 5e-06,
|
| 1187 |
+
"loss": 1.0961,
|
| 1188 |
+
"step": 19500
|
| 1189 |
+
},
|
| 1190 |
+
{
|
| 1191 |
+
"epoch": 16.01,
|
| 1192 |
+
"learning_rate": 5e-06,
|
| 1193 |
+
"loss": 1.0696,
|
| 1194 |
+
"step": 19600
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"epoch": 16.09,
|
| 1198 |
+
"learning_rate": 5e-06,
|
| 1199 |
+
"loss": 1.0574,
|
| 1200 |
+
"step": 19700
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"epoch": 16.18,
|
| 1204 |
+
"learning_rate": 5e-06,
|
| 1205 |
+
"loss": 1.0373,
|
| 1206 |
+
"step": 19800
|
| 1207 |
+
},
|
| 1208 |
+
{
|
| 1209 |
+
"epoch": 16.26,
|
| 1210 |
+
"learning_rate": 5e-06,
|
| 1211 |
+
"loss": 1.0526,
|
| 1212 |
+
"step": 19900
|
| 1213 |
+
},
|
| 1214 |
+
{
|
| 1215 |
+
"epoch": 16.34,
|
| 1216 |
+
"learning_rate": 5e-06,
|
| 1217 |
+
"loss": 1.0707,
|
| 1218 |
+
"step": 20000
|
| 1219 |
+
},
|
| 1220 |
+
{
|
| 1221 |
+
"epoch": 16.34,
|
| 1222 |
+
"eval_cer": 0.14808497441463792,
|
| 1223 |
+
"eval_loss": 0.6507723331451416,
|
| 1224 |
+
"eval_runtime": 8.0755,
|
| 1225 |
+
"eval_samples_per_second": 37.149,
|
| 1226 |
+
"eval_steps_per_second": 18.575,
|
| 1227 |
+
"eval_wer": 0.5153846153846153,
|
| 1228 |
+
"step": 20000
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"epoch": 16.42,
|
| 1232 |
+
"learning_rate": 5e-06,
|
| 1233 |
+
"loss": 1.0709,
|
| 1234 |
+
"step": 20100
|
| 1235 |
+
},
|
| 1236 |
+
{
|
| 1237 |
+
"epoch": 16.5,
|
| 1238 |
+
"learning_rate": 5e-06,
|
| 1239 |
+
"loss": 1.0897,
|
| 1240 |
+
"step": 20200
|
| 1241 |
+
},
|
| 1242 |
+
{
|
| 1243 |
+
"epoch": 16.58,
|
| 1244 |
+
"learning_rate": 5e-06,
|
| 1245 |
+
"loss": 1.0179,
|
| 1246 |
+
"step": 20300
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"epoch": 16.67,
|
| 1250 |
+
"learning_rate": 5e-06,
|
| 1251 |
+
"loss": 1.0557,
|
| 1252 |
+
"step": 20400
|
| 1253 |
+
},
|
| 1254 |
+
{
|
| 1255 |
+
"epoch": 16.75,
|
| 1256 |
+
"learning_rate": 5e-06,
|
| 1257 |
+
"loss": 1.0734,
|
| 1258 |
+
"step": 20500
|
| 1259 |
+
},
|
| 1260 |
+
{
|
| 1261 |
+
"epoch": 16.83,
|
| 1262 |
+
"learning_rate": 5e-06,
|
| 1263 |
+
"loss": 1.0456,
|
| 1264 |
+
"step": 20600
|
| 1265 |
+
},
|
| 1266 |
+
{
|
| 1267 |
+
"epoch": 16.91,
|
| 1268 |
+
"learning_rate": 5e-06,
|
| 1269 |
+
"loss": 1.0186,
|
| 1270 |
+
"step": 20700
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"epoch": 16.99,
|
| 1274 |
+
"learning_rate": 5e-06,
|
| 1275 |
+
"loss": 1.0345,
|
| 1276 |
+
"step": 20800
|
| 1277 |
+
},
|
| 1278 |
+
{
|
| 1279 |
+
"epoch": 17.08,
|
| 1280 |
+
"learning_rate": 5e-06,
|
| 1281 |
+
"loss": 1.093,
|
| 1282 |
+
"step": 20900
|
| 1283 |
+
},
|
| 1284 |
+
{
|
| 1285 |
+
"epoch": 17.16,
|
| 1286 |
+
"learning_rate": 5e-06,
|
| 1287 |
+
"loss": 1.0239,
|
| 1288 |
+
"step": 21000
|
| 1289 |
+
},
|
| 1290 |
+
{
|
| 1291 |
+
"epoch": 17.24,
|
| 1292 |
+
"learning_rate": 5e-06,
|
| 1293 |
+
"loss": 1.0453,
|
| 1294 |
+
"step": 21100
|
| 1295 |
+
},
|
| 1296 |
+
{
|
| 1297 |
+
"epoch": 17.32,
|
| 1298 |
+
"learning_rate": 5e-06,
|
| 1299 |
+
"loss": 1.01,
|
| 1300 |
+
"step": 21200
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"epoch": 17.4,
|
| 1304 |
+
"learning_rate": 5e-06,
|
| 1305 |
+
"loss": 1.0602,
|
| 1306 |
+
"step": 21300
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 17.48,
|
| 1310 |
+
"learning_rate": 5e-06,
|
| 1311 |
+
"loss": 1.0047,
|
| 1312 |
+
"step": 21400
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"epoch": 17.57,
|
| 1316 |
+
"learning_rate": 5e-06,
|
| 1317 |
+
"loss": 1.0379,
|
| 1318 |
+
"step": 21500
|
| 1319 |
+
},
|
| 1320 |
+
{
|
| 1321 |
+
"epoch": 17.65,
|
| 1322 |
+
"learning_rate": 5e-06,
|
| 1323 |
+
"loss": 1.0277,
|
| 1324 |
+
"step": 21600
|
| 1325 |
+
},
|
| 1326 |
+
{
|
| 1327 |
+
"epoch": 17.73,
|
| 1328 |
+
"learning_rate": 5e-06,
|
| 1329 |
+
"loss": 1.0434,
|
| 1330 |
+
"step": 21700
|
| 1331 |
+
},
|
| 1332 |
+
{
|
| 1333 |
+
"epoch": 17.81,
|
| 1334 |
+
"learning_rate": 5e-06,
|
| 1335 |
+
"loss": 0.9741,
|
| 1336 |
+
"step": 21800
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 17.89,
|
| 1340 |
+
"learning_rate": 5e-06,
|
| 1341 |
+
"loss": 0.9905,
|
| 1342 |
+
"step": 21900
|
| 1343 |
+
},
|
| 1344 |
+
{
|
| 1345 |
+
"epoch": 17.97,
|
| 1346 |
+
"learning_rate": 5e-06,
|
| 1347 |
+
"loss": 1.0115,
|
| 1348 |
+
"step": 22000
|
| 1349 |
+
},
|
| 1350 |
+
{
|
| 1351 |
+
"epoch": 18.06,
|
| 1352 |
+
"learning_rate": 5e-06,
|
| 1353 |
+
"loss": 0.9891,
|
| 1354 |
+
"step": 22100
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 18.14,
|
| 1358 |
+
"learning_rate": 5e-06,
|
| 1359 |
+
"loss": 1.0262,
|
| 1360 |
+
"step": 22200
|
| 1361 |
+
},
|
| 1362 |
+
{
|
| 1363 |
+
"epoch": 18.22,
|
| 1364 |
+
"learning_rate": 5e-06,
|
| 1365 |
+
"loss": 1.0048,
|
| 1366 |
+
"step": 22300
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"epoch": 18.3,
|
| 1370 |
+
"learning_rate": 5e-06,
|
| 1371 |
+
"loss": 1.0176,
|
| 1372 |
+
"step": 22400
|
| 1373 |
+
},
|
| 1374 |
+
{
|
| 1375 |
+
"epoch": 18.38,
|
| 1376 |
+
"learning_rate": 5e-06,
|
| 1377 |
+
"loss": 1.01,
|
| 1378 |
+
"step": 22500
|
| 1379 |
+
},
|
| 1380 |
+
{
|
| 1381 |
+
"epoch": 18.46,
|
| 1382 |
+
"learning_rate": 5e-06,
|
| 1383 |
+
"loss": 0.9899,
|
| 1384 |
+
"step": 22600
|
| 1385 |
+
},
|
| 1386 |
+
{
|
| 1387 |
+
"epoch": 18.55,
|
| 1388 |
+
"learning_rate": 5e-06,
|
| 1389 |
+
"loss": 0.9553,
|
| 1390 |
+
"step": 22700
|
| 1391 |
+
},
|
| 1392 |
+
{
|
| 1393 |
+
"epoch": 18.63,
|
| 1394 |
+
"learning_rate": 5e-06,
|
| 1395 |
+
"loss": 0.9849,
|
| 1396 |
+
"step": 22800
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 18.71,
|
| 1400 |
+
"learning_rate": 5e-06,
|
| 1401 |
+
"loss": 0.9745,
|
| 1402 |
+
"step": 22900
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"epoch": 18.79,
|
| 1406 |
+
"learning_rate": 5e-06,
|
| 1407 |
+
"loss": 0.9761,
|
| 1408 |
+
"step": 23000
|
| 1409 |
+
},
|
| 1410 |
+
{
|
| 1411 |
+
"epoch": 18.87,
|
| 1412 |
+
"learning_rate": 5e-06,
|
| 1413 |
+
"loss": 1.0133,
|
| 1414 |
+
"step": 23100
|
| 1415 |
+
},
|
| 1416 |
+
{
|
| 1417 |
+
"epoch": 18.95,
|
| 1418 |
+
"learning_rate": 5e-06,
|
| 1419 |
+
"loss": 0.9874,
|
| 1420 |
+
"step": 23200
|
| 1421 |
+
},
|
| 1422 |
+
{
|
| 1423 |
+
"epoch": 19.04,
|
| 1424 |
+
"learning_rate": 5e-06,
|
| 1425 |
+
"loss": 0.9607,
|
| 1426 |
+
"step": 23300
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"epoch": 19.12,
|
| 1430 |
+
"learning_rate": 5e-06,
|
| 1431 |
+
"loss": 0.9738,
|
| 1432 |
+
"step": 23400
|
| 1433 |
+
},
|
| 1434 |
+
{
|
| 1435 |
+
"epoch": 19.2,
|
| 1436 |
+
"learning_rate": 5e-06,
|
| 1437 |
+
"loss": 0.9543,
|
| 1438 |
+
"step": 23500
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"epoch": 19.28,
|
| 1442 |
+
"learning_rate": 5e-06,
|
| 1443 |
+
"loss": 0.9533,
|
| 1444 |
+
"step": 23600
|
| 1445 |
+
},
|
| 1446 |
+
{
|
| 1447 |
+
"epoch": 19.36,
|
| 1448 |
+
"learning_rate": 5e-06,
|
| 1449 |
+
"loss": 1.0028,
|
| 1450 |
+
"step": 23700
|
| 1451 |
+
},
|
| 1452 |
+
{
|
| 1453 |
+
"epoch": 19.44,
|
| 1454 |
+
"learning_rate": 5e-06,
|
| 1455 |
+
"loss": 0.937,
|
| 1456 |
+
"step": 23800
|
| 1457 |
+
},
|
| 1458 |
+
{
|
| 1459 |
+
"epoch": 19.53,
|
| 1460 |
+
"learning_rate": 5e-06,
|
| 1461 |
+
"loss": 0.9979,
|
| 1462 |
+
"step": 23900
|
| 1463 |
+
},
|
| 1464 |
+
{
|
| 1465 |
+
"epoch": 19.61,
|
| 1466 |
+
"learning_rate": 5e-06,
|
| 1467 |
+
"loss": 0.9872,
|
| 1468 |
+
"step": 24000
|
| 1469 |
+
},
|
| 1470 |
+
{
|
| 1471 |
+
"epoch": 19.69,
|
| 1472 |
+
"learning_rate": 5e-06,
|
| 1473 |
+
"loss": 0.9519,
|
| 1474 |
+
"step": 24100
|
| 1475 |
+
},
|
| 1476 |
+
{
|
| 1477 |
+
"epoch": 19.77,
|
| 1478 |
+
"learning_rate": 5e-06,
|
| 1479 |
+
"loss": 0.9624,
|
| 1480 |
+
"step": 24200
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"epoch": 19.85,
|
| 1484 |
+
"learning_rate": 5e-06,
|
| 1485 |
+
"loss": 1.0209,
|
| 1486 |
+
"step": 24300
|
| 1487 |
+
},
|
| 1488 |
+
{
|
| 1489 |
+
"epoch": 19.93,
|
| 1490 |
+
"learning_rate": 5e-06,
|
| 1491 |
+
"loss": 0.9874,
|
| 1492 |
+
"step": 24400
|
| 1493 |
+
},
|
| 1494 |
+
{
|
| 1495 |
+
"epoch": 20.02,
|
| 1496 |
+
"learning_rate": 5e-06,
|
| 1497 |
+
"loss": 0.9655,
|
| 1498 |
+
"step": 24500
|
| 1499 |
+
},
|
| 1500 |
+
{
|
| 1501 |
+
"epoch": 20.1,
|
| 1502 |
+
"learning_rate": 5e-06,
|
| 1503 |
+
"loss": 0.9492,
|
| 1504 |
+
"step": 24600
|
| 1505 |
+
},
|
| 1506 |
+
{
|
| 1507 |
+
"epoch": 20.18,
|
| 1508 |
+
"learning_rate": 5e-06,
|
| 1509 |
+
"loss": 0.9851,
|
| 1510 |
+
"step": 24700
|
| 1511 |
+
},
|
| 1512 |
+
{
|
| 1513 |
+
"epoch": 20.26,
|
| 1514 |
+
"learning_rate": 5e-06,
|
