kdcyberdude commited on
Commit
446a228
·
verified ·
1 Parent(s): d16b08f

Training in progress, step 10000, checkpoint

Browse files
checkpoint-10000/config.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "kdcyberdude/w2v-multilingual-v1.0",
3
+ "activation_dropout": 0.0,
4
+ "adapter_act": "relu",
5
+ "adapter_kernel_size": 3,
6
+ "adapter_stride": 2,
7
+ "add_adapter": true,
8
+ "apply_spec_augment": true,
9
+ "architectures": [
10
+ "Wav2Vec2BertForCTC"
11
+ ],
12
+ "attention_dropout": 0.0,
13
+ "bos_token_id": 1,
14
+ "classifier_proj_size": 768,
15
+ "codevector_dim": 768,
16
+ "conformer_conv_dropout": 0.1,
17
+ "contrastive_logits_temperature": 0.1,
18
+ "conv_depthwise_kernel_size": 31,
19
+ "ctc_loss_reduction": "mean",
20
+ "ctc_zero_infinity": true,
21
+ "diversity_loss_weight": 0.1,
22
+ "eos_token_id": 2,
23
+ "feat_proj_dropout": 0.0,
24
+ "feat_quantizer_dropout": 0.0,
25
+ "feature_projection_input_dim": 160,
26
+ "final_dropout": 0.1,
27
+ "hidden_act": "swish",
28
+ "hidden_dropout": 0.0,
29
+ "hidden_size": 1024,
30
+ "initializer_range": 0.02,
31
+ "intermediate_size": 4096,
32
+ "layer_norm_eps": 1e-05,
33
+ "layerdrop": 0.0,
34
+ "left_max_position_embeddings": 64,
35
+ "mask_feature_length": 10,
36
+ "mask_feature_min_masks": 0,
37
+ "mask_feature_prob": 0.0,
38
+ "mask_time_length": 10,
39
+ "mask_time_min_masks": 2,
40
+ "mask_time_prob": 0.0,
41
+ "max_source_positions": 5000,
42
+ "model_type": "wav2vec2-bert",
43
+ "num_adapter_layers": 1,
44
+ "num_attention_heads": 16,
45
+ "num_codevector_groups": 2,
46
+ "num_codevectors_per_group": 320,
47
+ "num_hidden_layers": 24,
48
+ "num_negatives": 100,
49
+ "output_hidden_size": 1024,
50
+ "pad_token_id": 221,
51
+ "position_embeddings_type": "relative_key",
52
+ "proj_codevector_dim": 768,
53
+ "right_max_position_embeddings": 8,
54
+ "rotary_embedding_base": 10000,
55
+ "tdnn_dilation": [
56
+ 1,
57
+ 2,
58
+ 3,
59
+ 1,
60
+ 1
61
+ ],
62
+ "tdnn_dim": [
63
+ 512,
64
+ 512,
65
+ 512,
66
+ 512,
67
+ 1500
68
+ ],
69
+ "tdnn_kernel": [
70
+ 5,
71
+ 3,
72
+ 3,
73
+ 1,
74
+ 1
75
+ ],
76
+ "torch_dtype": "float16",
77
+ "transformers_version": "4.46.2",
78
+ "use_intermediate_ffn_before_adapter": true,
79
+ "use_weighted_layer_sum": false,
80
+ "vocab_size": 224,
81
+ "xvector_output_dim": 512
82
+ }
checkpoint-10000/global_step10000/mp_rank_00_model_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:895bd1ba1ff9a4e0535aa9db4cf5fe3b9cae2f1a60320dc21d66c7fd8db64962
3
+ size 1228864248
checkpoint-10000/global_step10000/zero_pp_rank_0_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:448603f6741fa759e7ca0148512b387afc49e440b4fcfc8ea85c4f1e331f5d22
3
+ size 3685848510
checkpoint-10000/global_step10000/zero_pp_rank_1_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4391249779f7b9b8ba628fa74a8962227e3b29bc4e33b93bf5cc226a06d0c31d
3
+ size 3685847294
checkpoint-10000/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step10000
checkpoint-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27035cf27b7b21074f7039ae2e82c1e319b2dad8490797cad5f458382fa893e9
3
+ size 1228706240
checkpoint-10000/preprocessor_config.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "feature_extractor_type": "SeamlessM4TFeatureExtractor",
3
+ "feature_size": 80,
4
+ "num_mel_bins": 80,
5
+ "padding_side": "right",
6
+ "padding_value": 1,
7
+ "processor_class": "M4TProcessorWithLM",
8
+ "return_attention_mask": true,
9
+ "sampling_rate": 16000,
10
+ "stride": 2
11
+ }
checkpoint-10000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5aa69eba430a15caaabae2b48ca23b3b5bf187f563f1a16591a477ac089eaaf
3
+ size 14512
checkpoint-10000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11dc824f413404a63c1c52ae38e1c0472cc8de3540622d6846a6fea64ff65c70
3
+ size 14512
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e5c16b72cf537e95729bd2e2d8b16fc77549fb66bc26281be375bbd90fc8c49
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
@@ -0,0 +1,3578 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.23184117484036582,
3
+ "best_model_checkpoint": "./checkpoints/w2v-multilingual-v1.3/checkpoint-10000",
4
+ "epoch": 0.8238137082601055,
5
+ "eval_steps": 2000,
6
+ "global_step": 10000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.001647627416520211,
13
+ "grad_norm": 0.2767059803009033,
14
+ "learning_rate": 4.9431537320810675e-08,
15
+ "loss": 0.1923,
16
+ "step": 20
17
+ },
18
+ {
19
+ "epoch": 0.003295254833040422,
20
+ "grad_norm": 2.006911039352417,
21
+ "learning_rate": 1.977261492832427e-07,
22
+ "loss": 0.2947,
23
+ "step": 40
24
+ },
25
+ {
26
+ "epoch": 0.004942882249560633,
27
+ "grad_norm": 0.3714193105697632,
28
+ "learning_rate": 3.624979403526116e-07,
29
+ "loss": 0.2955,
30
+ "step": 60
31
+ },
32
+ {
33
+ "epoch": 0.006590509666080844,
34
+ "grad_norm": 0.3106062412261963,
35
+ "learning_rate": 5.272697314219806e-07,
36
+ "loss": 0.2754,
37
+ "step": 80
38
+ },
39
+ {
40
+ "epoch": 0.008238137082601054,
41
+ "grad_norm": 3.2078254222869873,
42
+ "learning_rate": 6.838029329378811e-07,
43
+ "loss": 0.3177,
44
+ "step": 100
45
+ },
46
+ {
47
+ "epoch": 0.009885764499121265,
48
+ "grad_norm": 0.38503074645996094,
49
+ "learning_rate": 8.485747240072501e-07,
50
+ "loss": 0.1905,
51
+ "step": 120
52
+ },
53
+ {
54
+ "epoch": 0.011533391915641476,
55
+ "grad_norm": 0.15724419057369232,
56
+ "learning_rate": 1.013346515076619e-06,
57
+ "loss": 0.2942,
58
+ "step": 140
59
+ },
60
+ {
61
+ "epoch": 0.013181019332161687,
62
+ "grad_norm": 0.5840986967086792,
63
+ "learning_rate": 1.1781183061459877e-06,
64
+ "loss": 0.2907,
65
+ "step": 160
66
+ },
67
+ {
68
+ "epoch": 0.014828646748681899,
69
+ "grad_norm": 0.23122897744178772,
70
+ "learning_rate": 1.3346515076618883e-06,
71
+ "loss": 0.2811,
72
+ "step": 180
73
+ },
74
+ {
75
+ "epoch": 0.016476274165202108,
76
+ "grad_norm": 2.7248637676239014,
77
+ "learning_rate": 1.4994232987312573e-06,
78
+ "loss": 0.3168,
79
+ "step": 200
80
+ },
81
+ {
82
+ "epoch": 0.01812390158172232,
83
+ "grad_norm": 1.0846350193023682,
84
+ "learning_rate": 1.6641950898006263e-06,
85
+ "loss": 0.1979,
86
+ "step": 220
87
+ },
88
+ {
89
+ "epoch": 0.01977152899824253,
90
+ "grad_norm": 1.7905633449554443,
91
+ "learning_rate": 1.828966880869995e-06,
92
+ "loss": 0.2656,
93
+ "step": 240
94
+ },
95
+ {
96
+ "epoch": 0.021419156414762743,
97
+ "grad_norm": 0.31305554509162903,
98
+ "learning_rate": 1.993738671939364e-06,
99
+ "loss": 0.272,
100
+ "step": 260
101
+ },
102
+ {
103
+ "epoch": 0.023066783831282953,
104
+ "grad_norm": 0.31258106231689453,
105
+ "learning_rate": 2.158510463008733e-06,
106
+ "loss": 0.2693,
107
+ "step": 280
108
+ },
109
+ {
110
+ "epoch": 0.024714411247803162,
111
+ "grad_norm": 1.783349871635437,
112
+ "learning_rate": 2.323282254078102e-06,
113
+ "loss": 0.3116,
114
+ "step": 300
115
+ },
116
+ {
117
+ "epoch": 0.026362038664323375,
118
+ "grad_norm": 0.6936488747596741,
119
+ "learning_rate": 2.488054045147471e-06,
120
+ "loss": 0.1931,
121
+ "step": 320
122
+ },
123
+ {
124
+ "epoch": 0.028009666080843584,
125
+ "grad_norm": 0.6185577511787415,
126
+ "learning_rate": 2.65282583621684e-06,
127
+ "loss": 0.258,
128
+ "step": 340
129
+ },
130
+ {
131
+ "epoch": 0.029657293497363797,
132
+ "grad_norm": 0.3592207729816437,
133
+ "learning_rate": 2.8175976272862085e-06,
134
+ "loss": 0.2401,
135
+ "step": 360
136
+ },
137
+ {
138
+ "epoch": 0.03130492091388401,
139
+ "grad_norm": 1.2324920892715454,
140
+ "learning_rate": 2.982369418355578e-06,
141
+ "loss": 0.2371,
142
+ "step": 380
143
+ },
144
+ {
145
+ "epoch": 0.032952548330404216,
146
+ "grad_norm": 1.5560214519500732,
147
+ "learning_rate": 3.147141209424947e-06,
148
+ "loss": 0.2414,
149
+ "step": 400
150
+ },
151
+ {
152
+ "epoch": 0.03460017574692443,
153
+ "grad_norm": 0.8133947253227234,
154
+ "learning_rate": 3.3119130004943154e-06,
155
+ "loss": 0.1744,
156
+ "step": 420
157
+ },
158
+ {
159
+ "epoch": 0.03624780316344464,
160
+ "grad_norm": 0.8599107265472412,
161
+ "learning_rate": 3.4766847915636844e-06,
162
+ "loss": 0.2536,
163
+ "step": 440
164
+ },
165
+ {
166
+ "epoch": 0.03789543057996485,
167
+ "grad_norm": 0.3699595034122467,
168
+ "learning_rate": 3.641456582633054e-06,
169
+ "loss": 0.2552,
170
+ "step": 460
171
+ },
172
+ {
173
+ "epoch": 0.03954305799648506,
174
+ "grad_norm": 1.2955116033554077,
175
+ "learning_rate": 3.797989784148954e-06,
176
+ "loss": 0.2532,
177
+ "step": 480
178
+ },
179
+ {
180
+ "epoch": 0.041190685413005274,
181
+ "grad_norm": 11.518170356750488,
182
+ "learning_rate": 3.962761575218322e-06,
183
+ "loss": 0.2507,
184
+ "step": 500
185
+ },
186
+ {
187
+ "epoch": 0.042838312829525486,
188
+ "grad_norm": 0.9779248833656311,
189
+ "learning_rate": 4.127533366287692e-06,
190
+ "loss": 0.1733,
191
+ "step": 520
192
+ },
193
+ {
194
+ "epoch": 0.04448594024604569,
195
+ "grad_norm": 0.5386682152748108,
196
+ "learning_rate": 4.29230515735706e-06,
197
+ "loss": 0.2415,
198
+ "step": 540
199
+ },
200
+ {
201
+ "epoch": 0.046133567662565905,
202
+ "grad_norm": 2.7366793155670166,
203
+ "learning_rate": 4.457076948426429e-06,
204
+ "loss": 0.2409,
205
+ "step": 560
206
+ },
207
+ {
208
+ "epoch": 0.04778119507908612,
209
+ "grad_norm": 1.9912066459655762,
210
+ "learning_rate": 4.621848739495799e-06,
211
+ "loss": 0.2134,
212
+ "step": 580
213
+ },
214
+ {
215
+ "epoch": 0.049428822495606324,
216
+ "grad_norm": 4.058573246002197,
217
+ "learning_rate": 4.786620530565167e-06,
218
+ "loss": 0.2456,
219
+ "step": 600
220
+ },
221
+ {
222
+ "epoch": 0.05107644991212654,
223
+ "grad_norm": 0.37456750869750977,
224
+ "learning_rate": 4.951392321634536e-06,
225
+ "loss": 0.1589,
226
+ "step": 620
227
+ },
228
+ {
229
+ "epoch": 0.05272407732864675,
230
+ "grad_norm": 0.7950440645217896,
231
+ "learning_rate": 5.116164112703905e-06,
232
+ "loss": 0.2485,
233
+ "step": 640
234
+ },
235
+ {
236
+ "epoch": 0.05437170474516696,
237
+ "grad_norm": 1.884665846824646,
238
+ "learning_rate": 5.280935903773274e-06,
239
+ "loss": 0.2197,
240
+ "step": 660
241
+ },
242
+ {
243
+ "epoch": 0.05601933216168717,
244
+ "grad_norm": 0.32170844078063965,
245
+ "learning_rate": 5.445707694842643e-06,
246
+ "loss": 0.2111,
247
+ "step": 680
248
+ },
249
+ {
250
+ "epoch": 0.05766695957820738,
251
+ "grad_norm": 1.8230172395706177,
252
+ "learning_rate": 5.610479485912012e-06,
253
+ "loss": 0.2379,
254
+ "step": 700
255
+ },
256
+ {
257
+ "epoch": 0.059314586994727594,
258
+ "grad_norm": 1.2472524642944336,
259
+ "learning_rate": 5.775251276981381e-06,
260
+ "loss": 0.1684,
261
+ "step": 720
262
+ },
263
+ {
264
+ "epoch": 0.0609622144112478,
265
+ "grad_norm": 0.29022061824798584,
266
+ "learning_rate": 5.94002306805075e-06,
267
+ "loss": 0.2201,
268
+ "step": 740
269
+ },
270
+ {
271
+ "epoch": 0.06260984182776802,
272
+ "grad_norm": 0.49721184372901917,
273
+ "learning_rate": 6.104794859120118e-06,
274
+ "loss": 0.2297,
275
+ "step": 760
276
+ },
277
+ {
278
+ "epoch": 0.06425746924428823,
279
+ "grad_norm": 0.6133716106414795,
280
+ "learning_rate": 6.269566650189487e-06,
281
+ "loss": 0.3173,
282
+ "step": 780
283
+ },
284
+ {
285
+ "epoch": 0.06590509666080843,
286
+ "grad_norm": 0.9667792320251465,
287
+ "learning_rate": 6.434338441258857e-06,
288
+ "loss": 0.2349,
289
+ "step": 800
290
+ },
291
+ {
292
+ "epoch": 0.06755272407732865,
293
+ "grad_norm": 0.3177216053009033,
294
+ "learning_rate": 6.599110232328226e-06,
295
+ "loss": 0.1635,
296
+ "step": 820
297
+ },
298
+ {
299
+ "epoch": 0.06920035149384886,
300
+ "grad_norm": 0.8457621335983276,
301
+ "learning_rate": 6.763882023397594e-06,
302
+ "loss": 0.2777,
303
+ "step": 840
304
+ },
305
+ {
306
+ "epoch": 0.07084797891036906,
307
+ "grad_norm": 0.3330087661743164,
308
+ "learning_rate": 6.928653814466963e-06,
309
+ "loss": 0.2154,
310
+ "step": 860
311
+ },
312
+ {
313
+ "epoch": 0.07249560632688928,
314
+ "grad_norm": 0.5845814943313599,
315
+ "learning_rate": 7.093425605536333e-06,
316
+ "loss": 0.2164,
317
+ "step": 880
318
+ },
319
+ {
320
+ "epoch": 0.07414323374340949,
321
+ "grad_norm": 2.325303554534912,
322
+ "learning_rate": 7.258197396605701e-06,
323
+ "loss": 0.2068,
324
+ "step": 900
325
+ },
326
+ {
327
+ "epoch": 0.0757908611599297,
328
+ "grad_norm": 0.21893823146820068,
329
+ "learning_rate": 7.414730598121602e-06,
330
+ "loss": 0.1621,
331
+ "step": 920
332
+ },
333
+ {
334
+ "epoch": 0.07743848857644992,
335
+ "grad_norm": 0.5854327082633972,
336
+ "learning_rate": 7.5795023891909705e-06,
337
+ "loss": 0.2281,
338
+ "step": 940
339
+ },
340
+ {
341
+ "epoch": 0.07908611599297012,
342
+ "grad_norm": 0.2406030148267746,
343
+ "learning_rate": 7.74427418026034e-06,
344
+ "loss": 0.2342,
345
+ "step": 960
346
+ },
347
+ {
348
+ "epoch": 0.08073374340949033,
349
+ "grad_norm": 1.3764126300811768,
350
+ "learning_rate": 7.90904597132971e-06,
351
+ "loss": 0.2018,