| 1515 |
+
"loss": 0.9355,
|
| 1516 |
+
"step": 24800
|
| 1517 |
+
},
|
| 1518 |
+
{
|
| 1519 |
+
"epoch": 20.34,
|
| 1520 |
+
"learning_rate": 5e-06,
|
| 1521 |
+
"loss": 0.972,
|
| 1522 |
+
"step": 24900
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"epoch": 20.42,
|
| 1526 |
+
"learning_rate": 5e-06,
|
| 1527 |
+
"loss": 0.9426,
|
| 1528 |
+
"step": 25000
|
| 1529 |
+
},
|
| 1530 |
+
{
|
| 1531 |
+
"epoch": 20.51,
|
| 1532 |
+
"learning_rate": 5e-06,
|
| 1533 |
+
"loss": 0.9908,
|
| 1534 |
+
"step": 25100
|
| 1535 |
+
},
|
| 1536 |
+
{
|
| 1537 |
+
"epoch": 20.59,
|
| 1538 |
+
"learning_rate": 5e-06,
|
| 1539 |
+
"loss": 0.9182,
|
| 1540 |
+
"step": 25200
|
| 1541 |
+
},
|
| 1542 |
+
{
|
| 1543 |
+
"epoch": 20.67,
|
| 1544 |
+
"learning_rate": 5e-06,
|
| 1545 |
+
"loss": 0.9191,
|
| 1546 |
+
"step": 25300
|
| 1547 |
+
},
|
| 1548 |
+
{
|
| 1549 |
+
"epoch": 20.75,
|
| 1550 |
+
"learning_rate": 5e-06,
|
| 1551 |
+
"loss": 0.9784,
|
| 1552 |
+
"step": 25400
|
| 1553 |
+
},
|
| 1554 |
+
{
|
| 1555 |
+
"epoch": 20.83,
|
| 1556 |
+
"learning_rate": 5e-06,
|
| 1557 |
+
"loss": 0.944,
|
| 1558 |
+
"step": 25500
|
| 1559 |
+
},
|
| 1560 |
+
{
|
| 1561 |
+
"epoch": 20.91,
|
| 1562 |
+
"learning_rate": 5e-06,
|
| 1563 |
+
"loss": 0.9475,
|
| 1564 |
+
"step": 25600
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"epoch": 21.0,
|
| 1568 |
+
"learning_rate": 5e-06,
|
| 1569 |
+
"loss": 0.9254,
|
| 1570 |
+
"step": 25700
|
| 1571 |
+
},
|
| 1572 |
+
{
|
| 1573 |
+
"epoch": 21.08,
|
| 1574 |
+
"learning_rate": 5e-06,
|
| 1575 |
+
"loss": 0.9553,
|
| 1576 |
+
"step": 25800
|
| 1577 |
+
},
|
| 1578 |
+
{
|
| 1579 |
+
"epoch": 21.16,
|
| 1580 |
+
"learning_rate": 5e-06,
|
| 1581 |
+
"loss": 0.9452,
|
| 1582 |
+
"step": 25900
|
| 1583 |
+
},
|
| 1584 |
+
{
|
| 1585 |
+
"epoch": 21.24,
|
| 1586 |
+
"learning_rate": 5e-06,
|
| 1587 |
+
"loss": 0.9666,
|
| 1588 |
+
"step": 26000
|
| 1589 |
+
},
|
| 1590 |
+
{
|
| 1591 |
+
"epoch": 21.32,
|
| 1592 |
+
"learning_rate": 5e-06,
|
| 1593 |
+
"loss": 0.9389,
|
| 1594 |
+
"step": 26100
|
| 1595 |
+
},
|
| 1596 |
+
{
|
| 1597 |
+
"epoch": 21.4,
|
| 1598 |
+
"learning_rate": 5e-06,
|
| 1599 |
+
"loss": 0.971,
|
| 1600 |
+
"step": 26200
|
| 1601 |
+
},
|
| 1602 |
+
{
|
| 1603 |
+
"epoch": 21.49,
|
| 1604 |
+
"learning_rate": 5e-06,
|
| 1605 |
+
"loss": 0.9212,
|
| 1606 |
+
"step": 26300
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"epoch": 21.57,
|
| 1610 |
+
"learning_rate": 5e-06,
|
| 1611 |
+
"loss": 0.9039,
|
| 1612 |
+
"step": 26400
|
| 1613 |
+
},
|
| 1614 |
+
{
|
| 1615 |
+
"epoch": 21.65,
|
| 1616 |
+
"learning_rate": 5e-06,
|
| 1617 |
+
"loss": 0.8837,
|
| 1618 |
+
"step": 26500
|
| 1619 |
+
},
|
| 1620 |
+
{
|
| 1621 |
+
"epoch": 21.73,
|
| 1622 |
+
"learning_rate": 5e-06,
|
| 1623 |
+
"loss": 0.9587,
|
| 1624 |
+
"step": 26600
|
| 1625 |
+
},
|
| 1626 |
+
{
|
| 1627 |
+
"epoch": 21.81,
|
| 1628 |
+
"learning_rate": 5e-06,
|
| 1629 |
+
"loss": 0.9101,
|
| 1630 |
+
"step": 26700
|
| 1631 |
+
},
|
| 1632 |
+
{
|
| 1633 |
+
"epoch": 21.89,
|
| 1634 |
+
"learning_rate": 5e-06,
|
| 1635 |
+
"loss": 0.928,
|
| 1636 |
+
"step": 26800
|
| 1637 |
+
},
|
| 1638 |
+
{
|
| 1639 |
+
"epoch": 21.98,
|
| 1640 |
+
"learning_rate": 5e-06,
|
| 1641 |
+
"loss": 0.889,
|
| 1642 |
+
"step": 26900
|
| 1643 |
+
},
|
| 1644 |
+
{
|
| 1645 |
+
"epoch": 22.06,
|
| 1646 |
+
"learning_rate": 5e-06,
|
| 1647 |
+
"loss": 0.9088,
|
| 1648 |
+
"step": 27000
|
| 1649 |
+
},
|
| 1650 |
+
{
|
| 1651 |
+
"epoch": 22.14,
|
| 1652 |
+
"learning_rate": 5e-06,
|
| 1653 |
+
"loss": 0.9235,
|
| 1654 |
+
"step": 27100
|
| 1655 |
+
},
|
| 1656 |
+
{
|
| 1657 |
+
"epoch": 22.22,
|
| 1658 |
+
"learning_rate": 5e-06,
|
| 1659 |
+
"loss": 0.9353,
|
| 1660 |
+
"step": 27200
|
| 1661 |
+
},
|
| 1662 |
+
{
|
| 1663 |
+
"epoch": 22.3,
|
| 1664 |
+
"learning_rate": 5e-06,
|
| 1665 |
+
"loss": 0.9046,
|
| 1666 |
+
"step": 27300
|
| 1667 |
+
},
|
| 1668 |
+
{
|
| 1669 |
+
"epoch": 22.39,
|
| 1670 |
+
"learning_rate": 5e-06,
|
| 1671 |
+
"loss": 0.9106,
|
| 1672 |
+
"step": 27400
|
| 1673 |
+
},
|
| 1674 |
+
{
|
| 1675 |
+
"epoch": 22.47,
|
| 1676 |
+
"learning_rate": 5e-06,
|
| 1677 |
+
"loss": 0.9404,
|
| 1678 |
+
"step": 27500
|
| 1679 |
+
},
|
| 1680 |
+
{
|
| 1681 |
+
"epoch": 22.55,
|
| 1682 |
+
"learning_rate": 5e-06,
|
| 1683 |
+
"loss": 0.9081,
|
| 1684 |
+
"step": 27600
|
| 1685 |
+
},
|
| 1686 |
+
{
|
| 1687 |
+
"epoch": 22.63,
|
| 1688 |
+
"learning_rate": 5e-06,
|
| 1689 |
+
"loss": 0.9042,
|
| 1690 |
+
"step": 27700
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"epoch": 22.71,
|
| 1694 |
+
"learning_rate": 5e-06,
|
| 1695 |
+
"loss": 0.8971,
|
| 1696 |
+
"step": 27800
|
| 1697 |
+
},
|
| 1698 |
+
{
|
| 1699 |
+
"epoch": 22.79,
|
| 1700 |
+
"learning_rate": 5e-06,
|
| 1701 |
+
"loss": 0.9116,
|
| 1702 |
+
"step": 27900
|
| 1703 |
+
},
|
| 1704 |
+
{
|
| 1705 |
+
"epoch": 22.88,
|
| 1706 |
+
"learning_rate": 5e-06,
|
| 1707 |
+
"loss": 0.9135,
|
| 1708 |
+
"step": 28000
|
| 1709 |
+
},
|
| 1710 |
+
{
|
| 1711 |
+
"epoch": 22.96,
|
| 1712 |
+
"learning_rate": 5e-06,
|
| 1713 |
+
"loss": 0.9011,
|
| 1714 |
+
"step": 28100
|
| 1715 |
+
},
|
| 1716 |
+
{
|
| 1717 |
+
"epoch": 23.04,
|
| 1718 |
+
"learning_rate": 5e-06,
|
| 1719 |
+
"loss": 0.9025,
|
| 1720 |
+
"step": 28200
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"epoch": 23.12,
|
| 1724 |
+
"learning_rate": 5e-06,
|
| 1725 |
+
"loss": 0.9275,
|
| 1726 |
+
"step": 28300
|
| 1727 |
+
},
|
| 1728 |
+
{
|
| 1729 |
+
"epoch": 23.2,
|
| 1730 |
+
"learning_rate": 5e-06,
|
| 1731 |
+
"loss": 0.9086,
|
| 1732 |
+
"step": 28400
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"epoch": 23.28,
|
| 1736 |
+
"learning_rate": 5e-06,
|
| 1737 |
+
"loss": 0.9075,
|
| 1738 |
+
"step": 28500
|
| 1739 |
+
},
|
| 1740 |
+
{
|
| 1741 |
+
"epoch": 23.37,
|
| 1742 |
+
"learning_rate": 5e-06,
|
| 1743 |
+
"loss": 0.9022,
|
| 1744 |
+
"step": 28600
|
| 1745 |
+
},
|
| 1746 |
+
{
|
| 1747 |
+
"epoch": 23.45,
|
| 1748 |
+
"learning_rate": 5e-06,
|
| 1749 |
+
"loss": 0.9067,
|
| 1750 |
+
"step": 28700
|
| 1751 |
+
},
|
| 1752 |
+
{
|
| 1753 |
+
"epoch": 23.53,
|
| 1754 |
+
"learning_rate": 5e-06,
|
| 1755 |
+
"loss": 0.895,
|
| 1756 |
+
"step": 28800
|
| 1757 |
+
},
|
| 1758 |
+
{
|
| 1759 |
+
"epoch": 23.61,
|
| 1760 |
+
"learning_rate": 5e-06,
|
| 1761 |
+
"loss": 0.891,
|
| 1762 |
+
"step": 28900
|
| 1763 |
+
},
|
| 1764 |
+
{
|
| 1765 |
+
"epoch": 23.69,
|
| 1766 |
+
"learning_rate": 5e-06,
|
| 1767 |
+
"loss": 0.8775,
|
| 1768 |
+
"step": 29000
|
| 1769 |
+
},
|
| 1770 |
+
{
|
| 1771 |
+
"epoch": 23.77,
|
| 1772 |
+
"learning_rate": 5e-06,
|
| 1773 |
+
"loss": 0.8691,
|
| 1774 |
+
"step": 29100
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"epoch": 23.86,
|
| 1778 |
+
"learning_rate": 5e-06,
|
| 1779 |
+
"loss": 0.9149,
|
| 1780 |
+
"step": 29200
|
| 1781 |
+
},
|
| 1782 |
+
{
|
| 1783 |
+
"epoch": 23.94,
|
| 1784 |
+
"learning_rate": 5e-06,
|
| 1785 |
+
"loss": 0.9142,
|
| 1786 |
+
"step": 29300
|
| 1787 |
+
},
|
| 1788 |
+
{
|
| 1789 |
+
"epoch": 24.02,
|
| 1790 |
+
"learning_rate": 5e-06,
|
| 1791 |
+
"loss": 0.9338,
|
| 1792 |
+
"step": 29400
|
| 1793 |
+
},
|
| 1794 |
+
{
|
| 1795 |
+
"epoch": 24.1,
|
| 1796 |
+
"learning_rate": 5e-06,
|
| 1797 |
+
"loss": 0.8882,
|
| 1798 |
+
"step": 29500
|
| 1799 |
+
},
|
| 1800 |
+
{
|
| 1801 |
+
"epoch": 24.18,
|
| 1802 |
+
"learning_rate": 5e-06,
|
| 1803 |
+
"loss": 0.8561,
|
| 1804 |
+
"step": 29600
|
| 1805 |
+
},
|
| 1806 |
+
{
|
| 1807 |
+
"epoch": 24.26,
|
| 1808 |
+
"learning_rate": 5e-06,
|
| 1809 |
+
"loss": 0.8855,
|
| 1810 |
+
"step": 29700
|
| 1811 |
+
},
|
| 1812 |
+
{
|
| 1813 |
+
"epoch": 24.35,
|
| 1814 |
+
"learning_rate": 5e-06,
|
| 1815 |
+
"loss": 0.8879,
|
| 1816 |
+
"step": 29800
|
| 1817 |
+
},
|
| 1818 |
+
{
|
| 1819 |
+
"epoch": 24.43,
|
| 1820 |
+
"learning_rate": 5e-06,
|
| 1821 |
+
"loss": 0.9106,
|
| 1822 |
+
"step": 29900
|
| 1823 |
+
},
|
| 1824 |
+
{
|
| 1825 |
+
"epoch": 24.51,
|
| 1826 |
+
"learning_rate": 5e-06,
|
| 1827 |
+
"loss": 0.902,
|
| 1828 |
+
"step": 30000
|
| 1829 |
+
},
|
| 1830 |
+
{
|
| 1831 |
+
"epoch": 24.51,
|
| 1832 |
+
"eval_cer": 0.12203442394169639,
|
| 1833 |
+
"eval_loss": 0.547027051448822,
|
| 1834 |
+
"eval_runtime": 8.3805,
|
| 1835 |
+
"eval_samples_per_second": 35.798,
|
| 1836 |
+
"eval_steps_per_second": 17.899,
|
| 1837 |
+
"eval_wer": 0.4239316239316239,
|
| 1838 |
+
"step": 30000
|
| 1839 |
+
},
|
| 1840 |
+
{
|
| 1841 |
+
"epoch": 24.59,
|
| 1842 |
+
"learning_rate": 4.985666666666667e-06,
|
| 1843 |
+
"loss": 0.9302,
|
| 1844 |
+
"step": 30100
|
| 1845 |
+
},
|
| 1846 |
+
{
|
| 1847 |
+
"epoch": 24.67,
|
| 1848 |
+
"learning_rate": 4.969e-06,
|
| 1849 |
+
"loss": 0.8792,
|
| 1850 |
+
"step": 30200
|
| 1851 |
+
},
|
| 1852 |
+
{
|
| 1853 |
+
"epoch": 24.75,
|
| 1854 |
+
"learning_rate": 4.952333333333334e-06,
|
| 1855 |
+
"loss": 0.8448,
|
| 1856 |
+
"step": 30300
|
| 1857 |
+
},
|
| 1858 |
+
{
|
| 1859 |
+
"epoch": 24.84,
|
| 1860 |
+
"learning_rate": 4.935666666666667e-06,
|
| 1861 |
+
"loss": 0.8973,
|
| 1862 |
+
"step": 30400
|
| 1863 |
+
},
|
| 1864 |
+
{
|
| 1865 |
+
"epoch": 24.92,
|
| 1866 |
+
"learning_rate": 4.919000000000001e-06,
|
| 1867 |
+
"loss": 0.8425,
|
| 1868 |
+
"step": 30500
|
| 1869 |
+
},
|
| 1870 |
+
{
|
| 1871 |
+
"epoch": 25.0,
|
| 1872 |
+
"learning_rate": 4.902333333333334e-06,
|
| 1873 |
+
"loss": 0.8707,
|
| 1874 |
+
"step": 30600
|
| 1875 |
+
},
|
| 1876 |
+
{
|
| 1877 |
+
"epoch": 25.08,
|
| 1878 |
+
"learning_rate": 4.885666666666667e-06,
|
| 1879 |
+
"loss": 0.8867,
|
| 1880 |
+
"step": 30700
|
| 1881 |
+
},
|
| 1882 |
+
{
|
| 1883 |
+
"epoch": 25.16,
|
| 1884 |
+
"learning_rate": 4.869e-06,
|
| 1885 |
+
"loss": 0.8502,
|
| 1886 |
+
"step": 30800
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"epoch": 25.24,
|
| 1890 |
+
"learning_rate": 4.8523333333333335e-06,
|
| 1891 |
+
"loss": 0.8851,
|
| 1892 |
+
"step": 30900
|
| 1893 |
+
},
|
| 1894 |
+
{
|
| 1895 |
+
"epoch": 25.33,
|
| 1896 |
+
"learning_rate": 4.835666666666667e-06,
|
| 1897 |
+
"loss": 0.858,
|
| 1898 |
+
"step": 31000
|
| 1899 |
+
},
|
| 1900 |
+
{
|
| 1901 |
+
"epoch": 25.41,
|
| 1902 |
+
"learning_rate": 4.819e-06,
|
| 1903 |
+
"loss": 0.8725,
|
| 1904 |
+
"step": 31100
|
| 1905 |
+
},
|
| 1906 |
+
{
|
| 1907 |
+
"epoch": 25.49,
|
| 1908 |
+
"learning_rate": 4.802333333333334e-06,
|
| 1909 |
+
"loss": 0.8324,
|
| 1910 |
+
"step": 31200
|
| 1911 |
+
},
|
| 1912 |
+
{
|
| 1913 |
+
"epoch": 25.57,
|
| 1914 |
+
"learning_rate": 4.785666666666667e-06,
|
| 1915 |
+
"loss": 0.866,
|
| 1916 |
+
"step": 31300
|
| 1917 |
+
},
|
| 1918 |
+
{
|
| 1919 |
+
"epoch": 25.65,
|
| 1920 |
+
"learning_rate": 4.769e-06,
|
| 1921 |
+
"loss": 0.8658,
|
| 1922 |
+
"step": 31400
|
| 1923 |
+
},
|
| 1924 |
+
{
|
| 1925 |
+
"epoch": 25.73,
|
| 1926 |
+
"learning_rate": 4.752333333333334e-06,
|
| 1927 |
+
"loss": 0.8725,
|
| 1928 |
+
"step": 31500
|
| 1929 |
+
},
|
| 1930 |
+
{
|
| 1931 |
+
"epoch": 25.82,
|
| 1932 |
+
"learning_rate": 4.735666666666668e-06,
|
| 1933 |
+
"loss": 0.8853,
|
| 1934 |
+
"step": 31600
|
| 1935 |
+
},
|
| 1936 |
+
{
|
| 1937 |
+
"epoch": 25.9,
|
| 1938 |
+
"learning_rate": 4.719e-06,
|
| 1939 |
+
"loss": 0.8961,
|
| 1940 |
+
"step": 31700
|
| 1941 |
+
},
|
| 1942 |
+
{
|
| 1943 |
+
"epoch": 25.98,
|
| 1944 |
+
"learning_rate": 4.702333333333334e-06,
|
| 1945 |
+
"loss": 0.8206,
|
| 1946 |
+
"step": 31800
|
| 1947 |
+
},
|
| 1948 |
+
{
|
| 1949 |
+
"epoch": 26.06,
|
| 1950 |
+
"learning_rate": 4.685666666666667e-06,
|
| 1951 |
+
"loss": 0.8634,
|
| 1952 |
+
"step": 31900
|
| 1953 |
+
},
|
| 1954 |
+
{
|
| 1955 |
+
"epoch": 26.14,
|
| 1956 |
+
"learning_rate": 4.6690000000000005e-06,
|
| 1957 |
+
"loss": 0.9202,
|
| 1958 |
+
"step": 32000
|
| 1959 |
+
},
|
| 1960 |
+
{
|
| 1961 |
+
"epoch": 26.23,
|
| 1962 |
+
"learning_rate": 4.652333333333334e-06,
|
| 1963 |
+
"loss": 0.8333,
|
| 1964 |
+
"step": 32100
|
| 1965 |
+
},
|
| 1966 |
+
{
|
| 1967 |
+
"epoch": 26.31,
|
| 1968 |
+
"learning_rate": 4.635666666666667e-06,
|
| 1969 |
+
"loss": 0.8525,
|
| 1970 |
+
"step": 32200
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"epoch": 26.39,
|
| 1974 |
+
"learning_rate": 4.619e-06,
|
| 1975 |
+
"loss": 0.8417,
|
| 1976 |
+
"step": 32300
|
| 1977 |
+
},
|
| 1978 |
+
{
|
| 1979 |
+
"epoch": 26.47,
|
| 1980 |
+
"learning_rate": 4.602333333333333e-06,
|
| 1981 |
+
"loss": 0.8567,
|
| 1982 |
+
"step": 32400
|
| 1983 |
+
},
|
| 1984 |
+
{
|
| 1985 |
+
"epoch": 26.55,
|
| 1986 |
+
"learning_rate": 4.5856666666666674e-06,
|
| 1987 |
+
"loss": 0.8276,
|
| 1988 |
+
"step": 32500
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 26.63,
|
| 1992 |
+
"learning_rate": 4.569e-06,
|
| 1993 |
+
"loss": 0.8267,
|
| 1994 |
+
"step": 32600
|
| 1995 |
+
},
|
| 1996 |
+
{
|
| 1997 |
+
"epoch": 26.72,
|
| 1998 |
+
"learning_rate": 4.552333333333334e-06,
|
| 1999 |
+
"loss": 0.8312,
|
| 2000 |
+
"step": 32700
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 26.8,
|
| 2004 |
+
"learning_rate": 4.535666666666667e-06,
|
| 2005 |
+
"loss": 0.8272,
|
| 2006 |
+
"step": 32800
|
| 2007 |
+
},
|
| 2008 |
+
{
|
| 2009 |
+
"epoch": 26.88,
|
| 2010 |
+
"learning_rate": 4.519e-06,
|
| 2011 |
+
"loss": 0.9122,
|
| 2012 |
+
"step": 32900
|
| 2013 |
+
},
|
| 2014 |
+
{
|
| 2015 |
+
"epoch": 26.96,
|
| 2016 |
+
"learning_rate": 4.5023333333333335e-06,
|
| 2017 |
+
"loss": 0.862,
|
| 2018 |
+
"step": 33000
|
| 2019 |
+
},
|
| 2020 |
+
{
|
| 2021 |
+
"epoch": 27.04,
|
| 2022 |
+
"learning_rate": 4.485666666666667e-06,
|
| 2023 |
+
"loss": 0.882,
|
| 2024 |
+
"step": 33100
|
| 2025 |
+
},
|
| 2026 |
+
{
|
| 2027 |
+
"epoch": 27.12,
|
| 2028 |
+
"learning_rate": 4.469000000000001e-06,
|
| 2029 |
+
"loss": 0.8624,
|
| 2030 |
+
"step": 33200
|
| 2031 |
+
},
|
| 2032 |
+
{
|
| 2033 |
+
"epoch": 27.21,
|
| 2034 |
+
"learning_rate": 4.452333333333333e-06,
|
| 2035 |
+
"loss": 0.8011,
|
| 2036 |
+
"step": 33300
|
| 2037 |
+
},
|
| 2038 |
+
{
|
| 2039 |
+
"epoch": 27.29,
|
| 2040 |
+
"learning_rate": 4.435666666666667e-06,
|
| 2041 |
+
"loss": 0.8589,
|
| 2042 |
+
"step": 33400
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"epoch": 27.37,
|
| 2046 |
+
"learning_rate": 4.419e-06,
|
| 2047 |
+
"loss": 0.8521,
|
| 2048 |
+
"step": 33500
|
| 2049 |
+
},
|
| 2050 |
+
{
|
| 2051 |
+
"epoch": 27.45,
|
| 2052 |
+
"learning_rate": 4.402333333333334e-06,
|
| 2053 |
+
"loss": 0.8436,
|
| 2054 |
+
"step": 33600
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"epoch": 27.53,
|
| 2058 |
+
"learning_rate": 4.385666666666667e-06,
|
| 2059 |
+
"loss": 0.8348,
|
| 2060 |
+
"step": 33700
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 27.61,
|
| 2064 |
+
"learning_rate": 4.369000000000001e-06,
|
| 2065 |
+
"loss": 0.847,
|
| 2066 |
+
"step": 33800
|
| 2067 |
+
},
|
| 2068 |
+
{
|
| 2069 |
+
"epoch": 27.7,
|
| 2070 |
+
"learning_rate": 4.352333333333334e-06,
|
| 2071 |
+
"loss": 0.8742,
|
| 2072 |
+
"step": 33900
|
| 2073 |
+
},
|
| 2074 |
+
{
|
| 2075 |
+
"epoch": 27.78,
|
| 2076 |
+
"learning_rate": 4.335666666666667e-06,
|
| 2077 |
+
"loss": 0.8275,
|
| 2078 |
+
"step": 34000
|
| 2079 |
+
},
|
| 2080 |
+
{
|
| 2081 |
+
"epoch": 27.86,
|
| 2082 |
+
"learning_rate": 4.3190000000000005e-06,
|
| 2083 |
+
"loss": 0.8436,
|
| 2084 |
+
"step": 34100
|
| 2085 |
+
},
|
| 2086 |
+
{
|
| 2087 |
+
"epoch": 27.94,
|
| 2088 |
+
"learning_rate": 4.302333333333334e-06,
|
| 2089 |
+
"loss": 0.8261,
|
| 2090 |
+
"step": 34200
|
| 2091 |
+
},
|
| 2092 |
+
{
|
| 2093 |
+
"epoch": 28.02,
|
| 2094 |
+
"learning_rate": 4.285666666666667e-06,
|
| 2095 |
+
"loss": 0.7982,
|
| 2096 |
+
"step": 34300
|
| 2097 |
+
},
|
| 2098 |
+
{
|
| 2099 |
+
"epoch": 28.1,
|
| 2100 |
+
"learning_rate": 4.269e-06,
|
| 2101 |
+
"loss": 0.8012,
|
| 2102 |
+
"step": 34400
|
| 2103 |
+
},
|
| 2104 |
+
{
|
| 2105 |
+
"epoch": 28.19,
|
| 2106 |
+
"learning_rate": 4.2525e-06,
|
| 2107 |
+
"loss": 0.8682,
|
| 2108 |
+
"step": 34500
|
| 2109 |
+
},
|
| 2110 |
+
{
|
| 2111 |
+
"epoch": 28.27,
|
| 2112 |
+
"learning_rate": 4.235833333333334e-06,
|
| 2113 |
+
"loss": 0.8125,
|
| 2114 |
+
"step": 34600
|
| 2115 |
+
},
|
| 2116 |
+
{
|
| 2117 |
+
"epoch": 28.35,
|
| 2118 |
+
"learning_rate": 4.219166666666667e-06,
|
| 2119 |
+
"loss": 0.8077,
|
| 2120 |
+
"step": 34700
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 28.43,
|
| 2124 |
+
"learning_rate": 4.202500000000001e-06,
|
| 2125 |
+
"loss": 0.8447,
|
| 2126 |
+
"step": 34800
|
| 2127 |
+
},
|
| 2128 |
+
{
|
| 2129 |
+
"epoch": 28.51,
|
| 2130 |
+
"learning_rate": 4.185833333333333e-06,
|
| 2131 |
+
"loss": 0.8205,
|
| 2132 |
+
"step": 34900
|
| 2133 |
+
},
|
| 2134 |
+
{
|
| 2135 |
+
"epoch": 28.59,
|
| 2136 |
+
"learning_rate": 4.169166666666667e-06,
|
| 2137 |
+
"loss": 0.8297,
|
| 2138 |
+
"step": 35000
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"epoch": 28.68,
|
| 2142 |
+
"learning_rate": 4.1525000000000005e-06,
|
| 2143 |
+
"loss": 0.8311,
|
| 2144 |
+
"step": 35100
|
| 2145 |
+
},
|
| 2146 |
+
{
|
| 2147 |
+
"epoch": 28.76,
|
| 2148 |
+
"learning_rate": 4.135833333333334e-06,
|
| 2149 |
+
"loss": 0.818,
|
| 2150 |
+
"step": 35200
|
| 2151 |
+
},
|
| 2152 |
+
{
|
| 2153 |
+