352
+ "step": 980
353
+ },
354
+ {
355
+ "epoch": 0.08238137082601055,
356
+ "grad_norm": 0.9587862491607666,
357
+ "learning_rate": 8.073817762399077e-06,
358
+ "loss": 0.2489,
359
+ "step": 1000
360
+ },
361
+ {
362
+ "epoch": 0.08402899824253075,
363
+ "grad_norm": 0.9726558923721313,
364
+ "learning_rate": 8.238589553468447e-06,
365
+ "loss": 0.154,
366
+ "step": 1020
367
+ },
368
+ {
369
+ "epoch": 0.08567662565905097,
370
+ "grad_norm": 1.9828400611877441,
371
+ "learning_rate": 8.403361344537817e-06,
372
+ "loss": 0.2336,
373
+ "step": 1040
374
+ },
375
+ {
376
+ "epoch": 0.08732425307557118,
377
+ "grad_norm": 0.14761961996555328,
378
+ "learning_rate": 8.568133135607183e-06,
379
+ "loss": 0.1949,
380
+ "step": 1060
381
+ },
382
+ {
383
+ "epoch": 0.08897188049209138,
384
+ "grad_norm": 0.39077144861221313,
385
+ "learning_rate": 8.732904926676553e-06,
386
+ "loss": 0.2067,
387
+ "step": 1080
388
+ },
389
+ {
390
+ "epoch": 0.0906195079086116,
391
+ "grad_norm": 2.3257837295532227,
392
+ "learning_rate": 8.897676717745921e-06,
393
+ "loss": 0.2184,
394
+ "step": 1100
395
+ },
396
+ {
397
+ "epoch": 0.09226713532513181,
398
+ "grad_norm": 1.0966060161590576,
399
+ "learning_rate": 9.062448508815291e-06,
400
+ "loss": 0.1655,
401
+ "step": 1120
402
+ },
403
+ {
404
+ "epoch": 0.09391476274165202,
405
+ "grad_norm": 0.5648412704467773,
406
+ "learning_rate": 9.227220299884661e-06,
407
+ "loss": 0.2097,
408
+ "step": 1140
409
+ },
410
+ {
411
+ "epoch": 0.09556239015817224,
412
+ "grad_norm": 0.49816444516181946,
413
+ "learning_rate": 9.39199209095403e-06,
414
+ "loss": 0.2052,
415
+ "step": 1160
416
+ },
417
+ {
418
+ "epoch": 0.09721001757469244,
419
+ "grad_norm": 1.1164054870605469,
420
+ "learning_rate": 9.556763882023397e-06,
421
+ "loss": 0.2298,
422
+ "step": 1180
423
+ },
424
+ {
425
+ "epoch": 0.09885764499121265,
426
+ "grad_norm": 0.9453270435333252,
427
+ "learning_rate": 9.721535673092767e-06,
428
+ "loss": 0.2157,
429
+ "step": 1200
430
+ },
431
+ {
432
+ "epoch": 0.10050527240773287,
433
+ "grad_norm": 0.4588276743888855,
434
+ "learning_rate": 9.886307464162135e-06,
435
+ "loss": 0.1433,
436
+ "step": 1220
437
+ },
438
+ {
439
+ "epoch": 0.10215289982425307,
440
+ "grad_norm": 0.47535696625709534,
441
+ "learning_rate": 1.0051079255231505e-05,
442
+ "loss": 0.2115,
443
+ "step": 1240
444
+ },
445
+ {
446
+ "epoch": 0.10380052724077328,
447
+ "grad_norm": 0.585959792137146,
448
+ "learning_rate": 1.0215851046300875e-05,
449
+ "loss": 0.2245,
450
+ "step": 1260
451
+ },
452
+ {
453
+ "epoch": 0.1054481546572935,
454
+ "grad_norm": 0.24861204624176025,
455
+ "learning_rate": 1.0380622837370241e-05,
456
+ "loss": 0.1981,
457
+ "step": 1280
458
+ },
459
+ {
460
+ "epoch": 0.1070957820738137,
461
+ "grad_norm": 31.602128982543945,
462
+ "learning_rate": 1.0545394628439611e-05,
463
+ "loss": 0.2265,
464
+ "step": 1300
465
+ },
466
+ {
467
+ "epoch": 0.10874340949033393,
468
+ "grad_norm": 0.6234269142150879,
469
+ "learning_rate": 1.0710166419508981e-05,
470
+ "loss": 0.154,
471
+ "step": 1320
472
+ },
473
+ {
474
+ "epoch": 0.11039103690685413,
475
+ "grad_norm": 1.2423540353775024,
476
+ "learning_rate": 1.087493821057835e-05,
477
+ "loss": 0.2062,
478
+ "step": 1340
479
+ },
480
+ {
481
+ "epoch": 0.11203866432337434,
482
+ "grad_norm": 0.2090279757976532,
483
+ "learning_rate": 1.1039710001647719e-05,
484
+ "loss": 0.2214,
485
+ "step": 1360
486
+ },
487
+ {
488
+ "epoch": 0.11368629173989456,
489
+ "grad_norm": 0.6145613193511963,
490
+ "learning_rate": 1.1204481792717087e-05,
491
+ "loss": 0.1942,
492
+ "step": 1380
493
+ },
494
+ {
495
+ "epoch": 0.11533391915641476,
496
+ "grad_norm": 0.9004138708114624,
497
+ "learning_rate": 1.1369253583786455e-05,
498
+ "loss": 0.2271,
499
+ "step": 1400
500
+ },
501
+ {
502
+ "epoch": 0.11698154657293497,
503
+ "grad_norm": 1.609165906906128,
504
+ "learning_rate": 1.1534025374855825e-05,
505
+ "loss": 0.1606,
506
+ "step": 1420
507
+ },
508
+ {
509
+ "epoch": 0.11862917398945519,
510
+ "grad_norm": 0.8725568652153015,
511
+ "learning_rate": 1.1698797165925195e-05,
512
+ "loss": 0.204,
513
+ "step": 1440
514
+ },
515
+ {
516
+ "epoch": 0.1202768014059754,
517
+ "grad_norm": 1.8169455528259277,
518
+ "learning_rate": 1.1863568956994563e-05,
519
+ "loss": 0.207,
520
+ "step": 1460
521
+ },
522
+ {
523
+ "epoch": 0.1219244288224956,
524
+ "grad_norm": 0.37334388494491577,
525
+ "learning_rate": 1.2028340748063933e-05,
526
+ "loss": 0.2033,
527
+ "step": 1480
528
+ },
529
+ {
530
+ "epoch": 0.12357205623901582,
531
+ "grad_norm": 15.727474212646484,
532
+ "learning_rate": 1.2193112539133301e-05,
533
+ "loss": 0.3298,
534
+ "step": 1500
535
+ },
536
+ {
537
+ "epoch": 0.12521968365553604,
538
+ "grad_norm": 4.70903205871582,
539
+ "learning_rate": 1.2357884330202669e-05,
540
+ "loss": 0.159,
541
+ "step": 1520
542
+ },
543
+ {
544
+ "epoch": 0.12686731107205623,
545
+ "grad_norm": 0.4843326807022095,
546
+ "learning_rate": 1.2522656121272039e-05,
547
+ "loss": 0.2141,
548
+ "step": 1540
549
+ },
550
+ {
551
+ "epoch": 0.12851493848857645,
552
+ "grad_norm": 0.572084367275238,
553
+ "learning_rate": 1.2687427912341407e-05,
554
+ "loss": 0.2182,
555
+ "step": 1560
556
+ },
557
+ {
558
+ "epoch": 0.13016256590509667,
559
+ "grad_norm": 0.31078797578811646,
560
+ "learning_rate": 1.2852199703410777e-05,
561
+ "loss": 0.1994,
562
+ "step": 1580
563
+ },
564
+ {
565
+ "epoch": 0.13181019332161686,
566
+ "grad_norm": 9.014205932617188,
567
+ "learning_rate": 1.3016971494480145e-05,
568
+ "loss": 0.2089,
569
+ "step": 1600
570
+ },
571
+ {
572
+ "epoch": 0.13345782073813708,
573
+ "grad_norm": 1.0150245428085327,
574
+ "learning_rate": 1.3181743285549513e-05,
575
+ "loss": 0.1521,
576
+ "step": 1620
577
+ },
578
+ {
579
+ "epoch": 0.1351054481546573,
580
+ "grad_norm": 0.25271451473236084,
581
+ "learning_rate": 1.3346515076618885e-05,
582
+ "loss": 0.1996,
583
+ "step": 1640
584
+ },
585
+ {
586
+ "epoch": 0.1367530755711775,
587
+ "grad_norm": 0.47118502855300903,
588
+ "learning_rate": 1.3511286867688253e-05,
589
+ "loss": 0.2059,
590
+ "step": 1660
591
+ },
592
+ {
593
+ "epoch": 0.13840070298769772,
594
+ "grad_norm": 0.5134350657463074,
595
+ "learning_rate": 1.3676058658757621e-05,
596
+ "loss": 0.1935,
597
+ "step": 1680
598
+ },
599
+ {
600
+ "epoch": 0.14004833040421794,
601
+ "grad_norm": 1.0354816913604736,
602
+ "learning_rate": 1.384083044982699e-05,
603
+ "loss": 0.2103,
604
+ "step": 1700
605
+ },
606
+ {
607
+ "epoch": 0.14169595782073813,
608
+ "grad_norm": 0.5588876605033875,
609
+ "learning_rate": 1.4005602240896359e-05,
610
+ "loss": 0.1598,
611
+ "step": 1720
612
+ },
613
+ {
614
+ "epoch": 0.14334358523725835,
615
+ "grad_norm": 0.7309175133705139,
616
+ "learning_rate": 1.4170374031965727e-05,
617
+ "loss": 0.2204,
618
+ "step": 1740
619
+ },
620
+ {
621
+ "epoch": 0.14499121265377857,
622
+ "grad_norm": 0.6155902743339539,
623
+ "learning_rate": 1.4335145823035099e-05,
624
+ "loss": 0.2133,
625
+ "step": 1760
626
+ },
627
+ {
628
+ "epoch": 0.14663884007029876,
629
+ "grad_norm": 0.7660940885543823,
630
+ "learning_rate": 1.4499917614104467e-05,
631
+ "loss": 0.2065,
632
+ "step": 1780
633
+ },
634
+ {
635
+ "epoch": 0.14828646748681898,
636
+ "grad_norm": 1.1954026222229004,
637
+ "learning_rate": 1.4664689405173835e-05,
638
+ "loss": 0.2147,
639
+ "step": 1800
640
+ },
641
+ {
642
+ "epoch": 0.1499340949033392,
643
+ "grad_norm": 0.4249323606491089,
644
+ "learning_rate": 1.4829461196243205e-05,
645
+ "loss": 0.1553,
646
+ "step": 1820
647
+ },
648
+ {
649
+ "epoch": 0.1515817223198594,
650
+ "grad_norm": 2.9014129638671875,
651
+ "learning_rate": 1.4994232987312573e-05,
652
+ "loss": 0.2208,
653
+ "step": 1840
654
+ },
655
+ {
656
+ "epoch": 0.1532293497363796,
657
+ "grad_norm": 1.6474498510360718,
658
+ "learning_rate": 1.5159004778381941e-05,
659
+ "loss": 0.209,
660
+ "step": 1860
661
+ },
662
+ {
663
+ "epoch": 0.15487697715289983,
664
+ "grad_norm": 0.1585623174905777,
665
+ "learning_rate": 1.532377656945131e-05,
666
+ "loss": 0.1873,
667
+ "step": 1880
668
+ },
669
+ {
670
+ "epoch": 0.15652460456942002,
671
+ "grad_norm": 1.171941876411438,
672
+ "learning_rate": 1.548854836052068e-05,
673
+ "loss": 0.2389,
674
+ "step": 1900
675
+ },
676
+ {
677
+ "epoch": 0.15817223198594024,
678
+ "grad_norm": 0.48890382051467896,
679
+ "learning_rate": 1.5653320151590047e-05,
680
+ "loss": 0.1679,
681
+ "step": 1920
682
+ },
683
+ {
684
+ "epoch": 0.15981985940246046,
685
+ "grad_norm": 0.5568016767501831,
686
+ "learning_rate": 1.581809194265942e-05,
687
+ "loss": 0.1968,
688
+ "step": 1940
689
+ },
690
+ {
691
+ "epoch": 0.16146748681898065,
692
+ "grad_norm": 0.9775394797325134,
693
+ "learning_rate": 1.5982863733728787e-05,
694
+ "loss": 0.2208,
695
+ "step": 1960
696
+ },
697
+ {
698
+ "epoch": 0.16311511423550087,
699
+ "grad_norm": 0.60302734375,
700
+ "learning_rate": 1.6147635524798155e-05,
701
+ "loss": 0.1929,
702
+ "step": 1980
703
+ },
704
+ {
705
+ "epoch": 0.1647627416520211,
706
+ "grad_norm": 1.7513552904129028,
707
+ "learning_rate": 1.6312407315867526e-05,
708
+ "loss": 0.2055,
709
+ "step": 2000
710
+ },
711
+ {
712
+ "epoch": 0.1647627416520211,
713
+ "eval_loss": 0.5699400305747986,
714
+ "eval_runtime": 686.8117,
715
+ "eval_samples_per_second": 31.725,
716
+ "eval_steps_per_second": 7.932,
717
+ "eval_wer": 0.23469093535410654,
718
+ "step": 2000
719
+ },
720
+ {
721
+ "epoch": 0.16641036906854131,
722
+ "grad_norm": 1.049734354019165,
723
+ "learning_rate": 1.6477179106936894e-05,
724
+ "loss": 0.1608,
725
+ "step": 2020
726
+ },
727
+ {
728
+ "epoch": 0.1680579964850615,
729
+ "grad_norm": 1.7113618850708008,
730
+ "learning_rate": 1.6641950898006263e-05,
731
+ "loss": 0.2008,
732
+ "step": 2040
733
+ },
734
+ {
735
+ "epoch": 0.16970562390158173,
736
+ "grad_norm": 0.3202134370803833,
737
+ "learning_rate": 1.6806722689075634e-05,
738
+ "loss": 0.1973,
739
+ "step": 2060
740
+ },
741
+ {
742
+ "epoch": 0.17135325131810195,
743
+ "grad_norm": 0.3612610697746277,
744
+ "learning_rate": 1.6971494480145e-05,
745
+ "loss": 0.1732,
746
+ "step": 2080
747
+ },
748
+ {
749
+ "epoch": 0.17300087873462214,
750
+ "grad_norm": 1.8115849494934082,
751
+ "learning_rate": 1.7136266271214367e-05,
752
+ "loss": 0.2138,
753
+ "step": 2100
754
+ },
755
+ {
756
+ "epoch": 0.17464850615114236,
757
+ "grad_norm": 0.7046949863433838,
758
+ "learning_rate": 1.7301038062283735e-05,
759
+ "loss": 0.1571,
760
+ "step": 2120
761
+ },
762
+ {
763
+ "epoch": 0.17629613356766258,
764
+ "grad_norm": 0.5983096957206726,
765
+ "learning_rate": 1.7465809853353107e-05,
766
+ "loss": 0.2092,
767
+ "step": 2140
768
+ },
769
+ {
770
+ "epoch": 0.17794376098418277,
771
+ "grad_norm": 0.17064958810806274,
772
+ "learning_rate": 1.7630581644422475e-05,
773
+ "loss": 0.2083,
774
+ "step": 2160
775
+ },
776
+ {
777
+ "epoch": 0.179591388400703,
778
+ "grad_norm": 1.422013759613037,
779
+ "learning_rate": 1.7795353435491843e-05,
780
+ "loss": 0.2087,
781
+ "step": 2180
782
+ },
783
+ {
784
+ "epoch": 0.1812390158172232,
785
+ "grad_norm": 0.982097864151001,
786
+ "learning_rate": 1.7960125226561214e-05,
787
+ "loss": 0.2161,
788
+ "step": 2200
789
+ },
790
+ {
791
+ "epoch": 0.1828866432337434,
792
+ "grad_norm": 0.2690947949886322,
793
+ "learning_rate": 1.8124897017630583e-05,
794
+ "loss": 0.1693,
795
+ "step": 2220
796
+ },
797
+ {
798
+ "epoch": 0.18453427065026362,
799
+ "grad_norm": 0.24867244064807892,
800
+ "learning_rate": 1.828966880869995e-05,
801
+ "loss": 0.2058,
802
+ "step": 2240
803
+ },
804
+ {
805
+ "epoch": 0.18618189806678384,
806
+ "grad_norm": 0.9435555934906006,
807
+ "learning_rate": 1.8454440599769322e-05,
808
+ "loss": 0.1991,
809
+ "step": 2260
810
+ },
811
+ {
812
+ "epoch": 0.18782952548330403,
813
+ "grad_norm": 0.23964335024356842,
814
+ "learning_rate": 1.861921239083869e-05,
815
+ "loss": 0.1932,
816
+ "step": 2280
817
+ },
818
+ {
819
+ "epoch": 0.18947715289982425,
820
+ "grad_norm": 2.1007418632507324,
821
+ "learning_rate": 1.878398418190806e-05,
822
+ "loss": 0.2075,
823
+ "step": 2300
824
+ },
825
+ {
826
+ "epoch": 0.19112478031634447,
827
+ "grad_norm": 0.31368857622146606,
828
+ "learning_rate": 1.8948755972977427e-05,
829
+ "loss": 0.1557,
830
+ "step": 2320
831
+ },
832
+ {
833
+ "epoch": 0.19277240773286466,
834
+ "grad_norm": 0.4029647409915924,
835
+ "learning_rate": 1.9113527764046795e-05,
836
+ "loss": 0.1952,
837
+ "step": 2340
838
+ },
839
+ {
840
+ "epoch": 0.19442003514938488,
841
+ "grad_norm": 0.38545289635658264,
842
+ "learning_rate": 1.9278299555116163e-05,
843
+ "loss": 0.1998,
844
+ "step": 2360
845
+ },
846
+ {
847
+ "epoch": 0.1960676625659051,
848
+ "grad_norm": 0.44573166966438293,
849
+ "learning_rate": 1.9443071346185534e-05,
850
+ "loss": 0.2078,
851
+ "step": 2380
852
+ },
853
+ {
854
+ "epoch": 0.1977152899824253,