"epoch": 28.84,
|
| 2154 |
+
"learning_rate": 4.119166666666667e-06,
|
| 2155 |
+
"loss": 0.8653,
|
| 2156 |
+
"step": 35300
|
| 2157 |
+
},
|
| 2158 |
+
{
|
| 2159 |
+
"epoch": 28.92,
|
| 2160 |
+
"learning_rate": 4.1025e-06,
|
| 2161 |
+
"loss": 0.8136,
|
| 2162 |
+
"step": 35400
|
| 2163 |
+
},
|
| 2164 |
+
{
|
| 2165 |
+
"epoch": 29.0,
|
| 2166 |
+
"learning_rate": 4.085833333333334e-06,
|
| 2167 |
+
"loss": 0.8276,
|
| 2168 |
+
"step": 35500
|
| 2169 |
+
},
|
| 2170 |
+
{
|
| 2171 |
+
"epoch": 29.08,
|
| 2172 |
+
"learning_rate": 4.0691666666666665e-06,
|
| 2173 |
+
"loss": 0.7686,
|
| 2174 |
+
"step": 35600
|
| 2175 |
+
},
|
| 2176 |
+
{
|
| 2177 |
+
"epoch": 29.17,
|
| 2178 |
+
"learning_rate": 4.052500000000001e-06,
|
| 2179 |
+
"loss": 0.856,
|
| 2180 |
+
"step": 35700
|
| 2181 |
+
},
|
| 2182 |
+
{
|
| 2183 |
+
"epoch": 29.25,
|
| 2184 |
+
"learning_rate": 4.035833333333334e-06,
|
| 2185 |
+
"loss": 0.803,
|
| 2186 |
+
"step": 35800
|
| 2187 |
+
},
|
| 2188 |
+
{
|
| 2189 |
+
"epoch": 29.33,
|
| 2190 |
+
"learning_rate": 4.019166666666667e-06,
|
| 2191 |
+
"loss": 0.827,
|
| 2192 |
+
"step": 35900
|
| 2193 |
+
},
|
| 2194 |
+
{
|
| 2195 |
+
"epoch": 29.41,
|
| 2196 |
+
"learning_rate": 4.0025e-06,
|
| 2197 |
+
"loss": 0.8293,
|
| 2198 |
+
"step": 36000
|
| 2199 |
+
},
|
| 2200 |
+
{
|
| 2201 |
+
"epoch": 29.49,
|
| 2202 |
+
"learning_rate": 3.9858333333333334e-06,
|
| 2203 |
+
"loss": 0.8098,
|
| 2204 |
+
"step": 36100
|
| 2205 |
+
},
|
| 2206 |
+
{
|
| 2207 |
+
"epoch": 29.57,
|
| 2208 |
+
"learning_rate": 3.969166666666667e-06,
|
| 2209 |
+
"loss": 0.8037,
|
| 2210 |
+
"step": 36200
|
| 2211 |
+
},
|
| 2212 |
+
{
|
| 2213 |
+
"epoch": 29.66,
|
| 2214 |
+
"learning_rate": 3.9525e-06,
|
| 2215 |
+
"loss": 0.8306,
|
| 2216 |
+
"step": 36300
|
| 2217 |
+
},
|
| 2218 |
+
{
|
| 2219 |
+
"epoch": 29.74,
|
| 2220 |
+
"learning_rate": 3.935833333333334e-06,
|
| 2221 |
+
"loss": 0.7893,
|
| 2222 |
+
"step": 36400
|
| 2223 |
+
},
|
| 2224 |
+
{
|
| 2225 |
+
"epoch": 29.82,
|
| 2226 |
+
"learning_rate": 3.919166666666667e-06,
|
| 2227 |
+
"loss": 0.8476,
|
| 2228 |
+
"step": 36500
|
| 2229 |
+
},
|
| 2230 |
+
{
|
| 2231 |
+
"epoch": 29.9,
|
| 2232 |
+
"learning_rate": 3.9025e-06,
|
| 2233 |
+
"loss": 0.8068,
|
| 2234 |
+
"step": 36600
|
| 2235 |
+
},
|
| 2236 |
+
{
|
| 2237 |
+
"epoch": 29.98,
|
| 2238 |
+
"learning_rate": 3.8858333333333335e-06,
|
| 2239 |
+
"loss": 0.8298,
|
| 2240 |
+
"step": 36700
|
| 2241 |
+
},
|
| 2242 |
+
{
|
| 2243 |
+
"epoch": 30.07,
|
| 2244 |
+
"learning_rate": 3.869333333333334e-06,
|
| 2245 |
+
"loss": 0.8144,
|
| 2246 |
+
"step": 36800
|
| 2247 |
+
},
|
| 2248 |
+
{
|
| 2249 |
+
"epoch": 30.15,
|
| 2250 |
+
"learning_rate": 3.852666666666667e-06,
|
| 2251 |
+
"loss": 0.7807,
|
| 2252 |
+
"step": 36900
|
| 2253 |
+
},
|
| 2254 |
+
{
|
| 2255 |
+
"epoch": 30.23,
|
| 2256 |
+
"learning_rate": 3.836e-06,
|
| 2257 |
+
"loss": 0.7881,
|
| 2258 |
+
"step": 37000
|
| 2259 |
+
},
|
| 2260 |
+
{
|
| 2261 |
+
"epoch": 30.31,
|
| 2262 |
+
"learning_rate": 3.819333333333333e-06,
|
| 2263 |
+
"loss": 0.8269,
|
| 2264 |
+
"step": 37100
|
| 2265 |
+
},
|
| 2266 |
+
{
|
| 2267 |
+
"epoch": 30.39,
|
| 2268 |
+
"learning_rate": 3.8026666666666666e-06,
|
| 2269 |
+
"loss": 0.8161,
|
| 2270 |
+
"step": 37200
|
| 2271 |
+
},
|
| 2272 |
+
{
|
| 2273 |
+
"epoch": 30.47,
|
| 2274 |
+
"learning_rate": 3.7860000000000003e-06,
|
| 2275 |
+
"loss": 0.8432,
|
| 2276 |
+
"step": 37300
|
| 2277 |
+
},
|
| 2278 |
+
{
|
| 2279 |
+
"epoch": 30.56,
|
| 2280 |
+
"learning_rate": 3.769333333333334e-06,
|
| 2281 |
+
"loss": 0.7769,
|
| 2282 |
+
"step": 37400
|
| 2283 |
+
},
|
| 2284 |
+
{
|
| 2285 |
+
"epoch": 30.64,
|
| 2286 |
+
"learning_rate": 3.752666666666667e-06,
|
| 2287 |
+
"loss": 0.79,
|
| 2288 |
+
"step": 37500
|
| 2289 |
+
},
|
| 2290 |
+
{
|
| 2291 |
+
"epoch": 30.72,
|
| 2292 |
+
"learning_rate": 3.7360000000000003e-06,
|
| 2293 |
+
"loss": 0.8203,
|
| 2294 |
+
"step": 37600
|
| 2295 |
+
},
|
| 2296 |
+
{
|
| 2297 |
+
"epoch": 30.8,
|
| 2298 |
+
"learning_rate": 3.7193333333333335e-06,
|
| 2299 |
+
"loss": 0.7892,
|
| 2300 |
+
"step": 37700
|
| 2301 |
+
},
|
| 2302 |
+
{
|
| 2303 |
+
"epoch": 30.88,
|
| 2304 |
+
"learning_rate": 3.702666666666667e-06,
|
| 2305 |
+
"loss": 0.8422,
|
| 2306 |
+
"step": 37800
|
| 2307 |
+
},
|
| 2308 |
+
{
|
| 2309 |
+
"epoch": 30.96,
|
| 2310 |
+
"learning_rate": 3.686e-06,
|
| 2311 |
+
"loss": 0.7969,
|
| 2312 |
+
"step": 37900
|
| 2313 |
+
},
|
| 2314 |
+
{
|
| 2315 |
+
"epoch": 31.05,
|
| 2316 |
+
"learning_rate": 3.6693333333333336e-06,
|
| 2317 |
+
"loss": 0.8289,
|
| 2318 |
+
"step": 38000
|
| 2319 |
+
},
|
| 2320 |
+
{
|
| 2321 |
+
"epoch": 31.13,
|
| 2322 |
+
"learning_rate": 3.6526666666666672e-06,
|
| 2323 |
+
"loss": 0.8019,
|
| 2324 |
+
"step": 38100
|
| 2325 |
+
},
|
| 2326 |
+
{
|
| 2327 |
+
"epoch": 31.21,
|
| 2328 |
+
"learning_rate": 3.636e-06,
|
| 2329 |
+
"loss": 0.8314,
|
| 2330 |
+
"step": 38200
|
| 2331 |
+
},
|
| 2332 |
+
{
|
| 2333 |
+
"epoch": 31.29,
|
| 2334 |
+
"learning_rate": 3.6193333333333336e-06,
|
| 2335 |
+
"loss": 0.8357,
|
| 2336 |
+
"step": 38300
|
| 2337 |
+
},
|
| 2338 |
+
{
|
| 2339 |
+
"epoch": 31.37,
|
| 2340 |
+
"learning_rate": 3.6026666666666673e-06,
|
| 2341 |
+
"loss": 0.7961,
|
| 2342 |
+
"step": 38400
|
| 2343 |
+
},
|
| 2344 |
+
{
|
| 2345 |
+
"epoch": 31.45,
|
| 2346 |
+
"learning_rate": 3.586e-06,
|
| 2347 |
+
"loss": 0.7876,
|
| 2348 |
+
"step": 38500
|
| 2349 |
+
},
|
| 2350 |
+
{
|
| 2351 |
+
"epoch": 31.54,
|
| 2352 |
+
"learning_rate": 3.5695000000000003e-06,
|
| 2353 |
+
"loss": 0.775,
|
| 2354 |
+
"step": 38600
|
| 2355 |
+
},
|
| 2356 |
+
{
|
| 2357 |
+
"epoch": 31.62,
|
| 2358 |
+
"learning_rate": 3.5528333333333335e-06,
|
| 2359 |
+
"loss": 0.8121,
|
| 2360 |
+
"step": 38700
|
| 2361 |
+
},
|
| 2362 |
+
{
|
| 2363 |
+
"epoch": 31.7,
|
| 2364 |
+
"learning_rate": 3.536166666666667e-06,
|
| 2365 |
+
"loss": 0.7963,
|
| 2366 |
+
"step": 38800
|
| 2367 |
+
},
|
| 2368 |
+
{
|
| 2369 |
+
"epoch": 31.78,
|
| 2370 |
+
"learning_rate": 3.5195e-06,
|
| 2371 |
+
"loss": 0.7908,
|
| 2372 |
+
"step": 38900
|
| 2373 |
+
},
|
| 2374 |
+
{
|
| 2375 |
+
"epoch": 31.86,
|
| 2376 |
+
"learning_rate": 3.5028333333333336e-06,
|
| 2377 |
+
"loss": 0.8165,
|
| 2378 |
+
"step": 39000
|
| 2379 |
+
},
|
| 2380 |
+
{
|
| 2381 |
+
"epoch": 31.94,
|
| 2382 |
+
"learning_rate": 3.486166666666667e-06,
|
| 2383 |
+
"loss": 0.765,
|
| 2384 |
+
"step": 39100
|
| 2385 |
+
},
|
| 2386 |
+
{
|
| 2387 |
+
"epoch": 32.03,
|
| 2388 |
+
"learning_rate": 3.4695e-06,
|
| 2389 |
+
"loss": 0.7875,
|
| 2390 |
+
"step": 39200
|
| 2391 |
+
},
|
| 2392 |
+
{
|
| 2393 |
+
"epoch": 32.11,
|
| 2394 |
+
"learning_rate": 3.4528333333333336e-06,
|
| 2395 |
+
"loss": 0.8024,
|
| 2396 |
+
"step": 39300
|
| 2397 |
+
},
|
| 2398 |
+
{
|
| 2399 |
+
"epoch": 32.19,
|
| 2400 |
+
"learning_rate": 3.4361666666666673e-06,
|
| 2401 |
+
"loss": 0.7563,
|
| 2402 |
+
"step": 39400
|
| 2403 |
+
},
|
| 2404 |
+
{
|
| 2405 |
+
"epoch": 32.27,
|
| 2406 |
+
"learning_rate": 3.4195e-06,
|
| 2407 |
+
"loss": 0.8206,
|
| 2408 |
+
"step": 39500
|
| 2409 |
+
},
|
| 2410 |
+
{
|
| 2411 |
+
"epoch": 32.35,
|
| 2412 |
+
"learning_rate": 3.4028333333333337e-06,
|
| 2413 |
+
"loss": 0.7858,
|
| 2414 |
+
"step": 39600
|
| 2415 |
+
},
|
| 2416 |
+
{
|
| 2417 |
+
"epoch": 32.43,
|
| 2418 |
+
"learning_rate": 3.386166666666667e-06,
|
| 2419 |
+
"loss": 0.7956,
|
| 2420 |
+
"step": 39700
|
| 2421 |
+
},
|
| 2422 |
+
{
|
| 2423 |
+
"epoch": 32.52,
|
| 2424 |
+
"learning_rate": 3.3695000000000005e-06,
|
| 2425 |
+
"loss": 0.7966,
|
| 2426 |
+
"step": 39800
|
| 2427 |
+
},
|
| 2428 |
+
{
|
| 2429 |
+
"epoch": 32.6,
|
| 2430 |
+
"learning_rate": 3.3528333333333333e-06,
|
| 2431 |
+
"loss": 0.8083,
|
| 2432 |
+
"step": 39900
|
| 2433 |
+
},
|
| 2434 |
+
{
|
| 2435 |
+
"epoch": 32.68,
|
| 2436 |
+
"learning_rate": 3.336166666666667e-06,
|
| 2437 |
+
"loss": 0.7696,
|
| 2438 |
+
"step": 40000
|
| 2439 |
+
},
|
| 2440 |
+
{
|
| 2441 |
+
"epoch": 32.68,
|
| 2442 |
+
"eval_cer": 0.11156768491238951,
|
| 2443 |
+
"eval_loss": 0.5117331147193909,
|
| 2444 |
+
"eval_runtime": 8.1077,
|
| 2445 |
+
"eval_samples_per_second": 37.002,
|
| 2446 |
+
"eval_steps_per_second": 18.501,
|
| 2447 |
+
"eval_wer": 0.3871794871794872,
|
| 2448 |
+
"step": 40000
|
| 2449 |
+
},
|
| 2450 |
+
{
|
| 2451 |
+
"epoch": 32.76,
|
| 2452 |
+
"learning_rate": 3.3195000000000006e-06,
|
| 2453 |
+
"loss": 0.7772,
|
| 2454 |
+
"step": 40100
|
| 2455 |
+
},
|
| 2456 |
+
{
|
| 2457 |
+
"epoch": 32.84,
|
| 2458 |
+
"learning_rate": 3.3028333333333334e-06,