855
+ "grad_norm": 9.132265090942383,
856
+ "learning_rate": 1.9607843137254903e-05,
857
+ "loss": 0.2175,
858
+ "step": 2400
859
+ },
860
+ {
861
+ "epoch": 0.19936291739894552,
862
+ "grad_norm": 0.42929643392562866,
863
+ "learning_rate": 1.977261492832427e-05,
864
+ "loss": 0.1762,
865
+ "step": 2420
866
+ },
867
+ {
868
+ "epoch": 0.20101054481546574,
869
+ "grad_norm": 0.6267173886299133,
870
+ "learning_rate": 1.9937386719393642e-05,
871
+ "loss": 0.2101,
872
+ "step": 2440
873
+ },
874
+ {
875
+ "epoch": 0.20265817223198593,
876
+ "grad_norm": 9.70997142791748,
877
+ "learning_rate": 2.010215851046301e-05,
878
+ "loss": 0.2101,
879
+ "step": 2460
880
+ },
881
+ {
882
+ "epoch": 0.20430579964850615,
883
+ "grad_norm": 0.47748956084251404,
884
+ "learning_rate": 2.026693030153238e-05,
885
+ "loss": 0.2039,
886
+ "step": 2480
887
+ },
888
+ {
889
+ "epoch": 0.20595342706502637,
890
+ "grad_norm": 1.3222582340240479,
891
+ "learning_rate": 2.043170209260175e-05,
892
+ "loss": 0.213,
893
+ "step": 2500
894
+ },
895
+ {
896
+ "epoch": 0.20760105448154656,
897
+ "grad_norm": 0.4152863919734955,
898
+ "learning_rate": 2.0596473883671115e-05,
899
+ "loss": 0.148,
900
+ "step": 2520
901
+ },
902
+ {
903
+ "epoch": 0.20924868189806678,
904
+ "grad_norm": 0.7384160757064819,
905
+ "learning_rate": 2.0761245674740483e-05,
906
+ "loss": 0.2138,
907
+ "step": 2540
908
+ },
909
+ {
910
+ "epoch": 0.210896309314587,
911
+ "grad_norm": 0.27651092410087585,
912
+ "learning_rate": 2.0926017465809854e-05,
913
+ "loss": 0.2046,
914
+ "step": 2560
915
+ },
916
+ {
917
+ "epoch": 0.21254393673110722,
918
+ "grad_norm": 0.226897194981575,
919
+ "learning_rate": 2.1090789256879222e-05,
920
+ "loss": 0.1904,
921
+ "step": 2580
922
+ },
923
+ {
924
+ "epoch": 0.2141915641476274,
925
+ "grad_norm": 1.2391464710235596,
926
+ "learning_rate": 2.125556104794859e-05,
927
+ "loss": 0.204,
928
+ "step": 2600
929
+ },
930
+ {
931
+ "epoch": 0.21583919156414763,
932
+ "grad_norm": 1.6048617362976074,
933
+ "learning_rate": 2.1420332839017962e-05,
934
+ "loss": 0.1548,
935
+ "step": 2620
936
+ },
937
+ {
938
+ "epoch": 0.21748681898066785,
939
+ "grad_norm": 0.28409889340400696,
940
+ "learning_rate": 2.158510463008733e-05,
941
+ "loss": 0.201,
942
+ "step": 2640
943
+ },
944
+ {
945
+ "epoch": 0.21913444639718804,
946
+ "grad_norm": 0.354885995388031,
947
+ "learning_rate": 2.17498764211567e-05,
948
+ "loss": 0.2083,
949
+ "step": 2660
950
+ },
951
+ {
952
+ "epoch": 0.22078207381370826,
953
+ "grad_norm": 0.2778099775314331,
954
+ "learning_rate": 2.191464821222607e-05,
955
+ "loss": 0.1891,
956
+ "step": 2680
957
+ },
958
+ {
959
+ "epoch": 0.22242970123022848,
960
+ "grad_norm": 1.007686734199524,
961
+ "learning_rate": 2.2079420003295438e-05,
962
+ "loss": 0.2152,
963
+ "step": 2700
964
+ },
965
+ {
966
+ "epoch": 0.22407732864674867,
967
+ "grad_norm": 0.9725649952888489,
968
+ "learning_rate": 2.2244191794364806e-05,
969
+ "loss": 0.1581,
970
+ "step": 2720
971
+ },
972
+ {
973
+ "epoch": 0.2257249560632689,
974
+ "grad_norm": 0.2451123297214508,
975
+ "learning_rate": 2.2408963585434174e-05,
976
+ "loss": 0.2019,
977
+ "step": 2740
978
+ },
979
+ {
980
+ "epoch": 0.22737258347978911,
981
+ "grad_norm": 0.3667006194591522,
982
+ "learning_rate": 2.2573735376503542e-05,
983
+ "loss": 0.2083,
984
+ "step": 2760
985
+ },
986
+ {
987
+ "epoch": 0.2290202108963093,
988
+ "grad_norm": 3.1283884048461914,
989
+ "learning_rate": 2.273850716757291e-05,
990
+ "loss": 0.194,
991
+ "step": 2780
992
+ },
993
+ {
994
+ "epoch": 0.23066783831282953,
995
+ "grad_norm": 0.7148507237434387,
996
+ "learning_rate": 2.2903278958642282e-05,
997
+ "loss": 0.1931,
998
+ "step": 2800
999
+ },
1000
+ {
1001
+ "epoch": 0.23231546572934975,
1002
+ "grad_norm": 0.5805519223213196,
1003
+ "learning_rate": 2.306805074971165e-05,
1004
+ "loss": 0.1552,
1005
+ "step": 2820
1006
+ },
1007
+ {
1008
+ "epoch": 0.23396309314586994,
1009
+ "grad_norm": 0.8168196082115173,
1010
+ "learning_rate": 2.323282254078102e-05,
1011
+ "loss": 0.2107,
1012
+ "step": 2840
1013
+ },
1014
+ {
1015
+ "epoch": 0.23561072056239016,
1016
+ "grad_norm": 0.17171867191791534,
1017
+ "learning_rate": 2.339759433185039e-05,
1018
+ "loss": 0.2195,
1019
+ "step": 2860
1020
+ },
1021
+ {
1022
+ "epoch": 0.23725834797891038,
1023
+ "grad_norm": 0.6692082285881042,
1024
+ "learning_rate": 2.3562366122919758e-05,
1025
+ "loss": 0.1975,
1026
+ "step": 2880
1027
+ },
1028
+ {
1029
+ "epoch": 0.23890597539543057,
1030
+ "grad_norm": 1.5185160636901855,
1031
+ "learning_rate": 2.3727137913989126e-05,
1032
+ "loss": 0.2086,
1033
+ "step": 2900
1034
+ },
1035
+ {
1036
+ "epoch": 0.2405536028119508,
1037
+ "grad_norm": 0.8978987336158752,
1038
+ "learning_rate": 2.3891909705058498e-05,
1039
+ "loss": 0.1537,
1040
+ "step": 2920
1041
+ },
1042
+ {
1043
+ "epoch": 0.242201230228471,
1044
+ "grad_norm": 1.1462221145629883,
1045
+ "learning_rate": 2.4056681496127866e-05,
1046
+ "loss": 0.215,
1047
+ "step": 2940
1048
+ },
1049
+ {
1050
+ "epoch": 0.2438488576449912,
1051
+ "grad_norm": 0.2455727905035019,
1052
+ "learning_rate": 2.422145328719723e-05,
1053
+ "loss": 0.2137,
1054
+ "step": 2960
1055
+ },
1056
+ {
1057
+ "epoch": 0.24549648506151142,
1058
+ "grad_norm": 0.21464231610298157,
1059
+ "learning_rate": 2.4386225078266602e-05,
1060
+ "loss": 0.203,
1061
+ "step": 2980
1062
+ },
1063
+ {
1064
+ "epoch": 0.24714411247803164,
1065
+ "grad_norm": 0.8478316068649292,
1066
+ "learning_rate": 2.455099686933597e-05,
1067
+ "loss": 0.2085,
1068
+ "step": 3000
1069
+ },
1070
+ {
1071
+ "epoch": 0.24879173989455183,
1072
+ "grad_norm": 0.37225690484046936,
1073
+ "learning_rate": 2.4715768660405338e-05,
1074
+ "loss": 0.1645,
1075
+ "step": 3020
1076
+ },
1077
+ {
1078
+ "epoch": 0.2504393673110721,
1079
+ "grad_norm": 1.3999593257904053,
1080
+ "learning_rate": 2.488054045147471e-05,
1081
+ "loss": 0.2129,
1082
+ "step": 3040
1083
+ },
1084
+ {
1085
+ "epoch": 0.2520869947275923,
1086
+ "grad_norm": 2.0909199714660645,
1087
+ "learning_rate": 2.5045312242544078e-05,
1088
+ "loss": 0.1926,
1089
+ "step": 3060
1090
+ },
1091
+ {
1092
+ "epoch": 0.25373462214411246,
1093
+ "grad_norm": 0.19655053317546844,
1094
+ "learning_rate": 2.5210084033613446e-05,
1095
+ "loss": 0.2007,
1096
+ "step": 3080
1097
+ },
1098
+ {
1099
+ "epoch": 0.2553822495606327,
1100
+ "grad_norm": 1.2680870294570923,
1101
+ "learning_rate": 2.5374855824682814e-05,
1102
+ "loss": 0.2107,
1103
+ "step": 3100
1104
+ },
1105
+ {
1106
+ "epoch": 0.2570298769771529,
1107
+ "grad_norm": 0.19821316003799438,
1108
+ "learning_rate": 2.5539627615752182e-05,
1109
+ "loss": 0.1511,
1110
+ "step": 3120
1111
+ },
1112
+ {
1113
+ "epoch": 0.2586775043936731,
1114
+ "grad_norm": 0.22427937388420105,
1115
+ "learning_rate": 2.5704399406821554e-05,
1116
+ "loss": 0.2139,
1117
+ "step": 3140
1118
+ },
1119
+ {
1120
+ "epoch": 0.26032513181019334,
1121
+ "grad_norm": 0.4531656503677368,
1122
+ "learning_rate": 2.5869171197890922e-05,
1123
+ "loss": 0.1995,
1124
+ "step": 3160
1125
+ },
1126
+ {
1127
+ "epoch": 0.26197275922671354,
1128
+ "grad_norm": 0.3967747986316681,
1129
+ "learning_rate": 2.603394298896029e-05,
1130
+ "loss": 0.1981,
1131
+ "step": 3180
1132
+ },
1133
+ {
1134
+ "epoch": 0.26362038664323373,
1135
+ "grad_norm": 1.0957462787628174,
1136
+ "learning_rate": 2.6198714780029658e-05,
1137
+ "loss": 0.2092,
1138
+ "step": 3200
1139
+ },
1140
+ {
1141
+ "epoch": 0.265268014059754,
1142
+ "grad_norm": 0.5567193627357483,
1143
+ "learning_rate": 2.6363486571099026e-05,
1144
+ "loss": 0.1642,
1145
+ "step": 3220
1146
+ },
1147
+ {
1148
+ "epoch": 0.26691564147627417,
1149
+ "grad_norm": 0.3523741066455841,
1150
+ "learning_rate": 2.6528258362168395e-05,
1151
+ "loss": 0.2059,
1152
+ "step": 3240
1153
+ },
1154
+ {
1155
+ "epoch": 0.26856326889279436,
1156
+ "grad_norm": 0.40257710218429565,
1157
+ "learning_rate": 2.669303015323777e-05,
1158
+ "loss": 0.195,
1159
+ "step": 3260
1160
+ },
1161
+ {
1162
+ "epoch": 0.2702108963093146,
1163
+ "grad_norm": 0.3187640309333801,
1164
+ "learning_rate": 2.6849563354753665e-05,
1165
+ "loss": 0.1827,
1166
+ "step": 3280
1167
+ },
1168
+ {
1169
+ "epoch": 0.2718585237258348,
1170
+ "grad_norm": 0.7375414967536926,
1171
+ "learning_rate": 2.701433514582304e-05,
1172
+ "loss": 0.2247,
1173
+ "step": 3300
1174
+ },
1175
+ {
1176
+ "epoch": 0.273506151142355,
1177
+ "grad_norm": 0.45597076416015625,
1178
+ "learning_rate": 2.7179106936892408e-05,
1179
+ "loss": 0.1654,
1180
+ "step": 3320
1181
+ },
1182
+ {
1183
+ "epoch": 0.27515377855887524,
1184
+ "grad_norm": 0.21507132053375244,
1185
+ "learning_rate": 2.7343878727961776e-05,
1186
+ "loss": 0.1858,
1187
+ "step": 3340
1188
+ },
1189
+ {
1190
+ "epoch": 0.27680140597539543,
1191
+ "grad_norm": 0.7203060388565063,
1192
+ "learning_rate": 2.7508650519031144e-05,
1193
+ "loss": 0.1908,
1194
+ "step": 3360
1195
+ },
1196
+ {
1197
+ "epoch": 0.2784490333919156,
1198
+ "grad_norm": 0.8007901906967163,
1199
+ "learning_rate": 2.7673422310100512e-05,
1200
+ "loss": 0.1793,
1201
+ "step": 3380
1202
+ },
1203
+ {
1204
+ "epoch": 0.28009666080843587,
1205
+ "grad_norm": 3.210064649581909,
1206
+ "learning_rate": 2.783819410116988e-05,
1207
+ "loss": 0.206,
1208
+ "step": 3400
1209
+ },
1210
+ {
1211
+ "epoch": 0.28174428822495606,
1212
+ "grad_norm": 0.8950255513191223,
1213
+ "learning_rate": 2.8002965892239252e-05,
1214
+ "loss": 0.1531,
1215
+ "step": 3420
1216
+ },
1217
+ {
1218
+ "epoch": 0.28339191564147626,
1219
+ "grad_norm": 0.4942973256111145,
1220
+ "learning_rate": 2.816773768330862e-05,
1221
+ "loss": 0.1916,
1222
+ "step": 3440
1223
+ },
1224
+ {
1225
+ "epoch": 0.2850395430579965,
1226
+ "grad_norm": 0.31426137685775757,
1227
+ "learning_rate": 2.8332509474377988e-05,
1228
+ "loss": 0.2013,
1229
+ "step": 3460
1230
+ },
1231
+ {
1232
+ "epoch": 0.2866871704745167,
1233
+ "grad_norm": 0.47154414653778076,
1234
+ "learning_rate": 2.8497281265447356e-05,
1235
+ "loss": 0.1831,
1236
+ "step": 3480
1237
+ },
1238
+ {
1239
+ "epoch": 0.2883347978910369,
1240
+ "grad_norm": 0.8456603288650513,
1241
+ "learning_rate": 2.8662053056516724e-05,
1242
+ "loss": 0.2043,
1243
+ "step": 3500
1244
+ },
1245
+ {
1246
+ "epoch": 0.28998242530755713,
1247
+ "grad_norm": 0.29031482338905334,
1248
+ "learning_rate": 2.8826824847586092e-05,
1249
+ "loss": 0.1607,
1250
+ "step": 3520
1251
+ },
1252
+ {
1253
+ "epoch": 0.2916300527240773,
1254
+ "grad_norm": 0.3170378804206848,
1255
+ "learning_rate": 2.8991596638655467e-05,
1256
+ "loss": 0.1916,
1257
+ "step": 3540
1258
+ },
1259
+ {
1260
+ "epoch": 0.2932776801405975,
1261
+ "grad_norm": 0.3800877332687378,
1262
+ "learning_rate": 2.9156368429724836e-05,
1263
+ "loss": 0.2051,
1264
+ "step": 3560
1265
+ },
1266
+ {
1267
+ "epoch": 0.29492530755711777,
1268
+ "grad_norm": 0.5847609639167786,
1269
+ "learning_rate": 2.9321140220794204e-05,
1270
+ "loss": 0.196,
1271
+ "step": 3580
1272
+ },
1273
+ {
1274
+ "epoch": 0.29657293497363796,
1275
+ "grad_norm": 1.0933667421340942,
1276
+ "learning_rate": 2.9485912011863572e-05,
1277
+ "loss": 0.2154,
1278
+ "step": 3600
1279
+ },
1280
+ {
1281
+ "epoch": 0.29822056239015815,
1282
+ "grad_norm": 4.349573135375977,
1283
+ "learning_rate": 2.9650683802932937e-05,
1284
+ "loss": 0.1606,
1285
+ "step": 3620
1286
+ },
1287
+ {
1288
+ "epoch": 0.2998681898066784,
1289
+ "grad_norm": 0.4264489710330963,
1290
+ "learning_rate": 2.9815455594002305e-05,
1291
+ "loss": 0.2117,
1292
+ "step": 3640
1293
+ },
1294
+ {
1295
+ "epoch": 0.3015158172231986,
1296
+ "grad_norm": 0.47935691475868225,
1297
+ "learning_rate": 2.998022738507168e-05,
1298
+ "loss": 0.1901,
1299
+ "step": 3660
1300
+ },
1301
+ {
1302
+ "epoch": 0.3031634446397188,
1303
+ "grad_norm": 0.7258153557777405,
1304
+ "learning_rate": 3.0144999176141048e-05,
1305
+ "loss": 0.189,
1306
+ "step": 3680
1307
+ },
1308
+ {
1309
+ "epoch": 0.30481107205623903,
1310
+ "grad_norm": 2.0093941688537598,
1311
+ "learning_rate": 3.0309770967210416e-05,
1312
+ "loss": 0.2104,
1313
+ "step": 3700
1314
+ },
1315
+ {
1316
+ "epoch": 0.3064586994727592,
1317
+ "grad_norm": 1.1718699932098389,
1318
+ "learning_rate": 3.0474542758279784e-05,
1319
+ "loss": 0.1577,
1320
+ "step": 3720
1321
+ },
1322
+ {
1323
+ "epoch": 0.3081063268892794,
1324
+ "grad_norm": 0.19388867914676666,
1325
+ "learning_rate": 3.063931454934915e-05,
1326
+ "loss": 0.2011,
1327
+ "step": 3740
1328
+ },
1329
+ {
1330
+ "epoch": 0.30975395430579966,
1331
+ "grad_norm": 0.2112320065498352,
1332
+ "learning_rate": 3.080408634041852e-05,
1333
+ "loss": 0.2077,
1334
+ "step": 3760
1335
+ },
1336
+ {
1337
+ "epoch": 0.31140158172231985,
1338
+ "grad_norm": 1.9554697275161743,
1339
+ "learning_rate": 3.096885813148789e-05,
1340
+ "loss": 0.1861,
1341
+ "step": 3780
1342
+ },
1343
+ {
1344