|
| 2459 |
+
"loss": 0.7818,
|
| 2460 |
+
"step": 40200
|
| 2461 |
+
},
|
| 2462 |
+
{
|
| 2463 |
+
"epoch": 32.92,
|
| 2464 |
+
"learning_rate": 3.286166666666667e-06,
|
| 2465 |
+
"loss": 0.8184,
|
| 2466 |
+
"step": 40300
|
| 2467 |
+
},
|
| 2468 |
+
{
|
| 2469 |
+
"epoch": 33.01,
|
| 2470 |
+
"learning_rate": 3.2695000000000006e-06,
|
| 2471 |
+
"loss": 0.8117,
|
| 2472 |
+
"step": 40400
|
| 2473 |
+
},
|
| 2474 |
+
{
|
| 2475 |
+
"epoch": 33.09,
|
| 2476 |
+
"learning_rate": 3.2528333333333334e-06,
|
| 2477 |
+
"loss": 0.7759,
|
| 2478 |
+
"step": 40500
|
| 2479 |
+
},
|
| 2480 |
+
{
|
| 2481 |
+
"epoch": 33.17,
|
| 2482 |
+
"learning_rate": 3.236166666666667e-06,
|
| 2483 |
+
"loss": 0.8274,
|
| 2484 |
+
"step": 40600
|
| 2485 |
+
},
|
| 2486 |
+
{
|
| 2487 |
+
"epoch": 33.25,
|
| 2488 |
+
"learning_rate": 3.2195000000000003e-06,
|
| 2489 |
+
"loss": 0.7923,
|
| 2490 |
+
"step": 40700
|
| 2491 |
+
},
|
| 2492 |
+
{
|
| 2493 |
+
"epoch": 33.33,
|
| 2494 |
+
"learning_rate": 3.2028333333333335e-06,
|
| 2495 |
+
"loss": 0.7858,
|
| 2496 |
+
"step": 40800
|
| 2497 |
+
},
|
| 2498 |
+
{
|
| 2499 |
+
"epoch": 33.41,
|
| 2500 |
+
"learning_rate": 3.1861666666666667e-06,
|
| 2501 |
+
"loss": 0.7894,
|
| 2502 |
+
"step": 40900
|
| 2503 |
+
},
|
| 2504 |
+
{
|
| 2505 |
+
"epoch": 33.5,
|
| 2506 |
+
"learning_rate": 3.169666666666667e-06,
|
| 2507 |
+
"loss": 0.7956,
|
| 2508 |
+
"step": 41000
|
| 2509 |
+
},
|
| 2510 |
+
{
|
| 2511 |
+
"epoch": 33.58,
|
| 2512 |
+
"learning_rate": 3.1530000000000006e-06,
|
| 2513 |
+
"loss": 0.7691,
|
| 2514 |
+
"step": 41100
|
| 2515 |
+
},
|
| 2516 |
+
{
|
| 2517 |
+
"epoch": 33.66,
|
| 2518 |
+
"learning_rate": 3.1363333333333333e-06,
|
| 2519 |
+
"loss": 0.7619,
|
| 2520 |
+
"step": 41200
|
| 2521 |
+
},
|
| 2522 |
+
{
|
| 2523 |
+
"epoch": 33.74,
|
| 2524 |
+
"learning_rate": 3.119666666666667e-06,
|
| 2525 |
+
"loss": 0.7676,
|
| 2526 |
+
"step": 41300
|
| 2527 |
+
},
|
| 2528 |
+
{
|
| 2529 |
+
"epoch": 33.82,
|
| 2530 |
+
"learning_rate": 3.1030000000000006e-06,
|
| 2531 |
+
"loss": 0.8083,
|
| 2532 |
+
"step": 41400
|
| 2533 |
+
},
|
| 2534 |
+
{
|
| 2535 |
+
"epoch": 33.9,
|
| 2536 |
+
"learning_rate": 3.0863333333333334e-06,
|
| 2537 |
+
"loss": 0.7691,
|
| 2538 |
+
"step": 41500
|
| 2539 |
+
},
|
| 2540 |
+
{
|
| 2541 |
+
"epoch": 33.99,
|
| 2542 |
+
"learning_rate": 3.069666666666667e-06,
|
| 2543 |
+
"loss": 0.746,
|
| 2544 |
+
"step": 41600
|
| 2545 |
+
},
|
| 2546 |
+
{
|
| 2547 |
+
"epoch": 34.07,
|
| 2548 |
+
"learning_rate": 3.0530000000000002e-06,
|
| 2549 |
+
"loss": 0.7611,
|
| 2550 |
+
"step": 41700
|
| 2551 |
+
},
|
| 2552 |
+
{
|
| 2553 |
+
"epoch": 34.15,
|
| 2554 |
+
"learning_rate": 3.0363333333333335e-06,
|
| 2555 |
+
"loss": 0.7673,
|
| 2556 |
+
"step": 41800
|
| 2557 |
+
},
|
| 2558 |
+
{
|
| 2559 |
+
"epoch": 34.23,
|
| 2560 |
+
"learning_rate": 3.0196666666666667e-06,
|
| 2561 |
+
"loss": 0.7686,
|
| 2562 |
+
"step": 41900
|
| 2563 |
+
},
|
| 2564 |
+
{
|
| 2565 |
+
"epoch": 34.31,
|
| 2566 |
+
"learning_rate": 3.0030000000000003e-06,
|
| 2567 |
+
"loss": 0.7694,
|
| 2568 |
+
"step": 42000
|
| 2569 |
+
},
|
| 2570 |
+
{
|
| 2571 |
+
"epoch": 34.4,
|
| 2572 |
+
"learning_rate": 2.9863333333333335e-06,
|
| 2573 |
+
"loss": 0.815,
|
| 2574 |
+
"step": 42100
|
| 2575 |
+
},
|
| 2576 |
+
{
|
| 2577 |
+
"epoch": 34.48,
|
| 2578 |
+
"learning_rate": 2.9696666666666667e-06,
|
| 2579 |
+
"loss": 0.7531,
|
| 2580 |
+
"step": 42200
|
| 2581 |
+
},
|
| 2582 |
+
{
|
| 2583 |
+
"epoch": 34.56,
|
| 2584 |
+
"learning_rate": 2.9530000000000004e-06,
|
| 2585 |
+
"loss": 0.753,
|
| 2586 |
+
"step": 42300
|
| 2587 |
+
},
|
| 2588 |
+
{
|
| 2589 |
+
"epoch": 34.64,
|
| 2590 |
+
"learning_rate": 2.936333333333334e-06,
|
| 2591 |
+
"loss": 0.7748,
|
| 2592 |
+
"step": 42400
|
| 2593 |
+
},
|
| 2594 |
+
{
|
| 2595 |
+
"epoch": 34.72,
|
| 2596 |
+
"learning_rate": 2.9196666666666668e-06,
|
| 2597 |
+
"loss": 0.7828,
|
| 2598 |
+
"step": 42500
|
| 2599 |
+
},
|
| 2600 |
+
{
|
| 2601 |
+
"epoch": 34.8,
|
| 2602 |
+
"learning_rate": 2.9030000000000004e-06,
|
| 2603 |
+
"loss": 0.7849,
|
| 2604 |
+
"step": 42600
|
| 2605 |
+
},
|
| 2606 |
+
{
|
| 2607 |
+
"epoch": 34.89,
|
| 2608 |
+
"learning_rate": 2.8863333333333336e-06,
|
| 2609 |
+
"loss": 0.8083,
|
| 2610 |
+
"step": 42700
|
| 2611 |
+
},
|
| 2612 |
+
{
|
| 2613 |
+
"epoch": 34.97,
|
| 2614 |
+
"learning_rate": 2.869666666666667e-06,
|
| 2615 |
+
"loss": 0.7838,
|
| 2616 |
+
"step": 42800
|
| 2617 |
+
},
|
| 2618 |
+
{
|
| 2619 |
+
"epoch": 35.05,
|
| 2620 |
+
"learning_rate": 2.853e-06,
|
| 2621 |
+
"loss": 0.8093,
|
| 2622 |
+
"step": 42900
|
| 2623 |
+
},
|
| 2624 |
+
{
|
| 2625 |
+
"epoch": 35.13,
|
| 2626 |
+
"learning_rate": 2.8363333333333337e-06,
|
| 2627 |
+
"loss": 0.7689,
|
| 2628 |
+
"step": 43000
|
| 2629 |
+
},
|
| 2630 |
+
{
|
| 2631 |
+
"epoch": 35.21,
|
| 2632 |
+
"learning_rate": 2.8196666666666665e-06,
|
| 2633 |
+
"loss": 0.7665,
|
| 2634 |
+
"step": 43100
|
| 2635 |
+
},
|
| 2636 |
+
{
|
| 2637 |
+
"epoch": 35.29,
|
| 2638 |
+
"learning_rate": 2.803e-06,
|
| 2639 |
+
"loss": 0.7172,
|
| 2640 |
+
"step": 43200
|
| 2641 |
+
},
|
| 2642 |
+
{
|
| 2643 |
+
"epoch": 35.38,
|
| 2644 |
+
"learning_rate": 2.7863333333333337e-06,
|
| 2645 |
+
"loss": 0.7646,
|
| 2646 |
+
"step": 43300
|
| 2647 |
+
},
|
| 2648 |
+
{
|
| 2649 |
+
"epoch": 35.46,
|
| 2650 |
+
"learning_rate": 2.7696666666666674e-06,
|
| 2651 |
+
"loss": 0.7874,
|
| 2652 |
+
"step": 43400
|
| 2653 |
+
},
|
| 2654 |
+
{
|
| 2655 |
+
"epoch": 35.54,
|
| 2656 |
+
"learning_rate": 2.753e-06,
|
| 2657 |
+
"loss": 0.8095,
|
| 2658 |
+
"step": 43500
|
| 2659 |
+
},
|
| 2660 |
+
{
|
| 2661 |
+
"epoch": 35.62,
|
| 2662 |
+
"learning_rate": 2.7363333333333338e-06,
|
| 2663 |
+
"loss": 0.7498,
|
| 2664 |
+
"step": 43600
|
| 2665 |
+
},
|
| 2666 |
+
{
|
| 2667 |
+
"epoch": 35.7,
|
| 2668 |
+
"learning_rate": 2.719666666666667e-06,
|
| 2669 |
+
"loss": 0.7713,
|
| 2670 |
+
"step": 43700
|
| 2671 |
+
},
|
| 2672 |
+
{
|
| 2673 |
+
"epoch": 35.78,
|
| 2674 |
+
"learning_rate": 2.703e-06,
|
| 2675 |
+
"loss": 0.7965,
|
| 2676 |
+
"step": 43800
|
| 2677 |
+
},
|
| 2678 |
+
{
|
| 2679 |
+
"epoch": 35.87,
|
| 2680 |
+
"learning_rate": 2.6863333333333334e-06,
|
| 2681 |
+
"loss": 0.7848,
|
| 2682 |
+
"step": 43900
|
| 2683 |
+
},
|
| 2684 |
+
{
|
| 2685 |
+
"epoch": 35.95,
|
| 2686 |
+
"learning_rate": 2.669666666666667e-06,
|
| 2687 |
+
"loss": 0.7648,
|
| 2688 |
+
"step": 44000
|
| 2689 |
+
},
|
| 2690 |
+
{
|
| 2691 |
+
"epoch": 36.03,
|
| 2692 |
+
"learning_rate": 2.653e-06,
|
| 2693 |
+
"loss": 0.757,
|
| 2694 |
+
"step": 44100
|
| 2695 |
+
},
|
| 2696 |
+
{
|
| 2697 |
+
"epoch": 36.11,
|
| 2698 |
+
"learning_rate": 2.6363333333333335e-06,
|
| 2699 |
+
"loss": 0.7863,
|
| 2700 |
+
"step": 44200
|
| 2701 |
+
},
|
| 2702 |
+
{
|
| 2703 |
+
"epoch": 36.19,
|
| 2704 |
+
"learning_rate": 2.619666666666667e-06,
|
| 2705 |
+
"loss": 0.7656,
|
| 2706 |
+
"step": 44300
|
| 2707 |
+
},
|
| 2708 |
+
{
|
| 2709 |
+
"epoch": 36.27,
|
| 2710 |
+
"learning_rate": 2.603e-06,
|
| 2711 |
+
"loss": 0.7338,
|
| 2712 |
+
"step": 44400
|
| 2713 |
+
},
|
| 2714 |
+
{
|
| 2715 |
+
"epoch": 36.36,
|
| 2716 |
+
"learning_rate": 2.5863333333333335e-06,
|
| 2717 |
+
"loss": 0.7977,
|
| 2718 |
+
"step": 44500
|
| 2719 |
+
},
|
| 2720 |
+
{
|
| 2721 |
+
"epoch": 36.44,
|
| 2722 |
+
"learning_rate": 2.569666666666667e-06,
|
| 2723 |
+
"loss": 0.7555,
|
| 2724 |
+
"step": 44600
|
| 2725 |
+
},
|
| 2726 |
+
{
|
| 2727 |
+
"epoch": 36.52,
|
| 2728 |
+
"learning_rate": 2.5530000000000004e-06,
|
| 2729 |
+
"loss": 0.7268,
|
| 2730 |
+
"step": 44700
|
| 2731 |
+
},
|
| 2732 |
+
{
|
| 2733 |
+
"epoch": 36.6,
|
| 2734 |
+
"learning_rate": 2.5363333333333336e-06,
|
| 2735 |
+
"loss": 0.7737,
|
| 2736 |
+
"step": 44800
|
| 2737 |
+
},
|
| 2738 |
+
{
|
| 2739 |
+
"epoch": 36.68,
|
| 2740 |
+
"learning_rate": 2.519666666666667e-06,
|
| 2741 |
+
"loss": 0.7716,
|
| 2742 |
+
"step": 44900
|
| 2743 |
+
},
|
| 2744 |
+
{
|
| 2745 |
+
"epoch": 36.76,
|
| 2746 |
+
"learning_rate": 2.5030000000000004e-06,
|
| 2747 |
+
"loss": 0.7961,
|
| 2748 |
+
"step": 45000
|
| 2749 |
+
},
|
| 2750 |
+
{
|
| 2751 |
+
"epoch": 36.85,
|
| 2752 |
+
"learning_rate": 2.4863333333333336e-06,
|
| 2753 |
+
"loss": 0.73,
|
| 2754 |
+
"step": 45100
|
| 2755 |
+
},
|
| 2756 |
+
{
|
| 2757 |
+
"epoch": 36.93,
|
| 2758 |
+
"learning_rate": 2.4698333333333335e-06,
|
| 2759 |
+
"loss": 0.7859,
|
| 2760 |
+
"step": 45200
|
| 2761 |
+
},
|
| 2762 |
+
{
|
| 2763 |
+
"epoch": 37.01,
|
| 2764 |
+
"learning_rate": 2.4531666666666667e-06,