+ "epoch": 0.31304920913884005,
1345
+ "grad_norm": 0.7065563201904297,
1346
+ "learning_rate": 3.113362992255726e-05,
1347
+ "loss": 0.2071,
1348
+ "step": 3800
1349
+ },
1350
+ {
1351
+ "epoch": 0.3146968365553603,
1352
+ "grad_norm": 0.4599238634109497,
1353
+ "learning_rate": 3.129840171362663e-05,
1354
+ "loss": 0.1655,
1355
+ "step": 3820
1356
+ },
1357
+ {
1358
+ "epoch": 0.3163444639718805,
1359
+ "grad_norm": 0.5441445708274841,
1360
+ "learning_rate": 3.1463173504695996e-05,
1361
+ "loss": 0.1971,
1362
+ "step": 3840
1363
+ },
1364
+ {
1365
+ "epoch": 0.3179920913884007,
1366
+ "grad_norm": 1.1993862390518188,
1367
+ "learning_rate": 3.1627945295765364e-05,
1368
+ "loss": 0.2048,
1369
+ "step": 3860
1370
+ },
1371
+ {
1372
+ "epoch": 0.3196397188049209,
1373
+ "grad_norm": 0.3095191717147827,
1374
+ "learning_rate": 3.179271708683473e-05,
1375
+ "loss": 0.2009,
1376
+ "step": 3880
1377
+ },
1378
+ {
1379
+ "epoch": 0.3212873462214411,
1380
+ "grad_norm": 1.0743999481201172,
1381
+ "learning_rate": 3.195748887790411e-05,
1382
+ "loss": 0.2371,
1383
+ "step": 3900
1384
+ },
1385
+ {
1386
+ "epoch": 0.3229349736379613,
1387
+ "grad_norm": 0.5000220537185669,
1388
+ "learning_rate": 3.2122260668973475e-05,
1389
+ "loss": 0.1722,
1390
+ "step": 3920
1391
+ },
1392
+ {
1393
+ "epoch": 0.32458260105448156,
1394
+ "grad_norm": 1.1417018175125122,
1395
+ "learning_rate": 3.2287032460042844e-05,
1396
+ "loss": 0.2079,
1397
+ "step": 3940
1398
+ },
1399
+ {
1400
+ "epoch": 0.32623022847100175,
1401
+ "grad_norm": 1.099433422088623,
1402
+ "learning_rate": 3.245180425111221e-05,
1403
+ "loss": 0.2009,
1404
+ "step": 3960
1405
+ },
1406
+ {
1407
+ "epoch": 0.32787785588752194,
1408
+ "grad_norm": 0.3827146589756012,
1409
+ "learning_rate": 3.261657604218158e-05,
1410
+ "loss": 0.2171,
1411
+ "step": 3980
1412
+ },
1413
+ {
1414
+ "epoch": 0.3295254833040422,
1415
+ "grad_norm": 1.1845418214797974,
1416
+ "learning_rate": 3.278134783325095e-05,
1417
+ "loss": 0.2195,
1418
+ "step": 4000
1419
+ },
1420
+ {
1421
+ "epoch": 0.3295254833040422,
1422
+ "eval_loss": 0.5868579149246216,
1423
+ "eval_runtime": 260.92,
1424
+ "eval_samples_per_second": 83.508,
1425
+ "eval_steps_per_second": 20.88,
1426
+ "eval_wer": 0.2350240606008012,
1427
+ "step": 4000
1428
+ },
1429
+ {
1430
+ "epoch": 0.3311731107205624,
1431
+ "grad_norm": 1.6073412895202637,
1432
+ "learning_rate": 3.2946119624320316e-05,
1433
+ "loss": 0.1695,
1434
+ "step": 4020
1435
+ },
1436
+ {
1437
+ "epoch": 0.33282073813708263,
1438
+ "grad_norm": 2.9597036838531494,
1439
+ "learning_rate": 3.311089141538969e-05,
1440
+ "loss": 0.2484,
1441
+ "step": 4040
1442
+ },
1443
+ {
1444
+ "epoch": 0.3344683655536028,
1445
+ "grad_norm": 1.4150543212890625,
1446
+ "learning_rate": 3.327566320645906e-05,
1447
+ "loss": 0.2087,
1448
+ "step": 4060
1449
+ },
1450
+ {
1451
+ "epoch": 0.336115992970123,
1452
+ "grad_norm": 0.255397230386734,
1453
+ "learning_rate": 3.344043499752843e-05,
1454
+ "loss": 0.2048,
1455
+ "step": 4080
1456
+ },
1457
+ {
1458
+ "epoch": 0.33776362038664326,
1459
+ "grad_norm": 1.9232691526412964,
1460
+ "learning_rate": 3.3605206788597795e-05,
1461
+ "loss": 0.2274,
1462
+ "step": 4100
1463
+ },
1464
+ {
1465
+ "epoch": 0.33941124780316345,
1466
+ "grad_norm": 1.0172194242477417,
1467
+ "learning_rate": 3.3769978579667164e-05,
1468
+ "loss": 0.1779,
1469
+ "step": 4120
1470
+ },
1471
+ {
1472
+ "epoch": 0.34105887521968364,
1473
+ "grad_norm": 1.3862395286560059,
1474
+ "learning_rate": 3.393475037073653e-05,
1475
+ "loss": 0.2095,
1476
+ "step": 4140
1477
+ },
1478
+ {
1479
+ "epoch": 0.3427065026362039,
1480
+ "grad_norm": 0.3353387117385864,
1481
+ "learning_rate": 3.40995221618059e-05,
1482
+ "loss": 0.2021,
1483
+ "step": 4160
1484
+ },
1485
+ {
1486
+ "epoch": 0.3443541300527241,
1487
+ "grad_norm": 0.9549083709716797,
1488
+ "learning_rate": 3.426429395287527e-05,
1489
+ "loss": 0.1997,
1490
+ "step": 4180
1491
+ },
1492
+ {
1493
+ "epoch": 0.3460017574692443,
1494
+ "grad_norm": 1.6077580451965332,
1495
+ "learning_rate": 3.4429065743944636e-05,
1496
+ "loss": 0.2562,
1497
+ "step": 4200
1498
+ },
1499
+ {
1500
+ "epoch": 0.3476493848857645,
1501
+ "grad_norm": 5.387716770172119,
1502
+ "learning_rate": 3.4593837535014004e-05,
1503
+ "loss": 0.1726,
1504
+ "step": 4220
1505
+ },
1506
+ {
1507
+ "epoch": 0.3492970123022847,
1508
+ "grad_norm": 0.5455642342567444,
1509
+ "learning_rate": 3.475860932608337e-05,
1510
+ "loss": 0.2342,
1511
+ "step": 4240
1512
+ },
1513
+ {
1514
+ "epoch": 0.3509446397188049,
1515
+ "grad_norm": 0.18990729749202728,
1516
+ "learning_rate": 3.492338111715274e-05,
1517
+ "loss": 0.2334,
1518
+ "step": 4260
1519
+ },
1520
+ {
1521
+ "epoch": 0.35259226713532515,
1522
+ "grad_norm": 0.4878564476966858,
1523
+ "learning_rate": 3.5088152908222115e-05,
1524
+ "loss": 0.2065,
1525
+ "step": 4280
1526
+ },
1527
+ {
1528
+ "epoch": 0.35423989455184535,
1529
+ "grad_norm": 1.3400063514709473,
1530
+ "learning_rate": 3.5252924699291483e-05,
1531
+ "loss": 0.237,
1532
+ "step": 4300
1533
+ },
1534
+ {
1535
+ "epoch": 0.35588752196836554,
1536
+ "grad_norm": 0.5822551250457764,
1537
+ "learning_rate": 3.541769649036085e-05,
1538
+ "loss": 0.1787,
1539
+ "step": 4320
1540
+ },
1541
+ {
1542
+ "epoch": 0.3575351493848858,
1543
+ "grad_norm": 0.4629223644733429,
1544
+ "learning_rate": 3.558246828143022e-05,
1545
+ "loss": 0.2129,
1546
+ "step": 4340
1547
+ },
1548
+ {
1549
+ "epoch": 0.359182776801406,
1550
+ "grad_norm": 1.4195072650909424,
1551
+ "learning_rate": 3.574724007249959e-05,
1552
+ "loss": 0.2215,
1553
+ "step": 4360
1554
+ },
1555
+ {
1556
+ "epoch": 0.36083040421792617,
1557
+ "grad_norm": 0.3443647623062134,
1558
+ "learning_rate": 3.5912011863568956e-05,
1559
+ "loss": 0.2002,
1560
+ "step": 4380
1561
+ },
1562
+ {
1563
+ "epoch": 0.3624780316344464,
1564
+ "grad_norm": 3.2932443618774414,
1565
+ "learning_rate": 3.607678365463833e-05,
1566
+ "loss": 0.2172,
1567
+ "step": 4400
1568
+ },
1569
+ {
1570
+ "epoch": 0.3641256590509666,
1571
+ "grad_norm": 0.4463866055011749,
1572
+ "learning_rate": 3.62415554457077e-05,
1573
+ "loss": 0.1889,
1574
+ "step": 4420
1575
+ },
1576
+ {
1577
+ "epoch": 0.3657732864674868,
1578
+ "grad_norm": 0.4873151183128357,
1579
+ "learning_rate": 3.640632723677707e-05,
1580
+ "loss": 0.2139,
1581
+ "step": 4440
1582
+ },
1583
+ {
1584
+ "epoch": 0.36742091388400705,
1585
+ "grad_norm": 1.6354761123657227,
1586
+ "learning_rate": 3.6571099027846435e-05,
1587
+ "loss": 0.1986,
1588
+ "step": 4460
1589
+ },
1590
+ {
1591
+ "epoch": 0.36906854130052724,
1592
+ "grad_norm": 0.5571808815002441,
1593
+ "learning_rate": 3.6735870818915803e-05,
1594
+ "loss": 0.2047,
1595
+ "step": 4480
1596
+ },
1597
+ {
1598
+ "epoch": 0.37071616871704743,
1599
+ "grad_norm": 0.7461993098258972,
1600
+ "learning_rate": 3.690064260998517e-05,
1601
+ "loss": 0.2196,
1602
+ "step": 4500
1603
+ },
1604
+ {
1605
+ "epoch": 0.3723637961335677,
1606
+ "grad_norm": 0.2061534970998764,
1607
+ "learning_rate": 3.7065414401054546e-05,
1608
+ "loss": 0.1657,
1609
+ "step": 4520
1610
+ },
1611
+ {
1612
+ "epoch": 0.3740114235500879,
1613
+ "grad_norm": 1.0977954864501953,
1614
+ "learning_rate": 3.7230186192123915e-05,
1615
+ "loss": 0.2062,
1616
+ "step": 4540
1617
+ },
1618
+ {
1619
+ "epoch": 0.37565905096660807,
1620
+ "grad_norm": 2.38232684135437,
1621
+ "learning_rate": 3.739495798319328e-05,
1622
+ "loss": 0.2319,
1623
+ "step": 4560
1624
+ },
1625
+ {
1626
+ "epoch": 0.3773066783831283,
1627
+ "grad_norm": 0.6182531118392944,
1628
+ "learning_rate": 3.755972977426265e-05,
1629
+ "loss": 0.202,
1630
+ "step": 4580
1631
+ },
1632
+ {
1633
+ "epoch": 0.3789543057996485,
1634
+ "grad_norm": 28.744009017944336,
1635
+ "learning_rate": 3.772450156533201e-05,
1636
+ "loss": 0.2235,
1637
+ "step": 4600
1638
+ },
1639
+ {
1640
+ "epoch": 0.3806019332161687,
1641
+ "grad_norm": 0.5899057984352112,
1642
+ "learning_rate": 3.788927335640138e-05,
1643
+ "loss": 0.1577,
1644
+ "step": 4620
1645
+ },
1646
+ {
1647
+ "epoch": 0.38224956063268895,
1648
+ "grad_norm": 0.5423290133476257,
1649
+ "learning_rate": 3.8054045147470755e-05,
1650
+ "loss": 0.2098,
1651
+ "step": 4640
1652
+ },
1653
+ {
1654
+ "epoch": 0.38389718804920914,
1655
+ "grad_norm": 0.23139849305152893,
1656
+ "learning_rate": 3.8218816938540123e-05,
1657
+ "loss": 0.1989,
1658
+ "step": 4660
1659
+ },
1660
+ {
1661
+ "epoch": 0.38554481546572933,
1662
+ "grad_norm": 0.3539600670337677,
1663
+ "learning_rate": 3.838358872960949e-05,
1664
+ "loss": 0.1968,
1665
+ "step": 4680
1666
+ },
1667
+ {
1668
+ "epoch": 0.3871924428822496,
1669
+ "grad_norm": 0.7127693295478821,
1670
+ "learning_rate": 3.854836052067886e-05,
1671
+ "loss": 0.2225,
1672
+ "step": 4700
1673
+ },
1674
+ {
1675
+ "epoch": 0.38884007029876977,
1676
+ "grad_norm": 0.3489459753036499,
1677
+ "learning_rate": 3.871313231174823e-05,
1678
+ "loss": 0.1606,
1679
+ "step": 4720
1680
+ },
1681
+ {
1682
+ "epoch": 0.39048769771528996,
1683
+ "grad_norm": 0.27798184752464294,
1684
+ "learning_rate": 3.886966551326413e-05,
1685
+ "loss": 0.2096,
1686
+ "step": 4740
1687
+ },
1688
+ {
1689
+ "epoch": 0.3921353251318102,
1690
+ "grad_norm": 0.4229481816291809,
1691
+ "learning_rate": 3.90344373043335e-05,
1692
+ "loss": 0.2195,
1693
+ "step": 4760
1694
+ },
1695
+ {
1696
+ "epoch": 0.3937829525483304,
1697
+ "grad_norm": 0.25523656606674194,
1698
+ "learning_rate": 3.9199209095402866e-05,
1699
+ "loss": 0.2106,
1700
+ "step": 4780
1701
+ },
1702
+ {
1703
+ "epoch": 0.3954305799648506,
1704
+ "grad_norm": 0.9676795601844788,
1705
+ "learning_rate": 3.936398088647224e-05,
1706
+ "loss": 0.2267,
1707
+ "step": 4800
1708
+ },
1709
+ {
1710
+ "epoch": 0.39707820738137084,
1711
+ "grad_norm": 0.40875479578971863,
1712
+ "learning_rate": 3.952875267754161e-05,
1713
+ "loss": 0.155,
1714
+ "step": 4820
1715
+ },
1716
+ {
1717
+ "epoch": 0.39872583479789103,
1718
+ "grad_norm": 0.335033655166626,
1719
+ "learning_rate": 3.969352446861098e-05,
1720
+ "loss": 0.2081,
1721
+ "step": 4840
1722
+ },
1723
+ {
1724
+ "epoch": 0.4003734622144112,
1725
+ "grad_norm": 0.3303823173046112,
1726
+ "learning_rate": 3.9858296259680345e-05,
1727
+ "loss": 0.2282,
1728
+ "step": 4860
1729
+ },
1730
+ {
1731
+ "epoch": 0.40202108963093147,
1732
+ "grad_norm": 0.20934663712978363,
1733
+ "learning_rate": 4.0023068050749714e-05,
1734
+ "loss": 0.1965,
1735
+ "step": 4880
1736
+ },
1737
+ {
1738
+ "epoch": 0.40366871704745166,
1739
+ "grad_norm": 1.4093564748764038,
1740
+ "learning_rate": 4.018783984181908e-05,
1741
+ "loss": 0.22,
1742
+ "step": 4900
1743
+ },
1744
+ {
1745
+ "epoch": 0.40531634446397186,
1746
+ "grad_norm": 0.4276560842990875,
1747
+ "learning_rate": 4.035261163288846e-05,
1748
+ "loss": 0.1705,
1749
+ "step": 4920
1750
+ },
1751
+ {
1752
+ "epoch": 0.4069639718804921,
1753
+ "grad_norm": 0.4243983030319214,
1754
+ "learning_rate": 4.0517383423957825e-05,
1755
+ "loss": 0.2285,
1756
+ "step": 4940
1757
+ },
1758
+ {
1759
+ "epoch": 0.4086115992970123,
1760
+ "grad_norm": 0.16951896250247955,
1761
+ "learning_rate": 4.068215521502719e-05,
1762
+ "loss": 0.2234,
1763
+ "step": 4960
1764
+ },
1765
+ {
1766
+ "epoch": 0.4102592267135325,
1767
+ "grad_norm": 0.30336254835128784,
1768
+ "learning_rate": 4.084692700609656e-05,
1769
+ "loss": 0.2152,
1770
+ "step": 4980
1771
+ },
1772
+ {
1773
+ "epoch": 0.41190685413005274,
1774
+ "grad_norm": 1.0761586427688599,
1775
+ "learning_rate": 4.101169879716593e-05,
1776
+ "loss": 0.2093,
1777
+ "step": 5000
1778
+ },
1779
+ {
1780
+ "epoch": 0.4135544815465729,
1781
+ "grad_norm": 0.7221740484237671,
1782
+ "learning_rate": 4.11764705882353e-05,
1783
+ "loss": 0.1538,
1784
+ "step": 5020
1785
+ },
1786
+ {
1787
+ "epoch": 0.4152021089630931,
1788
+ "grad_norm": 0.4801746904850006,
1789
+ "learning_rate": 4.1341242379304665e-05,
1790
+ "loss": 0.1911,
1791
+ "step": 5040
1792
+ },
1793
+ {
1794
+ "epoch": 0.41684973637961337,
1795
+ "grad_norm": 0.311234712600708,
1796
+ "learning_rate": 4.1506014170374034e-05,
1797
+ "loss": 0.2015,
1798
+ "step": 5060
1799
+ },
1800
+ {
1801
+ "epoch": 0.41849736379613356,
1802
+ "grad_norm": 0.6403760313987732,
1803
+ "learning_rate": 4.16707859614434e-05,
1804
+ "loss": 0.2014,
1805
+ "step": 5080
1806
+ },
1807
+ {
1808
+ "epoch": 0.42014499121265375,
1809
+ "grad_norm": 1.2653217315673828,
1810
+ "learning_rate": 4.183555775251277e-05,
1811
+ "loss": 0.2388,
1812
+ "step": 5100
1813
+ },
1814
+ {
1815
+ "epoch": 0.421792618629174,
1816
+ "grad_norm": 0.5536401867866516,
1817
+ "learning_rate": 4.200032954358214e-05,
1818
+ "loss": 0.1553,
1819
+ "step": 5120
1820
+ },
1821
+ {
1822
+ "epoch": 0.4234402460456942,
1823
+ "grad_norm": 0.4605076014995575,
1824
+ "learning_rate": 4.2165101334651506e-05,
1825
+ "loss": 0.2105,
1826
+ "step": 5140
1827
+ },
1828