|
| 2765 |
+
"loss": 0.7968,
|
| 2766 |
+
"step": 45300
|
| 2767 |
+
},
|
| 2768 |
+
{
|
| 2769 |
+
"epoch": 37.09,
|
| 2770 |
+
"learning_rate": 2.4365000000000003e-06,
|
| 2771 |
+
"loss": 0.7851,
|
| 2772 |
+
"step": 45400
|
| 2773 |
+
},
|
| 2774 |
+
{
|
| 2775 |
+
"epoch": 37.17,
|
| 2776 |
+
"learning_rate": 2.4198333333333335e-06,
|
| 2777 |
+
"loss": 0.7777,
|
| 2778 |
+
"step": 45500
|
| 2779 |
+
},
|
| 2780 |
+
{
|
| 2781 |
+
"epoch": 37.25,
|
| 2782 |
+
"learning_rate": 2.403166666666667e-06,
|
| 2783 |
+
"loss": 0.7303,
|
| 2784 |
+
"step": 45600
|
| 2785 |
+
},
|
| 2786 |
+
{
|
| 2787 |
+
"epoch": 37.34,
|
| 2788 |
+
"learning_rate": 2.3865000000000004e-06,
|
| 2789 |
+
"loss": 0.7871,
|
| 2790 |
+
"step": 45700
|
| 2791 |
+
},
|
| 2792 |
+
{
|
| 2793 |
+
"epoch": 37.42,
|
| 2794 |
+
"learning_rate": 2.3698333333333336e-06,
|
| 2795 |
+
"loss": 0.7566,
|
| 2796 |
+
"step": 45800
|
| 2797 |
+
},
|
| 2798 |
+
{
|
| 2799 |
+
"epoch": 37.5,
|
| 2800 |
+
"learning_rate": 2.3531666666666668e-06,
|
| 2801 |
+
"loss": 0.7616,
|
| 2802 |
+
"step": 45900
|
| 2803 |
+
},
|
| 2804 |
+
{
|
| 2805 |
+
"epoch": 37.58,
|
| 2806 |
+
"learning_rate": 2.3365e-06,
|
| 2807 |
+
"loss": 0.7324,
|
| 2808 |
+
"step": 46000
|
| 2809 |
+
},
|
| 2810 |
+
{
|
| 2811 |
+
"epoch": 37.66,
|
| 2812 |
+
"learning_rate": 2.3198333333333336e-06,
|
| 2813 |
+
"loss": 0.7485,
|
| 2814 |
+
"step": 46100
|
| 2815 |
+
},
|
| 2816 |
+
{
|
| 2817 |
+
"epoch": 37.74,
|
| 2818 |
+
"learning_rate": 2.303166666666667e-06,
|
| 2819 |
+
"loss": 0.7578,
|
| 2820 |
+
"step": 46200
|
| 2821 |
+
},
|
| 2822 |
+
{
|
| 2823 |
+
"epoch": 37.83,
|
| 2824 |
+
"learning_rate": 2.2865e-06,
|
| 2825 |
+
"loss": 0.7522,
|
| 2826 |
+
"step": 46300
|
| 2827 |
+
},
|
| 2828 |
+
{
|
| 2829 |
+
"epoch": 37.91,
|
| 2830 |
+
"learning_rate": 2.2698333333333337e-06,
|
| 2831 |
+
"loss": 0.7478,
|
| 2832 |
+
"step": 46400
|
| 2833 |
+
},
|
| 2834 |
+
{
|
| 2835 |
+
"epoch": 37.99,
|
| 2836 |
+
"learning_rate": 2.253166666666667e-06,
|
| 2837 |
+
"loss": 0.7818,
|
| 2838 |
+
"step": 46500
|
| 2839 |
+
},
|
| 2840 |
+
{
|
| 2841 |
+
"epoch": 38.07,
|
| 2842 |
+
"learning_rate": 2.2365e-06,
|
| 2843 |
+
"loss": 0.7424,
|
| 2844 |
+
"step": 46600
|
| 2845 |
+
},
|
| 2846 |
+
{
|
| 2847 |
+
"epoch": 38.15,
|
| 2848 |
+
"learning_rate": 2.2198333333333337e-06,
|
| 2849 |
+
"loss": 0.7343,
|
| 2850 |
+
"step": 46700
|
| 2851 |
+
},
|
| 2852 |
+
{
|
| 2853 |
+
"epoch": 38.24,
|
| 2854 |
+
"learning_rate": 2.203166666666667e-06,
|
| 2855 |
+
"loss": 0.779,
|
| 2856 |
+
"step": 46800
|
| 2857 |
+
},
|
| 2858 |
+
{
|
| 2859 |
+
"epoch": 38.32,
|
| 2860 |
+
"learning_rate": 2.1865e-06,
|
| 2861 |
+
"loss": 0.7509,
|
| 2862 |
+
"step": 46900
|
| 2863 |
+
},
|
| 2864 |
+
{
|
| 2865 |
+
"epoch": 38.4,
|
| 2866 |
+
"learning_rate": 2.1698333333333334e-06,
|
| 2867 |
+
"loss": 0.7452,
|
| 2868 |
+
"step": 47000
|
| 2869 |
+
},
|
| 2870 |
+
{
|
| 2871 |
+
"epoch": 38.48,
|
| 2872 |
+
"learning_rate": 2.1531666666666666e-06,
|
| 2873 |
+
"loss": 0.7771,
|
| 2874 |
+
"step": 47100
|
| 2875 |
+
},
|
| 2876 |
+
{
|
| 2877 |
+
"epoch": 38.56,
|
| 2878 |
+
"learning_rate": 2.1365000000000002e-06,
|
| 2879 |
+
"loss": 0.7662,
|
| 2880 |
+
"step": 47200
|
| 2881 |
+
},
|
| 2882 |
+
{
|
| 2883 |
+
"epoch": 38.64,
|
| 2884 |
+
"learning_rate": 2.1198333333333334e-06,
|
| 2885 |
+
"loss": 0.7476,
|
| 2886 |
+
"step": 47300
|
| 2887 |
+
},
|
| 2888 |
+
{
|
| 2889 |
+
"epoch": 38.73,
|
| 2890 |
+
"learning_rate": 2.103166666666667e-06,
|
| 2891 |
+
"loss": 0.7692,
|
| 2892 |
+
"step": 47400
|
| 2893 |
+
},
|
| 2894 |
+
{
|
| 2895 |
+
"epoch": 38.81,
|
| 2896 |
+
"learning_rate": 2.0865000000000003e-06,
|
| 2897 |
+
"loss": 0.7256,
|
| 2898 |
+
"step": 47500
|
| 2899 |
+
},
|
| 2900 |
+
{
|
| 2901 |
+
"epoch": 38.89,
|
| 2902 |
+
"learning_rate": 2.0698333333333335e-06,
|
| 2903 |
+
"loss": 0.7455,
|
| 2904 |
+
"step": 47600
|
| 2905 |
+
},
|
| 2906 |
+
{
|
| 2907 |
+
"epoch": 38.97,
|
| 2908 |
+
"learning_rate": 2.053166666666667e-06,
|
| 2909 |
+
"loss": 0.7437,
|
| 2910 |
+
"step": 47700
|
| 2911 |
+
},
|
| 2912 |
+
{
|
| 2913 |
+
"epoch": 39.05,
|
| 2914 |
+
"learning_rate": 2.0365000000000003e-06,
|
| 2915 |
+
"loss": 0.7618,
|
| 2916 |
+
"step": 47800
|
| 2917 |
+
},
|
| 2918 |
+
{
|
| 2919 |
+
"epoch": 39.13,
|
| 2920 |
+
"learning_rate": 2.0198333333333335e-06,
|
| 2921 |
+
"loss": 0.731,
|
| 2922 |
+
"step": 47900
|
| 2923 |
+
},
|
| 2924 |
+
{
|
| 2925 |
+
"epoch": 39.22,
|
| 2926 |
+
"learning_rate": 2.0031666666666667e-06,
|
| 2927 |
+
"loss": 0.7739,
|
| 2928 |
+
"step": 48000
|
| 2929 |
+
},
|
| 2930 |
+
{
|
| 2931 |
+
"epoch": 39.3,
|
| 2932 |
+
"learning_rate": 1.9865e-06,
|
| 2933 |
+
"loss": 0.7263,
|
| 2934 |
+
"step": 48100
|
| 2935 |
+
},
|
| 2936 |
+
{
|
| 2937 |
+
"epoch": 39.38,
|
| 2938 |
+
"learning_rate": 1.9698333333333336e-06,
|
| 2939 |
+
"loss": 0.7434,
|
| 2940 |
+
"step": 48200
|
| 2941 |
+
},
|
| 2942 |
+
{
|
| 2943 |
+
"epoch": 39.46,
|
| 2944 |
+
"learning_rate": 1.953166666666667e-06,
|
| 2945 |
+
"loss": 0.7491,
|
| 2946 |
+
"step": 48300
|
| 2947 |
+
},
|
| 2948 |
+
{
|
| 2949 |
+
"epoch": 39.54,
|
| 2950 |
+
"learning_rate": 1.9365e-06,
|
| 2951 |
+
"loss": 0.7346,
|
| 2952 |
+
"step": 48400
|
| 2953 |
+
},
|
| 2954 |
+
{
|
| 2955 |
+
"epoch": 39.62,
|
| 2956 |
+
"learning_rate": 1.9198333333333336e-06,
|
| 2957 |
+
"loss": 0.7308,
|
| 2958 |
+
"step": 48500
|
| 2959 |
+
},
|
| 2960 |
+
{
|
| 2961 |
+
"epoch": 39.71,
|
| 2962 |
+
"learning_rate": 1.9031666666666669e-06,
|
| 2963 |
+
"loss": 0.7484,
|
| 2964 |
+
"step": 48600
|
| 2965 |
+
},
|
| 2966 |
+
{
|
| 2967 |
+
"epoch": 39.79,
|
| 2968 |
+
"learning_rate": 1.8865000000000003e-06,
|
| 2969 |
+
"loss": 0.7621,
|
| 2970 |
+
"step": 48700
|
| 2971 |
+
},
|
| 2972 |
+
{
|
| 2973 |
+
"epoch": 39.87,
|
| 2974 |
+
"learning_rate": 1.8698333333333335e-06,
|
| 2975 |
+
"loss": 0.7626,
|
| 2976 |
+
"step": 48800
|
| 2977 |
+
},
|
| 2978 |
+
{
|
| 2979 |
+
"epoch": 39.95,
|
| 2980 |
+
"learning_rate": 1.8531666666666667e-06,
|
| 2981 |
+
"loss": 0.7763,
|
| 2982 |
+
"step": 48900
|
| 2983 |
+
},
|
| 2984 |
+
{
|
| 2985 |
+
"epoch": 40.03,
|
| 2986 |
+
"learning_rate": 1.8365000000000003e-06,
|
| 2987 |
+
"loss": 0.7559,
|
| 2988 |
+
"step": 49000
|
| 2989 |
+
},
|
| 2990 |
+
{
|
| 2991 |
+
"epoch": 40.11,
|
| 2992 |
+
"learning_rate": 1.8198333333333335e-06,
|
| 2993 |
+
"loss": 0.7356,
|
| 2994 |
+
"step": 49100
|
| 2995 |
+
},
|
| 2996 |
+
{
|
| 2997 |
+
"epoch": 40.2,
|
| 2998 |
+
"learning_rate": 1.8031666666666668e-06,
|
| 2999 |
+
"loss": 0.7245,
|
| 3000 |
+
"step": 49200
|
| 3001 |
+
},
|
| 3002 |
+
{
|
| 3003 |
+
"epoch": 40.28,
|
| 3004 |
+
"learning_rate": 1.7866666666666668e-06,
|
| 3005 |
+
"loss": 0.7425,
|
| 3006 |
+
"step": 49300
|
| 3007 |
+
},
|
| 3008 |
+
{
|
| 3009 |
+
"epoch": 40.36,
|
| 3010 |
+
"learning_rate": 1.77e-06,
|
| 3011 |
+
"loss": 0.7435,
|
| 3012 |
+
"step": 49400
|
| 3013 |
+
},
|
| 3014 |
+
{
|
| 3015 |
+
"epoch": 40.44,
|
| 3016 |
+
"learning_rate": 1.7535000000000002e-06,
|
| 3017 |
+
"loss": 0.7487,
|
| 3018 |
+
"step": 49500
|
| 3019 |
+
},
|
| 3020 |
+
{
|
| 3021 |
+
"epoch": 40.52,
|
| 3022 |
+
"learning_rate": 1.7368333333333334e-06,
|
| 3023 |
+
"loss": 0.7614,
|
| 3024 |
+
"step": 49600
|
| 3025 |
+
},
|
| 3026 |
+
{
|
| 3027 |
+
"epoch": 40.6,
|
| 3028 |
+
"learning_rate": 1.7201666666666669e-06,
|
| 3029 |
+
"loss": 0.7309,
|
| 3030 |
+
"step": 49700
|
| 3031 |
+
},
|
| 3032 |
+
{
|
| 3033 |
+
"epoch": 40.69,
|
| 3034 |
+
"learning_rate": 1.7035e-06,
|
| 3035 |
+
"loss": 0.7304,
|
| 3036 |
+
"step": 49800
|
| 3037 |
+
},
|
| 3038 |
+
{
|
| 3039 |
+
"epoch": 40.77,
|
| 3040 |
+
"learning_rate": 1.6868333333333333e-06,
|
| 3041 |
+
"loss": 0.7716,
|
| 3042 |
+
"step": 49900
|
| 3043 |
+
},
|
| 3044 |
+
{
|
| 3045 |
+
"epoch": 40.85,
|
| 3046 |
+
"learning_rate": 1.670166666666667e-06,
|
| 3047 |
+
"loss": 0.7319,
|
| 3048 |
+
"step": 50000
|
| 3049 |
+
},
|
| 3050 |
+
{
|
| 3051 |
+
"epoch": 40.85,
|
| 3052 |
+
"eval_cer": 0.10707086369979842,
|
| 3053 |
+
"eval_loss": 0.4964157044887543,
|
| 3054 |
+
"eval_runtime": 7.9633,
|
| 3055 |
+
"eval_samples_per_second": 37.673,
|
| 3056 |
+
"eval_steps_per_second": 18.836,
|
| 3057 |
+
"eval_wer": 0.3645299145299145,
|
| 3058 |
+
"step": 50000
|
| 3059 |
+
},
|
| 3060 |
+
{
|
| 3061 |
+
"epoch": 40.93,
|
| 3062 |
+
"learning_rate": 1.6535000000000001e-06,
|
| 3063 |
+
"loss": 0.7407,
|
| 3064 |
+
"step": 50100
|
| 3065 |
+
},
|
| 3066 |
+
{
|
| 3067 |
+
"epoch": 41.01,
|
| 3068 |
+
"learning_rate": 1.6368333333333333e-06,
|