+ {
1829
+ "epoch": 0.42508787346221444,
1830
+ "grad_norm": 0.28758004307746887,
1831
+ "learning_rate": 4.232987312572088e-05,
1832
+ "loss": 0.2067,
1833
+ "step": 5160
1834
+ },
1835
+ {
1836
+ "epoch": 0.42673550087873463,
1837
+ "grad_norm": 0.36655622720718384,
1838
+ "learning_rate": 4.249464491679025e-05,
1839
+ "loss": 0.1976,
1840
+ "step": 5180
1841
+ },
1842
+ {
1843
+ "epoch": 0.4283831282952548,
1844
+ "grad_norm": 0.9053062796592712,
1845
+ "learning_rate": 4.265941670785962e-05,
1846
+ "loss": 0.2102,
1847
+ "step": 5200
1848
+ },
1849
+ {
1850
+ "epoch": 0.43003075571177507,
1851
+ "grad_norm": 0.5088081359863281,
1852
+ "learning_rate": 4.2824188498928985e-05,
1853
+ "loss": 0.1639,
1854
+ "step": 5220
1855
+ },
1856
+ {
1857
+ "epoch": 0.43167838312829526,
1858
+ "grad_norm": 0.7981218695640564,
1859
+ "learning_rate": 4.2988960289998354e-05,
1860
+ "loss": 0.2024,
1861
+ "step": 5240
1862
+ },
1863
+ {
1864
+ "epoch": 0.43332601054481545,
1865
+ "grad_norm": 0.7993011474609375,
1866
+ "learning_rate": 4.315373208106772e-05,
1867
+ "loss": 0.204,
1868
+ "step": 5260
1869
+ },
1870
+ {
1871
+ "epoch": 0.4349736379613357,
1872
+ "grad_norm": 0.5183406472206116,
1873
+ "learning_rate": 4.331850387213709e-05,
1874
+ "loss": 0.2041,
1875
+ "step": 5280
1876
+ },
1877
+ {
1878
+ "epoch": 0.4366212653778559,
1879
+ "grad_norm": 1.2263312339782715,
1880
+ "learning_rate": 4.3483275663206465e-05,
1881
+ "loss": 0.2182,
1882
+ "step": 5300
1883
+ },
1884
+ {
1885
+ "epoch": 0.4382688927943761,
1886
+ "grad_norm": 0.7018775343894958,
1887
+ "learning_rate": 4.364804745427583e-05,
1888
+ "loss": 0.1685,
1889
+ "step": 5320
1890
+ },
1891
+ {
1892
+ "epoch": 0.43991652021089633,
1893
+ "grad_norm": 0.7381752133369446,
1894
+ "learning_rate": 4.38128192453452e-05,
1895
+ "loss": 0.2073,
1896
+ "step": 5340
1897
+ },
1898
+ {
1899
+ "epoch": 0.4415641476274165,
1900
+ "grad_norm": 0.4658122956752777,
1901
+ "learning_rate": 4.397759103641457e-05,
1902
+ "loss": 0.2031,
1903
+ "step": 5360
1904
+ },
1905
+ {
1906
+ "epoch": 0.4432117750439367,
1907
+ "grad_norm": 0.34789761900901794,
1908
+ "learning_rate": 4.414236282748394e-05,
1909
+ "loss": 0.2023,
1910
+ "step": 5380
1911
+ },
1912
+ {
1913
+ "epoch": 0.44485940246045697,
1914
+ "grad_norm": 0.9787063598632812,
1915
+ "learning_rate": 4.4307134618553305e-05,
1916
+ "loss": 0.2264,
1917
+ "step": 5400
1918
+ },
1919
+ {
1920
+ "epoch": 0.44650702987697716,
1921
+ "grad_norm": 0.3786025047302246,
1922
+ "learning_rate": 4.4471906409622673e-05,
1923
+ "loss": 0.1656,
1924
+ "step": 5420
1925
+ },
1926
+ {
1927
+ "epoch": 0.44815465729349735,
1928
+ "grad_norm": 0.4397692084312439,
1929
+ "learning_rate": 4.463667820069204e-05,
1930
+ "loss": 0.2023,
1931
+ "step": 5440
1932
+ },
1933
+ {
1934
+ "epoch": 0.4498022847100176,
1935
+ "grad_norm": 0.20323756337165833,
1936
+ "learning_rate": 4.480144999176141e-05,
1937
+ "loss": 0.2146,
1938
+ "step": 5460
1939
+ },
1940
+ {
1941
+ "epoch": 0.4514499121265378,
1942
+ "grad_norm": 0.2108180820941925,
1943
+ "learning_rate": 4.496622178283078e-05,
1944
+ "loss": 0.2012,
1945
+ "step": 5480
1946
+ },
1947
+ {
1948
+ "epoch": 0.453097539543058,
1949
+ "grad_norm": 1.6603738069534302,
1950
+ "learning_rate": 4.5130993573900146e-05,
1951
+ "loss": 0.2257,
1952
+ "step": 5500
1953
+ },
1954
+ {
1955
+ "epoch": 0.45474516695957823,
1956
+ "grad_norm": 1.6646907329559326,
1957
+ "learning_rate": 4.5295765364969514e-05,
1958
+ "loss": 0.1636,
1959
+ "step": 5520
1960
+ },
1961
+ {
1962
+ "epoch": 0.4563927943760984,
1963
+ "grad_norm": 0.98222416639328,
1964
+ "learning_rate": 4.546053715603889e-05,
1965
+ "loss": 0.2018,
1966
+ "step": 5540
1967
+ },
1968
+ {
1969
+ "epoch": 0.4580404217926186,
1970
+ "grad_norm": 0.6065666079521179,
1971
+ "learning_rate": 4.562530894710826e-05,
1972
+ "loss": 0.2351,
1973
+ "step": 5560
1974
+ },
1975
+ {
1976
+ "epoch": 0.45968804920913886,
1977
+ "grad_norm": 0.4321737587451935,
1978
+ "learning_rate": 4.5790080738177625e-05,
1979
+ "loss": 0.2139,
1980
+ "step": 5580
1981
+ },
1982
+ {
1983
+ "epoch": 0.46133567662565905,
1984
+ "grad_norm": 2.530203342437744,
1985
+ "learning_rate": 4.5954852529246993e-05,
1986
+ "loss": 0.232,
1987
+ "step": 5600
1988
+ },
1989
+ {
1990
+ "epoch": 0.46298330404217924,
1991
+ "grad_norm": 0.8252795934677124,
1992
+ "learning_rate": 4.611962432031636e-05,
1993
+ "loss": 0.1792,
1994
+ "step": 5620
1995
+ },
1996
+ {
1997
+ "epoch": 0.4646309314586995,
1998
+ "grad_norm": 4.3282880783081055,
1999
+ "learning_rate": 4.628439611138573e-05,
2000
+ "loss": 0.2079,
2001
+ "step": 5640
2002
+ },
2003
+ {
2004
+ "epoch": 0.4662785588752197,
2005
+ "grad_norm": 0.2798108756542206,
2006
+ "learning_rate": 4.6449167902455105e-05,
2007
+ "loss": 0.2114,
2008
+ "step": 5660
2009
+ },
2010
+ {
2011
+ "epoch": 0.4679261862917399,
2012
+ "grad_norm": 0.1499057412147522,
2013
+ "learning_rate": 4.661393969352447e-05,
2014
+ "loss": 0.2026,
2015
+ "step": 5680
2016
+ },
2017
+ {
2018
+ "epoch": 0.4695738137082601,
2019
+ "grad_norm": 0.8664823770523071,
2020
+ "learning_rate": 4.677871148459384e-05,
2021
+ "loss": 0.2208,
2022
+ "step": 5700
2023
+ },
2024
+ {
2025
+ "epoch": 0.4712214411247803,
2026
+ "grad_norm": 0.32247450947761536,
2027
+ "learning_rate": 4.694348327566321e-05,
2028
+ "loss": 0.175,
2029
+ "step": 5720
2030
+ },
2031
+ {
2032
+ "epoch": 0.4728690685413005,
2033
+ "grad_norm": 0.4217327833175659,
2034
+ "learning_rate": 4.710825506673258e-05,
2035
+ "loss": 0.2207,
2036
+ "step": 5740
2037
+ },
2038
+ {
2039
+ "epoch": 0.47451669595782076,
2040
+ "grad_norm": 0.9544996023178101,
2041
+ "learning_rate": 4.7273026857801945e-05,
2042
+ "loss": 0.2269,
2043
+ "step": 5760
2044
+ },
2045
+ {
2046
+ "epoch": 0.47616432337434095,
2047
+ "grad_norm": 0.3182899057865143,
2048
+ "learning_rate": 4.743779864887132e-05,
2049
+ "loss": 0.224,
2050
+ "step": 5780
2051
+ },
2052
+ {
2053
+ "epoch": 0.47781195079086114,
2054
+ "grad_norm": 0.669552743434906,
2055
+ "learning_rate": 4.760257043994069e-05,
2056
+ "loss": 0.2332,
2057
+ "step": 5800
2058
+ },
2059
+ {
2060
+ "epoch": 0.4794595782073814,
2061
+ "grad_norm": 0.5297145247459412,
2062
+ "learning_rate": 4.7767342231010056e-05,
2063
+ "loss": 0.2233,
2064
+ "step": 5820
2065
+ },
2066
+ {
2067
+ "epoch": 0.4811072056239016,
2068
+ "grad_norm": 0.5998116731643677,
2069
+ "learning_rate": 4.7932114022079425e-05,
2070
+ "loss": 0.2282,
2071
+ "step": 5840
2072
+ },
2073
+ {
2074
+ "epoch": 0.48275483304042177,
2075
+ "grad_norm": 0.4391906261444092,
2076
+ "learning_rate": 4.809688581314879e-05,
2077
+ "loss": 0.2094,
2078
+ "step": 5860
2079
+ },
2080
+ {
2081
+ "epoch": 0.484402460456942,
2082
+ "grad_norm": 0.559304416179657,
2083
+ "learning_rate": 4.826165760421816e-05,
2084
+ "loss": 0.2106,
2085
+ "step": 5880
2086
+ },
2087
+ {
2088
+ "epoch": 0.4860500878734622,
2089
+ "grad_norm": 2.914066791534424,
2090
+ "learning_rate": 4.842642939528753e-05,
2091
+ "loss": 0.2484,
2092
+ "step": 5900
2093
+ },
2094
+ {
2095
+ "epoch": 0.4876977152899824,
2096
+ "grad_norm": 0.29989850521087646,
2097
+ "learning_rate": 4.85912011863569e-05,
2098
+ "loss": 0.1631,
2099
+ "step": 5920
2100
+ },
2101
+ {
2102
+ "epoch": 0.48934534270650265,
2103
+ "grad_norm": 2.1414361000061035,
2104
+ "learning_rate": 4.8755972977426265e-05,
2105
+ "loss": 0.2315,
2106
+ "step": 5940
2107
+ },
2108
+ {
2109
+ "epoch": 0.49099297012302284,
2110
+ "grad_norm": 0.5668668746948242,
2111
+ "learning_rate": 4.892074476849563e-05,
2112
+ "loss": 0.218,
2113
+ "step": 5960
2114
+ },
2115
+ {
2116
+ "epoch": 0.49264059753954303,
2117
+ "grad_norm": 0.34356266260147095,
2118
+ "learning_rate": 4.9077277970011535e-05,
2119
+ "loss": 0.2121,
2120
+ "step": 5980
2121
+ },
2122
+ {
2123
+ "epoch": 0.4942882249560633,
2124
+ "grad_norm": 0.7654836773872375,
2125
+ "learning_rate": 4.9242049761080904e-05,
2126
+ "loss": 0.212,
2127
+ "step": 6000
2128
+ },
2129
+ {
2130
+ "epoch": 0.4942882249560633,
2131
+ "eval_loss": 0.5638302564620972,
2132
+ "eval_runtime": 264.9088,
2133
+ "eval_samples_per_second": 82.251,
2134
+ "eval_steps_per_second": 20.566,
2135
+ "eval_wer": 0.24026168527782366,
2136
+ "step": 6000
2137
+ },
2138
+ {
2139
+ "epoch": 0.4959358523725835,
2140
+ "grad_norm": 1.2249395847320557,
2141
+ "learning_rate": 4.940682155215027e-05,
2142
+ "loss": 0.1786,
2143
+ "step": 6020
2144
+ },
2145
+ {
2146
+ "epoch": 0.49758347978910367,
2147
+ "grad_norm": 0.834241509437561,
2148
+ "learning_rate": 4.957159334321964e-05,
2149
+ "loss": 0.2193,
2150
+ "step": 6040
2151
+ },
2152
+ {
2153
+ "epoch": 0.4992311072056239,
2154
+ "grad_norm": 1.9017058610916138,
2155
+ "learning_rate": 4.9736365134289015e-05,
2156
+ "loss": 0.2338,
2157
+ "step": 6060
2158
+ },
2159
+ {
2160
+ "epoch": 0.5008787346221442,
2161
+ "grad_norm": 3.1669952869415283,
2162
+ "learning_rate": 4.990113692535838e-05,
2163
+ "loss": 0.2245,
2164
+ "step": 6080
2165
+ },
2166
+ {
2167
+ "epoch": 0.5025263620386643,
2168
+ "grad_norm": 2.9794270992279053,
2169
+ "learning_rate": 4.9999997353508944e-05,
2170
+ "loss": 0.239,
2171
+ "step": 6100
2172
+ },
2173
+ {
2174
+ "epoch": 0.5041739894551845,
2175
+ "grad_norm": 1.2019259929656982,
2176
+ "learning_rate": 4.999996758049098e-05,
2177
+ "loss": 0.1871,
2178
+ "step": 6120
2179
+ },
2180
+ {
2181
+ "epoch": 0.5058216168717048,
2182
+ "grad_norm": 2.175334930419922,
2183
+ "learning_rate": 4.999990472638076e-05,
2184
+ "loss": 0.2265,
2185
+ "step": 6140
2186
+ },
2187
+ {
2188
+ "epoch": 0.5074692442882249,
2189
+ "grad_norm": 1.4895635843276978,
2190
+ "learning_rate": 4.999980879126146e-05,
2191
+ "loss": 0.2259,
2192
+ "step": 6160
2193
+ },
2194
+ {
2195
+ "epoch": 0.5091168717047452,
2196
+ "grad_norm": 3.355076789855957,
2197
+ "learning_rate": 4.9999679775260015e-05,
2198
+ "loss": 0.2316,
2199
+ "step": 6180
2200
+ },
2201
+ {
2202
+ "epoch": 0.5107644991212654,
2203
+ "grad_norm": 3.545771598815918,
2204
+ "learning_rate": 4.999951767854715e-05,
2205
+ "loss": 0.2426,
2206
+ "step": 6200
2207
+ },
2208
+ {
2209
+ "epoch": 0.5124121265377856,
2210
+ "grad_norm": 2.085190773010254,
2211
+ "learning_rate": 4.999932250133736e-05,
2212
+ "loss": 0.1884,
2213
+ "step": 6220
2214
+ },
2215
+ {
2216
+ "epoch": 0.5140597539543058,
2217
+ "grad_norm": 1.6277663707733154,
2218
+ "learning_rate": 4.999909424388892e-05,
2219
+ "loss": 0.2332,
2220
+ "step": 6240
2221
+ },
2222
+ {
2223
+ "epoch": 0.5157073813708261,
2224
+ "grad_norm": 3.9234516620635986,
2225
+ "learning_rate": 4.9998832906503856e-05,
2226
+ "loss": 0.2346,
2227
+ "step": 6260
2228
+ },
2229
+ {
2230
+ "epoch": 0.5173550087873462,
2231
+ "grad_norm": 1.7027534246444702,
2232
+ "learning_rate": 4.9998538489527984e-05,
2233
+ "loss": 0.2339,
2234
+ "step": 6280
2235
+ },
2236
+ {
2237
+ "epoch": 0.5190026362038664,
2238
+ "grad_norm": 2.7195184230804443,
2239
+ "learning_rate": 4.9998210993350895e-05,
2240
+ "loss": 0.2511,
2241
+ "step": 6300
2242
+ },
2243
+ {
2244
+ "epoch": 0.5206502636203867,
2245
+ "grad_norm": 0.7267903089523315,
2246
+ "learning_rate": 4.9997850418405945e-05,
2247
+ "loss": 0.1765,
2248
+ "step": 6320
2249
+ },
2250
+ {
2251
+ "epoch": 0.5222978910369068,
2252
+ "grad_norm": 0.751308798789978,
2253
+ "learning_rate": 4.999745676517027e-05,
2254
+ "loss": 0.2393,
2255
+ "step": 6340
2256
+ },
2257
+ {
2258
+ "epoch": 0.5239455184534271,
2259
+ "grad_norm": 1.0236356258392334,
2260
+ "learning_rate": 4.999703003416476e-05,
2261
+ "loss": 0.2292,
2262
+ "step": 6360
2263
+ },
2264
+ {
2265
+ "epoch": 0.5255931458699473,
2266
+ "grad_norm": 0.7693842649459839,
2267
+ "learning_rate": 4.999657022595409e-05,
2268
+ "loss": 0.2285,
2269
+ "step": 6380
2270
+ },
2271
+ {
2272
+ "epoch": 0.5272407732864675,
2273
+ "grad_norm": 3.4103684425354004,
2274
+ "learning_rate": 4.999607734114669e-05,
2275
+ "loss": 0.2635,
2276
+ "step": 6400
2277
+ },
2278
+ {
2279
+ "epoch": 0.5288884007029877,
2280
+ "grad_norm": 0.5887142419815063,
2281
+ "learning_rate": 4.999555138039478e-05,
2282
+ "loss": 0.1938,
2283
+ "step": 6420
2284
+ },
2285
+ {
2286
+ "epoch": 0.530536028119508,
2287
+ "grad_norm": 1.6326848268508911,
2288
+ "learning_rate": 4.999499234439433e-05,
2289
+ "loss": 0.2112,
2290
+ "step": 6440
2291
+ },
2292
+ {
2293
+ "epoch": 0.5321836555360281,
2294
+ "grad_norm": 0.4003719985485077,
2295
+ "learning_rate": 4.9994400233885086e-05,
2296
+ "loss": 0.2011,
2297
+ "step": 6460
2298
+ },
2299
+ {
2300
+ "epoch": 0.5338312829525483,
2301
+ "grad_norm": 1.0633249282836914,
2302
+ "learning_rate": 4.999377504965055e-05,
2303
+ "loss": 0.2388,
2304
+ "step": 6480
2305
+ },
2306
+ {
2307
+ "epoch": 0.5354789103690686,
2308
+ "grad_norm": 3.87857985496521,
2309
+ "learning_rate": 4.999311679251799e-05,
2310
+ "loss": 0.2222,
2311
+ "step": 6500
2312
+ },
2313
+ {
2314
+ "epoch": 0.5371265377855887,
2315
+ "grad_norm": 1.1982243061065674,
2316
+ "learning_rate": 4.999242546335845e-05,
2317
+ "loss": 0.1736,
2318
+ "step": 6520
2319
+ },
2320
+ {
2321
+ "epoch": 0.538774165202109,
2322
+ "grad_norm": 1.4149129390716553,
2323
+ "learning_rate": 4.999170106308673e-05,
2324
+ "loss": 0.2044,
2325
+ "step": 6540
2326
+ },
2327
+ {
2328
+ "epoch": 0.5404217926186292,
2329
+ "grad_norm": 0.9540938138961792,
2330
+ "learning_rate": 4.999094359266138e-05,
2331
+ "loss": 0.2343,
2332
+ "step": 6560
2333
+ },
2334
+ {
2335
+ "epoch": 0.5420694200351494,
2336
+ "grad_norm": 0.8150781989097595,
2337
+ "learning_rate": 4.999015305308472e-05,
2338
+ "loss": 0.2027,
2339
+ "step": 6580
2340
+ },
2341
+ {
2342
+ "epoch": 0.5437170474516696,
2343
+ "grad_norm": 2.2410573959350586,
2344
+ "learning_rate": 4.998932944540284e-05,
2345
+ "loss": 0.2191,
2346
+ "step": 6600
2347
+ },
2348
+ {
2349
+ "epoch": 0.5453646748681898,
2350
+ "grad_norm": 2.924471616744995,
2351
+ "learning_rate": 4.998847277070556e-05,
2352
+ "loss": 0.1776,
2353
+ "step": 6620
2354
+ },
2355
+ {
2356
+ "epoch": 0.54701230228471,
2357
+ "grad_norm": 2.5102221965789795,
2358
+ "learning_rate": 4.9987583030126484e-05,
2359
+ "loss": 0.2172,
2360
+ "step": 6640
2361
+ },
2362
+ {
2363
+ "epoch": 0.5486599297012302,
2364
+ "grad_norm": 1.2676066160202026,
2365
+ "learning_rate": 4.998666022484295e-05,
2366
+ "loss": 0.2235,
2367
+ "step": 6660
2368
+ },
2369
+ {
2370
+ "epoch": 0.5503075571177505,
2371
+ "grad_norm": 1.7459567785263062,
2372
+ "learning_rate": 4.998570435607605e-05,
2373
+ "loss": 0.2114,
2374
+ "step": 6680
2375
+ },
2376
+ {
2377
+ "epoch": 0.5519551845342706,
2378
+ "grad_norm": 2.4994170665740967,
2379
+ "learning_rate": 4.998476565684759e-05,
2380
+ "loss": 0.2321,
2381
+ "step": 6700
2382
+ },
2383
+ {
2384
+ "epoch": 0.5536028119507909,
2385
+ "grad_norm": 1.436635136604309,
2386
+ "learning_rate": 4.998374531796601e-05,
2387
+ "loss": 0.1715,
2388
+ "step": 6720
2389
+ },
2390
+ {
2391
+ "epoch": 0.5552504393673111,
2392
+ "grad_norm": 5.170281410217285,
2393
+ "learning_rate": 4.9982691919458215e-05,
2394
+ "loss": 0.2227,
2395
+ "step": 6740
2396
+ },
2397
+ {
2398
+ "epoch": 0.5568980667838312,
2399
+ "grad_norm": 2.329806327819824,
2400
+ "learning_rate": 4.9981605462718097e-05,
2401
+ "loss": 0.2103,
2402
+ "step": 6760
2403
+ },
2404
+ {
2405
+ "epoch": 0.5585456942003515,
2406
+ "grad_norm": 0.5105818510055542,
2407
+ "learning_rate": 4.998048594918331e-05,
2408
+ "loss": 0.2297,
2409
+ "step": 6780
2410
+ },
2411
+ {
2412
+ "epoch": 0.5601933216168717,
2413
+ "grad_norm": 2.74815034866333,
2414
+ "learning_rate": 4.997933338033525e-05,
2415
+ "loss": 0.2454,
2416
+ "step": 6800
2417
+ },
2418
+ {
2419
+ "epoch": 0.5618409490333919,
2420
+ "grad_norm": 1.0433367490768433,
2421
+ "learning_rate": 4.997814775769904e-05,
2422
+ "loss": 0.175,
2423
+ "step": 6820
2424
+ },
2425
+ {
2426
+ "epoch": 0.5634885764499121,
2427
+ "grad_norm": 0.5902445316314697,
2428
+ "learning_rate": 4.997692908284356e-05,
2429
+ "loss": 0.2153,
2430
+ "step": 6840
2431
+ },
2432
+ {
2433
+ "epoch": 0.5651362038664324,
2434
+ "grad_norm": 0.9332594275474548,
2435
+ "learning_rate": 4.997567735738141e-05,
2436
+ "loss": 0.2246,
2437
+ "step": 6860
2438
+ },
2439
+ {
2440
+ "epoch": 0.5667838312829525,
2441
+ "grad_norm": 1.2199161052703857,
2442
+ "learning_rate": 4.9974392582968934e-05,
2443
+ "loss": 0.2281,
2444
+ "step": 6880
2445
+ },
2446
+ {
2447
+ "epoch": 0.5684314586994728,
2448
+ "grad_norm": 2.2492709159851074,
2449
+ "learning_rate": 4.9973074761306186e-05,
2450
+ "loss": 0.238,
2451
+ "step": 6900
2452
+ },
2453
+ {
2454
+ "epoch": 0.570079086115993,
2455
+ "grad_norm": 9.928802490234375,
2456
+ "learning_rate": 4.997172389413699e-05,
2457
+ "loss": 0.1731,
2458
+ "step": 6920
2459
+ },
2460
+ {
2461
+ "epoch": 0.5717267135325131,
2462
+ "grad_norm": 1.4187493324279785,
2463
+ "learning_rate": 4.997033998324886e-05,
2464
+ "loss": 0.2267,
2465
+ "step": 6940
2466
+ },
2467
+ {
2468
+ "epoch": 0.5733743409490334,
2469
+ "grad_norm": 1.0871655941009521,
2470
+ "learning_rate": 4.996892303047306e-05,
2471
+ "loss": 0.2478,
2472
+ "step": 6960
2473
+ },
2474
+ {
2475
+ "epoch": 0.5750219683655536,
2476
+ "grad_norm": 0.9939442873001099,
2477
+ "learning_rate": 4.996747303768456e-05,
2478
+ "loss": 0.1992,
2479
+ "step": 6980
2480
+ },
2481
+ {
2482
+ "epoch": 0.5766695957820738,
2483
+ "grad_norm": 48.66152572631836,
2484
+ "learning_rate": 4.996599000680206e-05,
2485
+ "loss": 0.2341,
2486
+ "step": 7000
2487
+ },
2488
+ {
2489
+ "epoch": 0.578317223198594,
2490
+ "grad_norm": 0.8890938758850098,
2491
+ "learning_rate": 4.996447393978797e-05,
2492
+ "loss": 0.1687,
2493
+ "step": 7020
2494
+ },
2495
+ {
2496
+ "epoch": 0.5799648506151143,
2497
+ "grad_norm": 0.8341367840766907,
2498
+ "learning_rate": 4.996292483864843e-05,
2499
+ "loss": 0.2175,
2500
+ "step": 7040
2501
+ },
2502
+ {
2503
+ "epoch": 0.5816124780316344,
2504
+ "grad_norm": 1.6562838554382324,
2505
+ "learning_rate": 4.996134270543326e-05,
2506
+ "loss": 0.2136,
2507
+ "step": 7060
2508
+ },
2509
+ {
2510
+ "epoch": 0.5832601054481547,
2511
+ "grad_norm": 0.6537133455276489,
2512
+ "learning_rate": 4.9959727542236025e-05,
2513
+ "loss": 0.2139,
2514
+ "step": 7080
2515
+ },
2516
+ {
2517
+ "epoch": 0.5849077328646749,
2518
+ "grad_norm": 2.101271390914917,
2519
+ "learning_rate": 4.9958079351193976e-05,
2520
+ "loss": 0.2359,
2521
+ "step": 7100
2522
+ },
2523
+ {
2524
+ "epoch": 0.586555360281195,
2525
+ "grad_norm": 0.9616307616233826,
2526
+ "learning_rate": 4.995639813448808e-05,
2527
+ "loss": 0.1774,
2528
+ "step": 7120
2529
+ },
2530
+ {
2531
+ "epoch": 0.5882029876977153,
2532
+ "grad_norm": 1.0992920398712158,
2533
+ "learning_rate": 4.9954683894343e-05,
2534
+ "loss": 0.1962,
2535
+ "step": 7140
2536
+ },
2537
+ {
2538
+ "epoch": 0.5898506151142355,
2539
+ "grad_norm": 0.6329948902130127,
2540
+ "learning_rate": 4.995293663302709e-05,
2541
+ "loss": 0.2145,
2542
+ "step": 7160
2543
+ },
2544
+ {
2545
+ "epoch": 0.5914982425307557,
2546
+ "grad_norm": 1.0156137943267822,
2547
+ "learning_rate": 4.9951156352852415e-05,
2548
+ "loss": 0.2161,
2549
+ "step": 7180
2550
+ },
2551
+ {
2552
+ "epoch": 0.5931458699472759,
2553
+ "grad_norm": 2.6131973266601562,
2554
+ "learning_rate": 4.994943450511368e-05,
2555
+ "loss": 0.2234,
2556
+ "step": 7200
2557
+ },
2558
+ {
2559
+ "epoch": 0.5947934973637962,
2560
+ "grad_norm": 3.4186134338378906,
2561
+ "learning_rate": 4.99475898449799e-05,
2562
+ "loss": 0.171,
2563
+ "step": 7220
2564
+ },
2565
+ {
2566
+ "epoch": 0.5964411247803163,
2567
+ "grad_norm": 4.896924018859863,
2568
+ "learning_rate": 4.9945712173062477e-05,
2569
+ "loss": 0.2073,
2570
+ "step": 7240
2571
+ },
2572
+ {
2573
+ "epoch": 0.5980887521968365,
2574
+ "grad_norm": 1.2636560201644897,
2575
+ "learning_rate": 4.994380149184601e-05,
2576
+ "loss": 0.2144,
2577
+ "step": 7260
2578
+ },
2579
+ {
2580
+ "epoch": 0.5997363796133568,
2581
+ "grad_norm": 1.4861242771148682,
2582
+ "learning_rate": 4.99418578038588e-05,
2583
+ "loss": 0.2024,
2584
+ "step": 7280
2585
+ },
2586
+ {
2587
+ "epoch": 0.6013840070298769,
2588
+ "grad_norm": 6.527903079986572,
2589
+ "learning_rate": 4.993988111167284e-05,
2590
+ "loss": 0.2172,
2591
+ "step": 7300
2592
+ },
2593
+ {
2594
+ "epoch": 0.6030316344463972,
2595
+ "grad_norm": 1.9219621419906616,
2596
+ "learning_rate": 4.993787141790375e-05,
2597
+ "loss": 0.1865,
2598
+ "step": 7320
2599
+ },
2600
+ {
2601
+ "epoch": 0.6046792618629174,
2602
+ "grad_norm": 0.7625167369842529,
2603
+ "learning_rate": 4.9935828725210874e-05,
2604
+ "loss": 0.2238,
2605
+ "step": 7340
2606
+ },
2607
+ {
2608
+ "epoch": 0.6063268892794376,
2609
+ "grad_norm": 1.3375577926635742,
2610
+ "learning_rate": 4.9933753036297196e-05,
2611
+ "loss": 0.1878,
2612
+ "step": 7360
2613
+ },
2614
+ {
2615
+ "epoch": 0.6079745166959578,
2616
+ "grad_norm": 1.1067700386047363,
2617
+ "learning_rate": 4.993164435390935e-05,
2618
+ "loss": 0.2111,
2619
+ "step": 7380
2620
+ },
2621
+ {
2622
+ "epoch": 0.6096221441124781,
2623
+ "grad_norm": 4.094808578491211,
2624
+ "learning_rate": 4.992950268083764e-05,
2625
+ "loss": 0.2073,
2626
+ "step": 7400
2627
+ },
2628
+ {
2629
+ "epoch": 0.6112697715289982,
2630
+ "grad_norm": 2.683678150177002,
2631
+ "learning_rate": 4.992732801991602e-05,
2632
+ "loss": 0.172,
2633
+ "step": 7420
2634
+ },
2635
+ {
2636
+ "epoch": 0.6129173989455184,
2637
+ "grad_norm": 0.9752517342567444,
2638
+ "learning_rate": 4.992512037402212e-05,
2639
+ "loss": 0.1924,
2640
+ "step": 7440
2641
+ },
2642
+ {
2643
+ "epoch": 0.6145650263620387,
2644
+ "grad_norm": 0.8463476896286011,
2645
+ "learning_rate": 4.9922879746077176e-05,
2646
+ "loss": 0.2021,
2647
+ "step": 7460
2648
+ },
2649
+ {
2650
+ "epoch": 0.6162126537785588,
2651
+ "grad_norm": 3.053812026977539,
2652
+ "learning_rate": 4.992060613904611e-05,
2653
+ "loss": 0.2086,
2654
+ "step": 7480
2655
+ },
2656
+ {
2657
+ "epoch": 0.6178602811950791,
2658
+ "grad_norm": 2.9483275413513184,
2659
+ "learning_rate": 4.991829955593744e-05,
2660
+ "loss": 0.2154,
2661
+ "step": 7500
2662
+ },
2663
+ {
2664
+ "epoch": 0.6195079086115993,
2665
+ "grad_norm": 1.3794467449188232,
2666
+ "learning_rate": 4.9915959999803365e-05,
2667
+ "loss": 0.1599,
2668
+ "step": 7520
2669
+ },
2670
+ {
2671
+ "epoch": 0.6211555360281195,
2672
+ "grad_norm": 1.0529364347457886,
2673
+ "learning_rate": 4.9913587473739666e-05,
2674
+ "loss": 0.1891,
2675
+ "step": 7540
2676
+ },
2677
+ {
2678
+ "epoch": 0.6228031634446397,
2679
+ "grad_norm": 0.9026776552200317,
2680
+ "learning_rate": 4.991118198088579e-05,
2681
+ "loss": 0.2074,
2682
+ "step": 7560
2683
+ },
2684
+ {
2685
+ "epoch": 0.62445079086116,
2686
+ "grad_norm": 0.4504956603050232,
2687
+ "learning_rate": 4.9908743524424806e-05,
2688
+ "loss": 0.1931,
2689
+ "step": 7580
2690
+ },
2691
+ {
2692
+ "epoch": 0.6260984182776801,
2693
+ "grad_norm": 4.0763726234436035,
2694
+ "learning_rate": 4.9906272107583366e-05,
2695
+ "loss": 0.2192,
2696
+ "step": 7600
2697
+ },
2698
+ {
2699
+ "epoch": 0.6277460456942003,
2700
+ "grad_norm": 1.012601375579834,
2701
+ "learning_rate": 4.990376773363178e-05,
2702
+ "loss": 0.1604,
2703
+ "step": 7620
2704
+ },
2705
+ {
2706
+ "epoch": 0.6293936731107206,
2707
+ "grad_norm": 0.6183291673660278,
2708
+ "learning_rate": 4.990123040588395e-05,
2709
+ "loss": 0.2015,
2710
+ "step": 7640
2711
+ },
2712
+ {
2713
+ "epoch": 0.6310413005272407,
2714
+ "grad_norm": 0.718015730381012,
2715
+ "learning_rate": 4.989866012769736e-05,
2716
+ "loss": 0.2155,
2717
+ "step": 7660
2718
+ },
2719
+ {
2720
+ "epoch": 0.632688927943761,
2721
+ "grad_norm": 1.5611259937286377,
2722
+ "learning_rate": 4.989605690247315e-05,
2723
+ "loss": 0.1975,
2724
+ "step": 7680
2725
+ },
2726
+ {
2727
+ "epoch": 0.6343365553602812,
2728
+ "grad_norm": 3.2936527729034424,
2729
+ "learning_rate": 4.9893420733656e-05,
2730
+ "loss": 0.2305,
2731
+ "step": 7700
2732
+ },
2733
+ {
2734
+ "epoch": 0.6359841827768014,
2735
+ "grad_norm": 1.3771491050720215,
2736
+ "learning_rate": 4.9890751624734225e-05,
2737
+ "loss": 0.1575,
2738
+ "step": 7720
2739
+ },
2740
+ {
2741
+ "epoch": 0.6376318101933216,
2742
+ "grad_norm": 1.0577577352523804,
2743
+ "learning_rate": 4.98880495792397e-05,
2744
+ "loss": 0.2045,
2745
+ "step": 7740
2746
+ },
2747
+ {
2748
+ "epoch": 0.6392794376098418,
2749
+ "grad_norm": 3.6020660400390625,
2750
+ "learning_rate": 4.988531460074791e-05,
2751
+ "loss": 0.2034,
2752
+ "step": 7760
2753
+ },
2754
+ {
2755
+ "epoch": 0.640927065026362,
2756
+ "grad_norm": 0.7804440855979919,
2757
+ "learning_rate": 4.9882546692877885e-05,
2758
+ "loss": 0.1953,
2759
+ "step": 7780
2760
+ },
2761
+ {
2762
+ "epoch": 0.6425746924428822,
2763
+ "grad_norm": 57.011390686035156,
2764
+ "learning_rate": 4.987974585929226e-05,
2765
+ "loss": 0.2207,
2766
+ "step": 7800
2767
+ },
2768
+ {
2769
+ "epoch": 0.6442223198594025,
2770
+ "grad_norm": 1.5725603103637695,
2771
+ "learning_rate": 4.987691210369721e-05,
2772
+ "loss": 0.1664,
2773
+ "step": 7820
2774
+ },
2775
+ {
2776
+ "epoch": 0.6458699472759226,
2777
+ "grad_norm": 0.6347635984420776,
2778
+ "learning_rate": 4.98740454298425e-05,
2779
+ "loss": 0.2134,
2780
+ "step": 7840
2781
+ },
2782
+ {
2783
+ "epoch": 0.6475175746924429,
2784
+ "grad_norm": 1.3581891059875488,
2785
+ "learning_rate": 4.987114584152145e-05,
2786
+ "loss": 0.217,
2787
+ "step": 7860
2788
+ },
2789
+ {
2790
+ "epoch": 0.6491652021089631,
2791
+ "grad_norm": 5.659873008728027,
2792
+ "learning_rate": 4.986821334257091e-05,
2793
+ "loss": 0.1977,
2794
+ "step": 7880
2795
+ },
2796
+ {
2797
+ "epoch": 0.6508128295254832,
2798
+ "grad_norm": 11.7850980758667,
2799
+ "learning_rate": 4.986524793687131e-05,