| 3069 |
+
"loss": 0.759,
|
| 3070 |
+
"step": 50200
|
| 3071 |
+
},
|
| 3072 |
+
{
|
| 3073 |
+
"epoch": 41.09,
|
| 3074 |
+
"learning_rate": 1.6201666666666668e-06,
|
| 3075 |
+
"loss": 0.7313,
|
| 3076 |
+
"step": 50300
|
| 3077 |
+
},
|
| 3078 |
+
{
|
| 3079 |
+
"epoch": 41.18,
|
| 3080 |
+
"learning_rate": 1.6035e-06,
|
| 3081 |
+
"loss": 0.7404,
|
| 3082 |
+
"step": 50400
|
| 3083 |
+
},
|
| 3084 |
+
{
|
| 3085 |
+
"epoch": 41.26,
|
| 3086 |
+
"learning_rate": 1.5868333333333336e-06,
|
| 3087 |
+
"loss": 0.7065,
|
| 3088 |
+
"step": 50500
|
| 3089 |
+
},
|
| 3090 |
+
{
|
| 3091 |
+
"epoch": 41.34,
|
| 3092 |
+
"learning_rate": 1.5701666666666668e-06,
|
| 3093 |
+
"loss": 0.734,
|
| 3094 |
+
"step": 50600
|
| 3095 |
+
},
|
| 3096 |
+
{
|
| 3097 |
+
"epoch": 41.42,
|
| 3098 |
+
"learning_rate": 1.5535e-06,
|
| 3099 |
+
"loss": 0.7326,
|
| 3100 |
+
"step": 50700
|
| 3101 |
+
},
|
| 3102 |
+
{
|
| 3103 |
+
"epoch": 41.5,
|
| 3104 |
+
"learning_rate": 1.5368333333333335e-06,
|
| 3105 |
+
"loss": 0.7268,
|
| 3106 |
+
"step": 50800
|
| 3107 |
+
},
|
| 3108 |
+
{
|
| 3109 |
+
"epoch": 41.58,
|
| 3110 |
+
"learning_rate": 1.5201666666666667e-06,
|
| 3111 |
+
"loss": 0.7687,
|
| 3112 |
+
"step": 50900
|
| 3113 |
+
},
|
| 3114 |
+
{
|
| 3115 |
+
"epoch": 41.67,
|
| 3116 |
+
"learning_rate": 1.5035000000000003e-06,
|
| 3117 |
+
"loss": 0.7397,
|
| 3118 |
+
"step": 51000
|
| 3119 |
+
},
|
| 3120 |
+
{
|
| 3121 |
+
"epoch": 41.75,
|
| 3122 |
+
"learning_rate": 1.4868333333333335e-06,
|
| 3123 |
+
"loss": 0.7426,
|
| 3124 |
+
"step": 51100
|
| 3125 |
+
},
|
| 3126 |
+
{
|
| 3127 |
+
"epoch": 41.83,
|
| 3128 |
+
"learning_rate": 1.4701666666666667e-06,
|
| 3129 |
+
"loss": 0.7418,
|
| 3130 |
+
"step": 51200
|
| 3131 |
+
},
|
| 3132 |
+
{
|
| 3133 |
+
"epoch": 41.91,
|
| 3134 |
+
"learning_rate": 1.4535000000000001e-06,
|
| 3135 |
+
"loss": 0.7789,
|
| 3136 |
+
"step": 51300
|
| 3137 |
+
},
|
| 3138 |
+
{
|
| 3139 |
+
"epoch": 41.99,
|
| 3140 |
+
"learning_rate": 1.4368333333333334e-06,
|
| 3141 |
+
"loss": 0.7412,
|
| 3142 |
+
"step": 51400
|
| 3143 |
+
},
|
| 3144 |
+
{
|
| 3145 |
+
"epoch": 42.08,
|
| 3146 |
+
"learning_rate": 1.4203333333333336e-06,
|
| 3147 |
+
"loss": 0.7638,
|
| 3148 |
+
"step": 51500
|
| 3149 |
+
},
|
| 3150 |
+
{
|
| 3151 |
+
"epoch": 42.16,
|
| 3152 |
+
"learning_rate": 1.4036666666666668e-06,
|
| 3153 |
+
"loss": 0.7291,
|
| 3154 |
+
"step": 51600
|
| 3155 |
+
},
|
| 3156 |
+
{
|
| 3157 |
+
"epoch": 42.24,
|
| 3158 |
+
"learning_rate": 1.387e-06,
|
| 3159 |
+
"loss": 0.7531,
|
| 3160 |
+
"step": 51700
|
| 3161 |
+
},
|
| 3162 |
+
{
|
| 3163 |
+
"epoch": 42.32,
|
| 3164 |
+
"learning_rate": 1.3703333333333334e-06,
|
| 3165 |
+
"loss": 0.7164,
|
| 3166 |
+
"step": 51800
|
| 3167 |
+
},
|
| 3168 |
+
{
|
| 3169 |
+
"epoch": 42.4,
|
| 3170 |
+
"learning_rate": 1.3536666666666667e-06,
|
| 3171 |
+
"loss": 0.7621,
|
| 3172 |
+
"step": 51900
|
| 3173 |
+
},
|
| 3174 |
+
{
|
| 3175 |
+
"epoch": 42.48,
|
| 3176 |
+
"learning_rate": 1.3370000000000003e-06,
|
| 3177 |
+
"loss": 0.7719,
|
| 3178 |
+
"step": 52000
|
| 3179 |
+
},
|
| 3180 |
+
{
|
| 3181 |
+
"epoch": 42.57,
|
| 3182 |
+
"learning_rate": 1.3203333333333335e-06,
|
| 3183 |
+
"loss": 0.7192,
|
| 3184 |
+
"step": 52100
|
| 3185 |
+
},
|
| 3186 |
+
{
|
| 3187 |
+
"epoch": 42.65,
|
| 3188 |
+
"learning_rate": 1.3036666666666667e-06,
|
| 3189 |
+
"loss": 0.73,
|
| 3190 |
+
"step": 52200
|
| 3191 |
+
},
|
| 3192 |
+
{
|
| 3193 |
+
"epoch": 42.73,
|
| 3194 |
+
"learning_rate": 1.2870000000000001e-06,
|
| 3195 |
+
"loss": 0.756,
|
| 3196 |
+
"step": 52300
|
| 3197 |
+
},
|
| 3198 |
+
{
|
| 3199 |
+
"epoch": 42.81,
|
| 3200 |
+
"learning_rate": 1.2703333333333333e-06,
|
| 3201 |
+
"loss": 0.7576,
|
| 3202 |
+
"step": 52400
|
| 3203 |
+
},
|
| 3204 |
+
{
|
| 3205 |
+
"epoch": 42.89,
|
| 3206 |
+
"learning_rate": 1.2536666666666666e-06,
|
| 3207 |
+
"loss": 0.7161,
|
| 3208 |
+
"step": 52500
|
| 3209 |
+
},
|
| 3210 |
+
{
|
| 3211 |
+
"epoch": 42.97,
|
| 3212 |
+
"learning_rate": 1.2370000000000002e-06,
|
| 3213 |
+
"loss": 0.7123,
|
| 3214 |
+
"step": 52600
|
| 3215 |
+
},
|
| 3216 |
+
{
|
| 3217 |
+
"epoch": 43.06,
|
| 3218 |
+
"learning_rate": 1.2203333333333334e-06,
|
| 3219 |
+
"loss": 0.7612,
|
| 3220 |
+
"step": 52700
|
| 3221 |
+
},
|
| 3222 |
+
{
|
| 3223 |
+
"epoch": 43.14,
|
| 3224 |
+
"learning_rate": 1.2036666666666668e-06,
|
| 3225 |
+
"loss": 0.7216,
|
| 3226 |
+
"step": 52800
|
| 3227 |
+
},
|
| 3228 |
+
{
|
| 3229 |
+
"epoch": 43.22,
|
| 3230 |
+
"learning_rate": 1.187e-06,
|
| 3231 |
+
"loss": 0.7363,
|
| 3232 |
+
"step": 52900
|
| 3233 |
+
},
|
| 3234 |
+
{
|
| 3235 |
+
"epoch": 43.3,
|
| 3236 |
+
"learning_rate": 1.1705e-06,
|
| 3237 |
+
"loss": 0.7175,
|
| 3238 |
+
"step": 53000
|
| 3239 |
+
},
|
| 3240 |
+
{
|
| 3241 |
+
"epoch": 43.38,
|
| 3242 |
+
"learning_rate": 1.1538333333333335e-06,
|
| 3243 |
+
"loss": 0.7358,
|
| 3244 |
+
"step": 53100
|
| 3245 |
+
},
|
| 3246 |
+
{
|
| 3247 |
+
"epoch": 43.46,
|
| 3248 |
+
"learning_rate": 1.1371666666666667e-06,
|
| 3249 |
+
"loss": 0.7449,
|
| 3250 |
+
"step": 53200
|
| 3251 |
+
},
|
| 3252 |
+
{
|
| 3253 |
+
"epoch": 43.55,
|
| 3254 |
+
"learning_rate": 1.1205000000000001e-06,
|
| 3255 |
+
"loss": 0.7414,
|
| 3256 |
+
"step": 53300
|
| 3257 |
+
},
|
| 3258 |
+
{
|
| 3259 |
+
"epoch": 43.63,
|
| 3260 |
+
"learning_rate": 1.1038333333333333e-06,
|
| 3261 |
+
"loss": 0.7333,
|
| 3262 |
+
"step": 53400
|
| 3263 |
+
},
|
| 3264 |
+
{
|
| 3265 |
+
"epoch": 43.71,
|
| 3266 |
+
"learning_rate": 1.0871666666666667e-06,
|
| 3267 |
+
"loss": 0.7042,
|
| 3268 |
+
"step": 53500
|
| 3269 |
+
},
|
| 3270 |
+
{
|
| 3271 |
+
"epoch": 43.79,
|
| 3272 |
+
"learning_rate": 1.0705000000000002e-06,
|
| 3273 |
+
"loss": 0.7239,
|
| 3274 |
+
"step": 53600
|
| 3275 |
+
},
|
| 3276 |
+
{
|
| 3277 |
+
"epoch": 43.87,
|
| 3278 |
+
"learning_rate": 1.0538333333333334e-06,
|
| 3279 |
+
"loss": 0.7493,
|
| 3280 |
+
"step": 53700
|
| 3281 |
+
},
|
| 3282 |
+
{
|
| 3283 |
+
"epoch": 43.95,
|
| 3284 |
+
"learning_rate": 1.0371666666666668e-06,
|
| 3285 |
+
"loss": 0.7588,
|
| 3286 |
+
"step": 53800
|
| 3287 |
+
},
|
| 3288 |
+
{
|
| 3289 |
+
"epoch": 44.04,
|
| 3290 |
+
"learning_rate": 1.0205e-06,
|
| 3291 |
+
"loss": 0.7311,
|
| 3292 |
+
"step": 53900
|
| 3293 |
+
},
|
| 3294 |
+
{
|
| 3295 |
+
"epoch": 44.12,
|
| 3296 |
+
"learning_rate": 1.0038333333333334e-06,
|
| 3297 |
+
"loss": 0.7177,
|
| 3298 |
+
"step": 54000
|
| 3299 |
+
},
|
| 3300 |
+
{
|
| 3301 |
+
"epoch": 44.2,
|
| 3302 |
+
"learning_rate": 9.871666666666666e-07,
|
| 3303 |
+
"loss": 0.7187,
|
| 3304 |
+
"step": 54100
|
| 3305 |
+
},
|
| 3306 |
+
{
|
| 3307 |
+
"epoch": 44.28,
|
| 3308 |
+
"learning_rate": 9.705e-07,
|
| 3309 |
+
"loss": 0.7348,
|
| 3310 |
+
"step": 54200
|
| 3311 |
+
},
|
| 3312 |
+
{
|
| 3313 |
+
"epoch": 44.36,
|
| 3314 |
+
"learning_rate": 9.538333333333335e-07,
|
| 3315 |
+
"loss": 0.7477,
|
| 3316 |
+
"step": 54300
|
| 3317 |
+
},
|
| 3318 |
+
{
|
| 3319 |
+
"epoch": 44.44,
|
| 3320 |
+
"learning_rate": 9.371666666666668e-07,
|
| 3321 |
+
"loss": 0.7348,
|
| 3322 |
+
"step": 54400
|
| 3323 |
+
},
|
| 3324 |
+
{
|
| 3325 |
+
"epoch": 44.53,
|
| 3326 |
+
"learning_rate": 9.205000000000001e-07,
|
| 3327 |
+
"loss": 0.7133,
|
| 3328 |
+
"step": 54500
|
| 3329 |
+
},
|
| 3330 |
+
{
|
| 3331 |
+
"epoch": 44.61,
|
| 3332 |
+
"learning_rate": 9.038333333333333e-07,
|
| 3333 |
+
"loss": 0.7707,
|
| 3334 |
+
"step": 54600
|
| 3335 |
+
},
|
| 3336 |
+
{
|
| 3337 |
+
"epoch": 44.69,
|
| 3338 |
+
"learning_rate": 8.871666666666668e-07,
|
| 3339 |
+
"loss": 0.7422,
|
| 3340 |
+
"step": 54700
|
| 3341 |
+
},
|
| 3342 |
+
{
|
| 3343 |
+
"epoch": 44.77,
|
| 3344 |
+
"learning_rate": 8.705000000000001e-07,
|
| 3345 |
+
"loss": 0.7397,
|
| 3346 |
+
"step": 54800
|
| 3347 |
+
},
|
| 3348 |
+
{
|
| 3349 |
+
"epoch": 44.85,
|
| 3350 |
+
"learning_rate": 8.538333333333335e-07,
|
| 3351 |
+
"loss": 0.7233,
|
| 3352 |
+
"step": 54900
|
| 3353 |
+
},
|
| 3354 |
+
{
|
| 3355 |
+
"epoch": 44.93,
|
| 3356 |
+
"learning_rate": 8.371666666666667e-07,
|
| 3357 |
+
"loss": 0.7672,
|
| 3358 |
+
"step": 55000
|
| 3359 |
+
},
|
| 3360 |
+
{
|
| 3361 |
+
"epoch": 45.02,
|
| 3362 |
+
"learning_rate": 8.205e-07,
|
| 3363 |
+
"loss": 0.7235,
|
| 3364 |
+
"step": 55100
|
| 3365 |
+
},
|
| 3366 |
+
{
|
| 3367 |
+
"epoch": 45.1,
|
| 3368 |
+
"learning_rate": 8.038333333333334e-07,
|
| 3369 |
+
"loss": 0.7442,
|
| 3370 |
+
"step": 55200
|
| 3371 |
+
},
|
| 3372 |
+
{
|
| 3373 |
+
"epoch": 45.18,
|
| 3374 |
+
"learning_rate": 7.871666666666668e-07,
|
| 3375 |
+
"loss": 0.7244,
|
| 3376 |