2800
+ "loss": 0.218,
2801
+ "step": 7900
2802
+ },
2803
+ {
2804
+ "epoch": 0.6524604569420035,
2805
+ "grad_norm": 1.2856826782226562,
2806
+ "learning_rate": 4.986224962834659e-05,
2807
+ "loss": 0.1675,
2808
+ "step": 7920
2809
+ },
2810
+ {
2811
+ "epoch": 0.6541080843585237,
2812
+ "grad_norm": 1.055503249168396,
2813
+ "learning_rate": 4.985921842096427e-05,
2814
+ "loss": 0.2113,
2815
+ "step": 7940
2816
+ },
2817
+ {
2818
+ "epoch": 0.6557557117750439,
2819
+ "grad_norm": 1.5587449073791504,
2820
+ "learning_rate": 4.9856154318735374e-05,
2821
+ "loss": 0.1939,
2822
+ "step": 7960
2823
+ },
2824
+ {
2825
+ "epoch": 0.6574033391915641,
2826
+ "grad_norm": 2.40641450881958,
2827
+ "learning_rate": 4.985305732571446e-05,
2828
+ "loss": 0.206,
2829
+ "step": 7980
2830
+ },
2831
+ {
2832
+ "epoch": 0.6590509666080844,
2833
+ "grad_norm": 4.06686544418335,
2834
+ "learning_rate": 4.98499274459996e-05,
2835
+ "loss": 0.2107,
2836
+ "step": 8000
2837
+ },
2838
+ {
2839
+ "epoch": 0.6590509666080844,
2840
+ "eval_loss": 0.5655311346054077,
2841
+ "eval_runtime": 835.4807,
2842
+ "eval_samples_per_second": 26.08,
2843
+ "eval_steps_per_second": 6.521,
2844
+ "eval_wer": 0.2457764477651215,
2845
+ "step": 8000
2846
+ },
2847
+ {
2848
+ "epoch": 0.6606985940246046,
2849
+ "grad_norm": 0.8585990071296692,
2850
+ "learning_rate": 4.984676468373241e-05,
2851
+ "loss": 0.1667,
2852
+ "step": 8020
2853
+ },
2854
+ {
2855
+ "epoch": 0.6623462214411248,
2856
+ "grad_norm": 1.2851459980010986,
2857
+ "learning_rate": 4.984356904309799e-05,
2858
+ "loss": 0.2184,
2859
+ "step": 8040
2860
+ },
2861
+ {
2862
+ "epoch": 0.663993848857645,
2863
+ "grad_norm": 0.8445897698402405,
2864
+ "learning_rate": 4.984034052832496e-05,
2865
+ "loss": 0.2169,
2866
+ "step": 8060
2867
+ },
2868
+ {
2869
+ "epoch": 0.6656414762741653,
2870
+ "grad_norm": 1.4265161752700806,
2871
+ "learning_rate": 4.983707914368544e-05,
2872
+ "loss": 0.2004,
2873
+ "step": 8080
2874
+ },
2875
+ {
2876
+ "epoch": 0.6672891036906854,
2877
+ "grad_norm": 3.0620169639587402,
2878
+ "learning_rate": 4.983378489349504e-05,
2879
+ "loss": 0.2467,
2880
+ "step": 8100
2881
+ },
2882
+ {
2883
+ "epoch": 0.6689367311072056,
2884
+ "grad_norm": 1.1079747676849365,
2885
+ "learning_rate": 4.983045778211286e-05,
2886
+ "loss": 0.1587,
2887
+ "step": 8120
2888
+ },
2889
+ {
2890
+ "epoch": 0.6705843585237259,
2891
+ "grad_norm": 0.8565286993980408,
2892
+ "learning_rate": 4.982709781394148e-05,
2893
+ "loss": 0.2101,
2894
+ "step": 8140
2895
+ },
2896
+ {
2897
+ "epoch": 0.672231985940246,
2898
+ "grad_norm": 2.957345962524414,
2899
+ "learning_rate": 4.982370499342698e-05,
2900
+ "loss": 0.212,
2901
+ "step": 8160
2902
+ },
2903
+ {
2904
+ "epoch": 0.6738796133567663,
2905
+ "grad_norm": 1.744805932044983,
2906
+ "learning_rate": 4.982027932505887e-05,
2907
+ "loss": 0.209,
2908
+ "step": 8180
2909
+ },
2910
+ {
2911
+ "epoch": 0.6755272407732865,
2912
+ "grad_norm": 3.963977336883545,
2913
+ "learning_rate": 4.9816820813370166e-05,
2914
+ "loss": 0.214,
2915
+ "step": 8200
2916
+ },
2917
+ {
2918
+ "epoch": 0.6771748681898067,
2919
+ "grad_norm": 1.3982582092285156,
2920
+ "learning_rate": 4.981332946293733e-05,
2921
+ "loss": 0.1671,
2922
+ "step": 8220
2923
+ },
2924
+ {
2925
+ "epoch": 0.6788224956063269,
2926
+ "grad_norm": 0.5959907174110413,
2927
+ "learning_rate": 4.9809805278380264e-05,
2928
+ "loss": 0.2061,
2929
+ "step": 8240
2930
+ },
2931
+ {
2932
+ "epoch": 0.6804701230228472,
2933
+ "grad_norm": 1.2237777709960938,
2934
+ "learning_rate": 4.980624826436233e-05,
2935
+ "loss": 0.2188,
2936
+ "step": 8260
2937
+ },
2938
+ {
2939
+ "epoch": 0.6821177504393673,
2940
+ "grad_norm": 2.6652753353118896,
2941
+ "learning_rate": 4.9802658425590344e-05,
2942
+ "loss": 0.1964,
2943
+ "step": 8280
2944
+ },
2945
+ {
2946
+ "epoch": 0.6837653778558875,
2947
+ "grad_norm": 31.890893936157227,
2948
+ "learning_rate": 4.979903576681453e-05,
2949
+ "loss": 0.2133,
2950
+ "step": 8300
2951
+ },
2952
+ {
2953
+ "epoch": 0.6854130052724078,
2954
+ "grad_norm": 1.1191130876541138,
2955
+ "learning_rate": 4.979538029282855e-05,
2956
+ "loss": 0.1623,
2957
+ "step": 8320
2958
+ },
2959
+ {
2960
+ "epoch": 0.6870606326889279,
2961
+ "grad_norm": 1.4404329061508179,
2962
+ "learning_rate": 4.9791692008469514e-05,
2963
+ "loss": 0.2087,
2964
+ "step": 8340
2965
+ },
2966
+ {
2967
+ "epoch": 0.6887082601054482,
2968
+ "grad_norm": 1.4656809568405151,
2969
+ "learning_rate": 4.9787970918617914e-05,
2970
+ "loss": 0.2134,
2971
+ "step": 8360
2972
+ },
2973
+ {
2974
+ "epoch": 0.6903558875219684,
2975
+ "grad_norm": 2.310316801071167,
2976
+ "learning_rate": 4.978421702819767e-05,
2977
+ "loss": 0.215,
2978
+ "step": 8380
2979
+ },
2980
+ {
2981
+ "epoch": 0.6920035149384886,
2982
+ "grad_norm": 21.141889572143555,
2983
+ "learning_rate": 4.978043034217609e-05,
2984
+ "loss": 0.2206,
2985
+ "step": 8400
2986
+ },
2987
+ {
2988
+ "epoch": 0.6936511423550088,
2989
+ "grad_norm": 1.2799972295761108,
2990
+ "learning_rate": 4.977680261809319e-05,
2991
+ "loss": 0.1553,
2992
+ "step": 8420
2993
+ },
2994
+ {
2995
+ "epoch": 0.695298769771529,
2996
+ "grad_norm": 0.927029550075531,
2997
+ "learning_rate": 4.97729519951006e-05,
2998
+ "loss": 0.2205,
2999
+ "step": 8440
3000
+ },
3001
+ {
3002
+ "epoch": 0.6969463971880492,
3003
+ "grad_norm": 1.8685028553009033,
3004
+ "learning_rate": 4.976906859141309e-05,
3005
+ "loss": 0.1938,
3006
+ "step": 8460
3007
+ },
3008
+ {
3009
+ "epoch": 0.6985940246045694,
3010
+ "grad_norm": 0.38425010442733765,
3011
+ "learning_rate": 4.976515241216936e-05,
3012
+ "loss": 0.1964,
3013
+ "step": 8480
3014
+ },
3015
+ {
3016
+ "epoch": 0.7002416520210897,
3017
+ "grad_norm": 2.6303980350494385,
3018
+ "learning_rate": 4.976120346255146e-05,
3019
+ "loss": 0.1984,
3020
+ "step": 8500
3021
+ },
3022
+ {
3023
+ "epoch": 0.7018892794376098,
3024
+ "grad_norm": 7.791947841644287,
3025
+ "learning_rate": 4.975722174778482e-05,
3026
+ "loss": 0.1678,
3027
+ "step": 8520
3028
+ },
3029
+ {
3030
+ "epoch": 0.7035369068541301,
3031
+ "grad_norm": 1.193328857421875,
3032
+ "learning_rate": 4.9753207273138245e-05,
3033
+ "loss": 0.2182,
3034
+ "step": 8540
3035
+ },
3036
+ {
3037
+ "epoch": 0.7051845342706503,
3038
+ "grad_norm": 1.1064016819000244,
3039
+ "learning_rate": 4.974916004392385e-05,
3040
+ "loss": 0.2065,
3041
+ "step": 8560
3042
+ },
3043
+ {
3044
+ "epoch": 0.7068321616871704,
3045
+ "grad_norm": 3.304832696914673,
3046
+ "learning_rate": 4.974508006549711e-05,
3047
+ "loss": 0.1872,
3048
+ "step": 8580
3049
+ },
3050
+ {
3051
+ "epoch": 0.7084797891036907,
3052
+ "grad_norm": 2.5478882789611816,
3053
+ "learning_rate": 4.974096734325686e-05,
3054
+ "loss": 0.2295,
3055
+ "step": 8600
3056
+ },
3057
+ {
3058
+ "epoch": 0.7101274165202109,
3059
+ "grad_norm": 14.596063613891602,
3060
+ "learning_rate": 4.9736821882645226e-05,
3061
+ "loss": 0.1628,
3062
+ "step": 8620
3063
+ },
3064
+ {
3065
+ "epoch": 0.7117750439367311,
3066
+ "grad_norm": 0.9999619126319885,
3067
+ "learning_rate": 4.973264368914766e-05,
3068
+ "loss": 0.2107,
3069
+ "step": 8640
3070
+ },
3071
+ {
3072
+ "epoch": 0.7134226713532513,
3073
+ "grad_norm": 1.053412675857544,
3074
+ "learning_rate": 4.972843276829296e-05,
3075
+ "loss": 0.2085,
3076
+ "step": 8660
3077
+ },
3078
+ {
3079
+ "epoch": 0.7150702987697716,
3080
+ "grad_norm": 2.1676464080810547,
3081
+ "learning_rate": 4.9724189125653195e-05,
3082
+ "loss": 0.2048,
3083
+ "step": 8680
3084
+ },
3085
+ {
3086
+ "epoch": 0.7167179261862917,
3087
+ "grad_norm": 10.32970142364502,
3088
+ "learning_rate": 4.9719912766843746e-05,
3089
+ "loss": 0.2224,
3090
+ "step": 8700
3091
+ },
3092
+ {
3093
+ "epoch": 0.718365553602812,
3094
+ "grad_norm": 1.9829623699188232,
3095
+ "learning_rate": 4.971560369752328e-05,
3096
+ "loss": 0.1686,
3097
+ "step": 8720
3098
+ },
3099
+ {
3100
+ "epoch": 0.7200131810193322,
3101
+ "grad_norm": 2.7383005619049072,
3102
+ "learning_rate": 4.971126192339377e-05,
3103
+ "loss": 0.2088,
3104
+ "step": 8740
3105
+ },
3106
+ {
3107
+ "epoch": 0.7216608084358523,
3108
+ "grad_norm": 0.6413734555244446,
3109
+ "learning_rate": 4.970688745020043e-05,
3110
+ "loss": 0.2111,
3111
+ "step": 8760
3112
+ },
3113
+ {
3114
+ "epoch": 0.7233084358523726,
3115
+ "grad_norm": 0.8469798564910889,
3116
+ "learning_rate": 4.970248028373178e-05,
3117
+ "loss": 0.2057,
3118
+ "step": 8780
3119
+ },
3120
+ {
3121
+ "epoch": 0.7249560632688928,
3122
+ "grad_norm": 8.44117259979248,
3123
+ "learning_rate": 4.969804042981956e-05,
3124
+ "loss": 0.2179,
3125
+ "step": 8800
3126
+ },
3127
+ {
3128
+ "epoch": 0.726603690685413,
3129
+ "grad_norm": 1.40958571434021,
3130
+ "learning_rate": 4.969356789433881e-05,
3131
+ "loss": 0.1604,
3132
+ "step": 8820
3133
+ },
3134
+ {
3135
+ "epoch": 0.7282513181019332,
3136
+ "grad_norm": 5.328885078430176,
3137
+ "learning_rate": 4.968906268320777e-05,
3138
+ "loss": 0.2075,
3139
+ "step": 8840
3140
+ },
3141
+ {
3142
+ "epoch": 0.7298989455184535,
3143
+ "grad_norm": 0.6396345496177673,
3144
+ "learning_rate": 4.9684524802387956e-05,
3145
+ "loss": 0.2057,
3146
+ "step": 8860
3147
+ },
3148
+ {
3149
+ "epoch": 0.7315465729349736,
3150
+ "grad_norm": 0.7661327123641968,
3151
+ "learning_rate": 4.967995425788409e-05,
3152
+ "loss": 0.2064,
3153
+ "step": 8880
3154
+ },
3155
+ {
3156
+ "epoch": 0.7331942003514939,
3157
+ "grad_norm": 2.8300130367279053,
3158
+ "learning_rate": 4.9675351055744134e-05,
3159
+ "loss": 0.2109,
3160
+ "step": 8900
3161
+ },
3162
+ {
3163
+ "epoch": 0.7348418277680141,
3164
+ "grad_norm": 1.3380045890808105,
3165
+ "learning_rate": 4.9670715202059235e-05,
3166
+ "loss": 0.1492,
3167
+ "step": 8920
3168
+ },
3169
+ {
3170
+ "epoch": 0.7364894551845342,
3171
+ "grad_norm": 1.478825569152832,
3172
+ "learning_rate": 4.9666046702963784e-05,
3173
+ "loss": 0.2133,
3174
+ "step": 8940
3175
+ },
3176
+ {
3177
+ "epoch": 0.7381370826010545,
3178
+ "grad_norm": 1.3596333265304565,
3179
+ "learning_rate": 4.9661345564635356e-05,
3180
+ "loss": 0.1999,
3181
+ "step": 8960
3182
+ },
3183
+ {
3184
+ "epoch": 0.7397847100175747,
3185
+ "grad_norm": 0.3950871229171753,
3186
+ "learning_rate": 4.965661179329468e-05,
3187
+ "loss": 0.1925,
3188
+ "step": 8980
3189
+ },
3190
+ {
3191
+ "epoch": 0.7414323374340949,
3192
+ "grad_norm": 3.47658634185791,
3193
+ "learning_rate": 4.965184539520572e-05,
3194
+ "loss": 0.2016,
3195
+ "step": 9000
3196
+ },
3197
+ {
3198
+ "epoch": 0.7430799648506151,
3199
+ "grad_norm": 1.6241281032562256,
3200
+ "learning_rate": 4.9647046376675586e-05,
3201
+ "loss": 0.1522,
3202
+ "step": 9020
3203
+ },
3204
+ {
3205
+ "epoch": 0.7447275922671354,
3206
+ "grad_norm": 0.582032322883606,
3207
+ "learning_rate": 4.964221474405456e-05,
3208
+ "loss": 0.2161,
3209
+ "step": 9040
3210
+ },
3211
+ {
3212
+ "epoch": 0.7463752196836555,
3213
+ "grad_norm": 1.1760663986206055,
3214
+ "learning_rate": 4.963735050373608e-05,
3215
+ "loss": 0.1986,
3216
+ "step": 9060
3217
+ },
3218
+ {
3219
+ "epoch": 0.7480228471001757,
3220
+ "grad_norm": 1.9206633567810059,
3221
+ "learning_rate": 4.963245366215672e-05,
3222
+ "loss": 0.195,
3223
+ "step": 9080
3224
+ },
3225
+ {
3226
+ "epoch": 0.749670474516696,
3227
+ "grad_norm": 2.0300755500793457,
3228
+ "learning_rate": 4.9627524225796206e-05,
3229
+ "loss": 0.2044,
3230
+ "step": 9100
3231
+ },
3232
+ {
3233
+ "epoch": 0.7513181019332161,
3234
+ "grad_norm": 0.91159987449646,
3235
+ "learning_rate": 4.962256220117739e-05,
3236
+ "loss": 0.156,
3237
+ "step": 9120
3238
+ },
3239
+ {
3240
+ "epoch": 0.7529657293497364,
3241
+ "grad_norm": 0.6553214192390442,
3242
+ "learning_rate": 4.961756759486625e-05,
3243
+ "loss": 0.1918,
3244
+ "step": 9140
3245
+ },
3246
+ {
3247
+ "epoch": 0.7546133567662566,
3248
+ "grad_norm": 2.1197195053100586,
3249
+ "learning_rate": 4.961254041347189e-05,
3250
+ "loss": 0.1942,
3251
+ "step": 9160
3252
+ },
3253
+ {
3254
+ "epoch": 0.7562609841827768,
3255
+ "grad_norm": 0.5275347232818604,
3256
+ "learning_rate": 4.9607480663646487e-05,
3257
+ "loss": 0.1975,
3258
+ "step": 9180
3259
+ },
3260
+ {
3261
+ "epoch": 0.757908611599297,
3262
+ "grad_norm": 2.276308536529541,
3263
+ "learning_rate": 4.9602388352085337e-05,
3264
+ "loss": 0.2035,
3265
+ "step": 9200
3266
+ },
3267
+ {
3268
+ "epoch": 0.7595562390158173,
3269
+ "grad_norm": 1.4071913957595825,
3270
+ "learning_rate": 4.9597263485526826e-05,
3271
+ "loss": 0.1636,
3272
+ "step": 9220
3273
+ },
3274
+ {
3275
+ "epoch": 0.7612038664323374,
3276
+ "grad_norm": 5.9329023361206055,
3277
+ "learning_rate": 4.959210607075239e-05,
3278
+ "loss": 0.1954,
3279
+ "step": 9240
3280
+ },
3281
+ {
3282
+ "epoch": 0.7628514938488576,
3283
+ "grad_norm": 1.0558881759643555,
3284
+ "learning_rate": 4.958691611458657e-05,
3285
+ "loss": 0.2089,
3286
+ "step": 9260
3287