+
"step": 55300
|
| 3377 |
+
},
|
| 3378 |
+
{
|
| 3379 |
+
"epoch": 45.26,
|
| 3380 |
+
"learning_rate": 7.705e-07,
|
| 3381 |
+
"loss": 0.7272,
|
| 3382 |
+
"step": 55400
|
| 3383 |
+
},
|
| 3384 |
+
{
|
| 3385 |
+
"epoch": 45.34,
|
| 3386 |
+
"learning_rate": 7.538333333333334e-07,
|
| 3387 |
+
"loss": 0.7283,
|
| 3388 |
+
"step": 55500
|
| 3389 |
+
},
|
| 3390 |
+
{
|
| 3391 |
+
"epoch": 45.42,
|
| 3392 |
+
"learning_rate": 7.371666666666667e-07,
|
| 3393 |
+
"loss": 0.7126,
|
| 3394 |
+
"step": 55600
|
| 3395 |
+
},
|
| 3396 |
+
{
|
| 3397 |
+
"epoch": 45.51,
|
| 3398 |
+
"learning_rate": 7.205000000000001e-07,
|
| 3399 |
+
"loss": 0.7342,
|
| 3400 |
+
"step": 55700
|
| 3401 |
+
},
|
| 3402 |
+
{
|
| 3403 |
+
"epoch": 45.59,
|
| 3404 |
+
"learning_rate": 7.038333333333334e-07,
|
| 3405 |
+
"loss": 0.7334,
|
| 3406 |
+
"step": 55800
|
| 3407 |
+
},
|
| 3408 |
+
{
|
| 3409 |
+
"epoch": 45.67,
|
| 3410 |
+
"learning_rate": 6.871666666666667e-07,
|
| 3411 |
+
"loss": 0.7709,
|
| 3412 |
+
"step": 55900
|
| 3413 |
+
},
|
| 3414 |
+
{
|
| 3415 |
+
"epoch": 45.75,
|
| 3416 |
+
"learning_rate": 6.705000000000001e-07,
|
| 3417 |
+
"loss": 0.7267,
|
| 3418 |
+
"step": 56000
|
| 3419 |
+
},
|
| 3420 |
+
{
|
| 3421 |
+
"epoch": 45.83,
|
| 3422 |
+
"learning_rate": 6.538333333333334e-07,
|
| 3423 |
+
"loss": 0.7248,
|
| 3424 |
+
"step": 56100
|
| 3425 |
+
},
|
| 3426 |
+
{
|
| 3427 |
+
"epoch": 45.91,
|
| 3428 |
+
"learning_rate": 6.371666666666667e-07,
|
| 3429 |
+
"loss": 0.7289,
|
| 3430 |
+
"step": 56200
|
| 3431 |
+
},
|
| 3432 |
+
{
|
| 3433 |
+
"epoch": 46.0,
|
| 3434 |
+
"learning_rate": 6.205e-07,
|
| 3435 |
+
"loss": 0.7192,
|
| 3436 |
+
"step": 56300
|
| 3437 |
+
},
|
| 3438 |
+
{
|
| 3439 |
+
"epoch": 46.08,
|
| 3440 |
+
"learning_rate": 6.038333333333333e-07,
|
| 3441 |
+
"loss": 0.7347,
|
| 3442 |
+
"step": 56400
|
| 3443 |
+
},
|
| 3444 |
+
{
|
| 3445 |
+
"epoch": 46.16,
|
| 3446 |
+
"learning_rate": 5.871666666666668e-07,
|
| 3447 |
+
"loss": 0.7493,
|
| 3448 |
+
"step": 56500
|
| 3449 |
+
},
|
| 3450 |
+
{
|
| 3451 |
+
"epoch": 46.24,
|
| 3452 |
+
"learning_rate": 5.705e-07,
|
| 3453 |
+
"loss": 0.7034,
|
| 3454 |
+
"step": 56600
|
| 3455 |
+
},
|
| 3456 |
+
{
|
| 3457 |
+
"epoch": 46.32,
|
| 3458 |
+
"learning_rate": 5.538333333333334e-07,
|
| 3459 |
+
"loss": 0.7668,
|
| 3460 |
+
"step": 56700
|
| 3461 |
+
},
|
| 3462 |
+
{
|
| 3463 |
+
"epoch": 46.4,
|
| 3464 |
+
"learning_rate": 5.371666666666667e-07,
|
| 3465 |
+
"loss": 0.7516,
|
| 3466 |
+
"step": 56800
|
| 3467 |
+
},
|
| 3468 |
+
{
|
| 3469 |
+
"epoch": 46.49,
|
| 3470 |
+
"learning_rate": 5.205e-07,
|
| 3471 |
+
"loss": 0.7294,
|
| 3472 |
+
"step": 56900
|
| 3473 |
+
},
|
| 3474 |
+
{
|
| 3475 |
+
"epoch": 46.57,
|
| 3476 |
+
"learning_rate": 5.038333333333334e-07,
|
| 3477 |
+
"loss": 0.7343,
|
| 3478 |
+
"step": 57000
|
| 3479 |
+
},
|
| 3480 |
+
{
|
| 3481 |
+
"epoch": 46.65,
|
| 3482 |
+
"learning_rate": 4.871666666666667e-07,
|
| 3483 |
+
"loss": 0.6948,
|
| 3484 |
+
"step": 57100
|
| 3485 |
+
},
|
| 3486 |
+
{
|
| 3487 |
+
"epoch": 46.73,
|
| 3488 |
+
"learning_rate": 4.7050000000000004e-07,
|
| 3489 |
+
"loss": 0.7239,
|
| 3490 |
+
"step": 57200
|
| 3491 |
+
},
|
| 3492 |
+
{
|
| 3493 |
+
"epoch": 46.81,
|
| 3494 |
+
"learning_rate": 4.5383333333333335e-07,
|
| 3495 |
+
"loss": 0.7533,
|
| 3496 |
+
"step": 57300
|
| 3497 |
+
},
|
| 3498 |
+
{
|
| 3499 |
+
"epoch": 46.89,
|
| 3500 |
+
"learning_rate": 4.371666666666667e-07,
|
| 3501 |
+
"loss": 0.7433,
|
| 3502 |
+
"step": 57400
|
| 3503 |
+
},
|
| 3504 |
+
{
|
| 3505 |
+
"epoch": 46.98,
|
| 3506 |
+
"learning_rate": 4.205e-07,
|
| 3507 |
+
"loss": 0.737,
|
| 3508 |
+
"step": 57500
|
| 3509 |
+
},
|
| 3510 |
+
{
|
| 3511 |
+
"epoch": 47.06,
|
| 3512 |
+
"learning_rate": 4.0383333333333336e-07,
|
| 3513 |
+
"loss": 0.7363,
|
| 3514 |
+
"step": 57600
|
| 3515 |
+
},
|
| 3516 |
+
{
|
| 3517 |
+
"epoch": 47.14,
|
| 3518 |
+
"learning_rate": 3.871666666666667e-07,
|
| 3519 |
+
"loss": 0.7304,
|
| 3520 |
+
"step": 57700
|
| 3521 |
+
},
|
| 3522 |
+
{
|
| 3523 |
+
"epoch": 47.22,
|
| 3524 |
+
"learning_rate": 3.7050000000000004e-07,
|
| 3525 |
+
"loss": 0.7017,
|
| 3526 |
+
"step": 57800
|
| 3527 |
+
},
|
| 3528 |
+
{
|
| 3529 |
+
"epoch": 47.3,
|
| 3530 |
+
"learning_rate": 3.538333333333334e-07,
|
| 3531 |
+
"loss": 0.7194,
|
| 3532 |
+
"step": 57900
|
| 3533 |
+
},
|
| 3534 |
+
{
|
| 3535 |
+
"epoch": 47.39,
|
| 3536 |
+
"learning_rate": 3.371666666666667e-07,
|
| 3537 |
+
"loss": 0.7061,
|
| 3538 |
+
"step": 58000
|
| 3539 |
+
},
|
| 3540 |
+
{
|
| 3541 |
+
"epoch": 47.47,
|
| 3542 |
+
"learning_rate": 3.2050000000000004e-07,
|
| 3543 |
+
"loss": 0.7565,
|
| 3544 |
+
"step": 58100
|
| 3545 |
+
},
|
| 3546 |
+
{
|
| 3547 |
+
"epoch": 47.55,
|
| 3548 |
+
"learning_rate": 3.04e-07,
|
| 3549 |
+
"loss": 0.7279,
|
| 3550 |
+
"step": 58200
|
| 3551 |
+
},
|
| 3552 |
+
{
|
| 3553 |
+
"epoch": 47.63,
|
| 3554 |
+
"learning_rate": 2.8733333333333334e-07,
|
| 3555 |
+
"loss": 0.748,
|
| 3556 |
+
"step": 58300
|
| 3557 |
+
},
|
| 3558 |
+
{
|
| 3559 |
+
"epoch": 47.71,
|
| 3560 |
+
"learning_rate": 2.706666666666667e-07,
|
| 3561 |
+
"loss": 0.7418,
|
| 3562 |
+
"step": 58400
|
| 3563 |
+
},
|
| 3564 |
+
{
|
| 3565 |
+
"epoch": 47.79,
|
| 3566 |
+
"learning_rate": 2.54e-07,
|
| 3567 |
+
"loss": 0.7147,
|
| 3568 |
+
"step": 58500
|
| 3569 |
+
},
|
| 3570 |
+
{
|
| 3571 |
+
"epoch": 47.88,
|
| 3572 |
+
"learning_rate": 2.3733333333333334e-07,
|
| 3573 |
+
"loss": 0.7313,
|
| 3574 |
+
"step": 58600
|
| 3575 |
+
},
|
| 3576 |
+
{
|
| 3577 |
+
"epoch": 47.96,
|
| 3578 |
+
"learning_rate": 2.2066666666666666e-07,
|
| 3579 |
+
"loss": 0.7291,
|
| 3580 |
+
"step": 58700
|
| 3581 |
+
},
|
| 3582 |
+
{
|
| 3583 |
+
"epoch": 48.04,
|
| 3584 |
+
"learning_rate": 2.0400000000000003e-07,
|
| 3585 |
+
"loss": 0.7145,
|
| 3586 |
+
"step": 58800
|
| 3587 |
+
},
|
| 3588 |
+
{
|
| 3589 |
+
"epoch": 48.12,
|
| 3590 |
+
"learning_rate": 1.8733333333333337e-07,
|
| 3591 |
+
"loss": 0.6993,
|
| 3592 |
+
"step": 58900
|
| 3593 |
+
},
|
| 3594 |
+
{
|
| 3595 |
+
"epoch": 48.2,
|
| 3596 |
+
"learning_rate": 1.7066666666666669e-07,
|
| 3597 |
+
"loss": 0.706,
|
| 3598 |
+
"step": 59000
|
| 3599 |
+
},
|
| 3600 |
+
{
|
| 3601 |
+
"epoch": 48.28,
|
| 3602 |
+
"learning_rate": 1.5400000000000003e-07,
|
| 3603 |
+
"loss": 0.7443,
|
| 3604 |
+
"step": 59100
|
| 3605 |
+
},
|
| 3606 |
+
{
|
| 3607 |
+
"epoch": 48.37,
|
| 3608 |
+
"learning_rate": 1.3733333333333335e-07,
|
| 3609 |
+
"loss": 0.7421,
|
| 3610 |
+
"step": 59200
|
| 3611 |
+
},
|
| 3612 |
+
{
|
| 3613 |
+
"epoch": 48.45,
|
| 3614 |
+
"learning_rate": 1.2066666666666666e-07,
|
| 3615 |
+
"loss": 0.7562,
|
| 3616 |
+
"step": 59300
|
| 3617 |
+
},
|
| 3618 |
+
{
|
| 3619 |
+
"epoch": 48.53,
|
| 3620 |
+
"learning_rate": 1.04e-07,
|
| 3621 |
+
"loss": 0.7263,
|
| 3622 |
+
"step": 59400
|
| 3623 |
+
},
|
| 3624 |
+
{
|
| 3625 |
+
"epoch": 48.61,
|
| 3626 |
+
"learning_rate": 8.733333333333333e-08,
|
| 3627 |
+
"loss": 0.7286,
|
| 3628 |
+
"step": 59500
|
| 3629 |
+
},
|
| 3630 |
+
{
|
| 3631 |
+
"epoch": 48.69,
|
| 3632 |
+
"learning_rate": 7.066666666666666e-08,
|
| 3633 |
+
"loss": 0.7011,
|
| 3634 |
+
"step": 59600
|
| 3635 |
+
},
|
| 3636 |
+
{
|
| 3637 |
+
"epoch": 48.77,
|
| 3638 |
+
"learning_rate": 5.400000000000001e-08,
|
| 3639 |
+
"loss": 0.7223,
|
| 3640 |
+
"step": 59700
|
| 3641 |
+
},
|
| 3642 |
+
{
|
| 3643 |
+
"epoch": 48.86,
|
| 3644 |
+
"learning_rate": 3.733333333333334e-08,
|
| 3645 |
+
"loss": 0.6973,
|
| 3646 |
+
"step": 59800
|
| 3647 |
+
},
|
| 3648 |
+
{
|
| 3649 |
+
"epoch": 48.94,
|
| 3650 |
+
"learning_rate": 2.066666666666667e-08,
|
| 3651 |
+
"loss": 0.7321,
|
| 3652 |
+
"step": 59900
|
| 3653 |
+
},
|
| 3654 |
+
{
|
| 3655 |
+
"epoch": 49.02,
|
| 3656 |
+
"learning_rate": 4e-09,
|
| 3657 |
+
"loss": 0.7392,
|
| 3658 |
+
"step": 60000
|
| 3659 |
+
},
|
| 3660 |
+
{
|
| 3661 |
+
"epoch": 49.02,
|
| 3662 |
+
"eval_cer": 0.10614048689719337,
|
| 3663 |
+
"eval_loss": 0.49394938349723816,
|
| 3664 |
+
"eval_runtime": 8.0445,
|
| 3665 |
+
"eval_samples_per_second": 37.292,
|
| 3666 |
+
"eval_steps_per_second": 18.646,
|
| 3667 |
+
"eval_wer": 0.35683760683760685,
|
| 3668 |
+
"step": 60000
|
| 3669 |
+
}
|
| 3670 |
+
],
|
| 3671 |
+
"max_steps": 60000,
|
| 3672 |
+
"num_train_epochs": 50,
|
| 3673 |
+
"total_flos": 6.280569561121633e+19,
|
| 3674 |
+
"trial_name": null,
|
| 3675 |
+
"trial_params": null
|
| 3676 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90b6c5e0527d5c0d75920fda5e62f89121513164ca3ffff10ff4252e62fec82c
|
| 3 |
+
size 3451
|