+ },
3288
+ {
3289
+ "epoch": 0.7644991212653779,
3290
+ "grad_norm": 1.1962648630142212,
3291
+ "learning_rate": 4.958169362389695e-05,
3292
+ "loss": 0.1966,
3293
+ "step": 9280
3294
+ },
3295
+ {
3296
+ "epoch": 0.766146748681898,
3297
+ "grad_norm": 27.050457000732422,
3298
+ "learning_rate": 4.957643860559417e-05,
3299
+ "loss": 0.2254,
3300
+ "step": 9300
3301
+ },
3302
+ {
3303
+ "epoch": 0.7677943760984183,
3304
+ "grad_norm": 1.5262819528579712,
3305
+ "learning_rate": 4.95711510666319e-05,
3306
+ "loss": 0.1672,
3307
+ "step": 9320
3308
+ },
3309
+ {
3310
+ "epoch": 0.7694420035149385,
3311
+ "grad_norm": 1.5432629585266113,
3312
+ "learning_rate": 4.956583101400685e-05,
3313
+ "loss": 0.2201,
3314
+ "step": 9340
3315
+ },
3316
+ {
3317
+ "epoch": 0.7710896309314587,
3318
+ "grad_norm": 1.4080497026443481,
3319
+ "learning_rate": 4.956047845475877e-05,
3320
+ "loss": 0.2034,
3321
+ "step": 9360
3322
+ },
3323
+ {
3324
+ "epoch": 0.7727372583479789,
3325
+ "grad_norm": 1.6339974403381348,
3326
+ "learning_rate": 4.9555093395970396e-05,
3327
+ "loss": 0.1918,
3328
+ "step": 9380
3329
+ },
3330
+ {
3331
+ "epoch": 0.7743848857644992,
3332
+ "grad_norm": 3.812486410140991,
3333
+ "learning_rate": 4.954967584476748e-05,
3334
+ "loss": 0.2221,
3335
+ "step": 9400
3336
+ },
3337
+ {
3338
+ "epoch": 0.7760325131810193,
3339
+ "grad_norm": 2.0563764572143555,
3340
+ "learning_rate": 4.954422580831879e-05,
3341
+ "loss": 0.1611,
3342
+ "step": 9420
3343
+ },
3344
+ {
3345
+ "epoch": 0.7776801405975395,
3346
+ "grad_norm": 0.7051644325256348,
3347
+ "learning_rate": 4.9538743293836046e-05,
3348
+ "loss": 0.2053,
3349
+ "step": 9440
3350
+ },
3351
+ {
3352
+ "epoch": 0.7793277680140598,
3353
+ "grad_norm": 0.837563157081604,
3354
+ "learning_rate": 4.9533228308573966e-05,
3355
+ "loss": 0.2078,
3356
+ "step": 9460
3357
+ },
3358
+ {
3359
+ "epoch": 0.7809753954305799,
3360
+ "grad_norm": 1.1410398483276367,
3361
+ "learning_rate": 4.952768085983023e-05,
3362
+ "loss": 0.2071,
3363
+ "step": 9480
3364
+ },
3365
+ {
3366
+ "epoch": 0.7826230228471002,
3367
+ "grad_norm": 2.372843027114868,
3368
+ "learning_rate": 4.952210095494546e-05,
3369
+ "loss": 0.2154,
3370
+ "step": 9500
3371
+ },
3372
+ {
3373
+ "epoch": 0.7842706502636204,
3374
+ "grad_norm": 25.939308166503906,
3375
+ "learning_rate": 4.9516488601303255e-05,
3376
+ "loss": 0.1618,
3377
+ "step": 9520
3378
+ },
3379
+ {
3380
+ "epoch": 0.7859182776801406,
3381
+ "grad_norm": 1.780004858970642,
3382
+ "learning_rate": 4.951084380633013e-05,
3383
+ "loss": 0.1993,
3384
+ "step": 9540
3385
+ },
3386
+ {
3387
+ "epoch": 0.7875659050966608,
3388
+ "grad_norm": 1.1589267253875732,
3389
+ "learning_rate": 4.9505166577495546e-05,
3390
+ "loss": 0.1939,
3391
+ "step": 9560
3392
+ },
3393
+ {
3394
+ "epoch": 0.789213532513181,
3395
+ "grad_norm": 1.0380936861038208,
3396
+ "learning_rate": 4.949945692231185e-05,
3397
+ "loss": 0.1863,
3398
+ "step": 9580
3399
+ },
3400
+ {
3401
+ "epoch": 0.7908611599297012,
3402
+ "grad_norm": 3.860309600830078,
3403
+ "learning_rate": 4.949371484833433e-05,
3404
+ "loss": 0.2199,
3405
+ "step": 9600
3406
+ },
3407
+ {
3408
+ "epoch": 0.7925087873462214,
3409
+ "grad_norm": 2.6430704593658447,
3410
+ "learning_rate": 4.9487940363161155e-05,
3411
+ "loss": 0.161,
3412
+ "step": 9620
3413
+ },
3414
+ {
3415
+ "epoch": 0.7941564147627417,
3416
+ "grad_norm": 2.1502444744110107,
3417
+ "learning_rate": 4.948213347443339e-05,
3418
+ "loss": 0.2101,
3419
+ "step": 9640
3420
+ },
3421
+ {
3422
+ "epoch": 0.7958040421792618,
3423
+ "grad_norm": 0.9621193408966064,
3424
+ "learning_rate": 4.9476294189834974e-05,
3425
+ "loss": 0.2105,
3426
+ "step": 9660
3427
+ },
3428
+ {
3429
+ "epoch": 0.7974516695957821,
3430
+ "grad_norm": 0.8426349759101868,
3431
+ "learning_rate": 4.9470422517092696e-05,
3432
+ "loss": 0.1821,
3433
+ "step": 9680
3434
+ },
3435
+ {
3436
+ "epoch": 0.7990992970123023,
3437
+ "grad_norm": 2.1861345767974854,
3438
+ "learning_rate": 4.9464518463976246e-05,
3439
+ "loss": 0.2157,
3440
+ "step": 9700
3441
+ },
3442
+ {
3443
+ "epoch": 0.8007469244288224,
3444
+ "grad_norm": 1.074504017829895,
3445
+ "learning_rate": 4.945858203829812e-05,
3446
+ "loss": 0.1593,
3447
+ "step": 9720
3448
+ },
3449
+ {
3450
+ "epoch": 0.8023945518453427,
3451
+ "grad_norm": 1.4774161577224731,
3452
+ "learning_rate": 4.945261324791367e-05,
3453
+ "loss": 0.1981,
3454
+ "step": 9740
3455
+ },
3456
+ {
3457
+ "epoch": 0.8040421792618629,
3458
+ "grad_norm": 0.9464648962020874,
3459
+ "learning_rate": 4.944661210072107e-05,
3460
+ "loss": 0.2107,
3461
+ "step": 9760
3462
+ },
3463
+ {
3464
+ "epoch": 0.8056898066783831,
3465
+ "grad_norm": 0.34157031774520874,
3466
+ "learning_rate": 4.94405786046613e-05,
3467
+ "loss": 0.209,
3468
+ "step": 9780
3469
+ },
3470
+ {
3471
+ "epoch": 0.8073374340949033,
3472
+ "grad_norm": 3.6989963054656982,
3473
+ "learning_rate": 4.943451276771818e-05,
3474
+ "loss": 0.221,
3475
+ "step": 9800
3476
+ },
3477
+ {
3478
+ "epoch": 0.8089850615114236,
3479
+ "grad_norm": 0.6388671398162842,
3480
+ "learning_rate": 4.942841459791828e-05,
3481
+ "loss": 0.1661,
3482
+ "step": 9820
3483
+ },
3484
+ {
3485
+ "epoch": 0.8106326889279437,
3486
+ "grad_norm": 0.6647291779518127,
3487
+ "learning_rate": 4.9422284103330985e-05,
3488
+ "loss": 0.192,
3489
+ "step": 9840
3490
+ },
3491
+ {
3492
+ "epoch": 0.812280316344464,
3493
+ "grad_norm": 0.540134847164154,
3494
+ "learning_rate": 4.941612129206844e-05,
3495
+ "loss": 0.2126,
3496
+ "step": 9860
3497
+ },
3498
+ {
3499
+ "epoch": 0.8139279437609842,
3500
+ "grad_norm": 0.5511148571968079,
3501
+ "learning_rate": 4.940992617228556e-05,
3502
+ "loss": 0.2018,
3503
+ "step": 9880
3504
+ },
3505
+ {
3506
+ "epoch": 0.8155755711775043,
3507
+ "grad_norm": 2.1634738445281982,
3508
+ "learning_rate": 4.9403698752180006e-05,
3509
+ "loss": 0.2087,
3510
+ "step": 9900
3511
+ },
3512
+ {
3513
+ "epoch": 0.8172231985940246,
3514
+ "grad_norm": 1.3764668703079224,
3515
+ "learning_rate": 4.939743903999218e-05,
3516
+ "loss": 0.1561,
3517
+ "step": 9920
3518
+ },
3519
+ {
3520
+ "epoch": 0.8188708260105448,
3521
+ "grad_norm": 0.8314900398254395,
3522
+ "learning_rate": 4.939114704400523e-05,
3523
+ "loss": 0.2226,
3524
+ "step": 9940
3525
+ },
3526
+ {
3527
+ "epoch": 0.820518453427065,
3528
+ "grad_norm": 1.0714060068130493,
3529
+ "learning_rate": 4.9384822772544994e-05,
3530
+ "loss": 0.191,
3531
+ "step": 9960
3532
+ },
3533
+ {
3534
+ "epoch": 0.8221660808435852,
3535
+ "grad_norm": 0.5301602482795715,
3536
+ "learning_rate": 4.937846623398003e-05,
3537
+ "loss": 0.2069,
3538
+ "step": 9980
3539
+ },
3540
+ {
3541
+ "epoch": 0.8238137082601055,
3542
+ "grad_norm": 2.4229841232299805,
3543
+ "learning_rate": 4.9372077436721634e-05,
3544
+ "loss": 0.2132,
3545
+ "step": 10000
3546
+ },
3547
+ {
3548
+ "epoch": 0.8238137082601055,
3549
+ "eval_loss": 0.5383469462394714,
3550
+ "eval_runtime": 248.9799,
3551
+ "eval_samples_per_second": 87.513,
3552
+ "eval_steps_per_second": 21.881,
3553
+ "eval_wer": 0.23184117484036582,
3554
+ "step": 10000
3555
+ }
3556
+ ],
3557
+ "logging_steps": 20,
3558
+ "max_steps": 60690,
3559
+ "num_input_tokens_seen": 0,
3560
+ "num_train_epochs": 5,
3561
+ "save_steps": 2000,
3562
+ "stateful_callbacks": {
3563
+ "TrainerControl": {
3564
+ "args": {
3565
+ "should_epoch_stop": false,
3566
+ "should_evaluate": false,
3567
+ "should_log": false,
3568
+ "should_save": true,
3569
+ "should_training_stop": false
3570
+ },
3571
+ "attributes": {}
3572
+ }
3573
+ },
3574
+ "total_flos": 5.389121772584792e+20,
3575
+ "train_batch_size": 24,
3576
+ "trial_name": null,
3577
+ "trial_params": null
3578
+ }
checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a54ae952e51708d931d3b1bd9f52b61601a0073209f0242acda50b4c93d3f95
3
+ size 6584
checkpoint-10000/zero_to_fp32.py ADDED
@@ -0,0 +1,604 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
215
+ exclude_frozen_parameters)
216
+ elif zero_stage == 3:
217
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
218
+ exclude_frozen_parameters)
219
+
220
+
221
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
222
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
223
+ return
224
+
225
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
226
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
227
+
228
+ if debug:
229
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
230
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
231
+
232
+ wanted_params = len(frozen_param_shapes)
233
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
234
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
235
+ print(f'Frozen params: Have {avail_numel} numels to process.')
236
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
237
+
238
+ total_params = 0
239
+ total_numel = 0
240
+ for name, shape in frozen_param_shapes.items():
241
+ total_params += 1
242
+ unpartitioned_numel = shape.numel()
243
+ total_numel += unpartitioned_numel
244
+
245
+ state_dict[name] = frozen_param_fragments[name]
246
+
247
+ if debug:
248
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
249
+
250
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
251
+
252
+
253
+ def _has_callable(obj, fn):
254
+ attr = getattr(obj, fn, None)
255
+ return callable(attr)
256
+
257
+
258
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
259
+ param_shapes = zero_model_states[0].param_shapes
260
+
261
+ # Reconstruction protocol:
262
+ #
263
+ # XXX: document this
264
+
265
+ if debug:
266
+ for i in range(world_size):
267
+ for j in range(len(fp32_flat_groups[0])):
268
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
269
+
270
+ # XXX: memory usage doubles here (zero2)
271
+ num_param_groups = len(fp32_flat_groups[0])
272
+ merged_single_partition_of_fp32_groups = []
273
+ for i in range(num_param_groups):
274
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
275
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
276
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
277
+ avail_numel = sum(
278
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
279
+
280
+ if debug:
281
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
282
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
283
+ # not asserting if there is a mismatch due to possible padding
284
+ print(f"Have {avail_numel} numels to process.")
285
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
286
+
287
+ # params
288
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
289
+ # out-of-core computing solution
290
+ total_numel = 0
291
+ total_params = 0
292
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
293
+ offset = 0
294
+ avail_numel = full_single_fp32_vector.numel()
295
+ for name, shape in shapes.items():
296
+
297
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
298
+ total_numel += unpartitioned_numel
299
+ total_params += 1
300
+
301
+ if debug:
302
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
303
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
304
+ offset += unpartitioned_numel
305
+
306
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
307
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
308
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
309
+ # live optimizer object, so we are checking that the numbers are within the right range
310
+ align_to = 2 * world_size
311
+
312
+ def zero2_align(x):
313
+ return align_to * math.ceil(x / align_to)
314
+
315
+ if debug:
316
+ print(f"original offset={offset}, avail_numel={avail_numel}")
317
+
318
+ offset = zero2_align(offset)
319
+ avail_numel = zero2_align(avail_numel)
320
+
321
+ if debug:
322
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
323
+
324
+ # Sanity check
325
+ if offset != avail_numel:
326
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
327
+
328
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
329
+
330
+
331
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
332
+ exclude_frozen_parameters):
333
+ state_dict = OrderedDict()
334
+
335
+ # buffers
336
+ buffers = zero_model_states[0].buffers
337
+ state_dict.update(buffers)
338
+ if debug:
339
+ print(f"added {len(buffers)} buffers")
340
+
341
+ if not exclude_frozen_parameters:
342
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
343
+
344
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
345
+
346
+ # recover shared parameters
347
+ for pair in zero_model_states[0].shared_params:
348
+ if pair[1] in state_dict:
349
+ state_dict[pair[0]] = state_dict[pair[1]]
350
+
351
+ return state_dict
352
+
353
+
354
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
355
+ remainder = unpartitioned_numel % world_size
356
+ padding_numel = (world_size - remainder) if remainder else 0
357
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
358
+ return partitioned_numel, padding_numel
359
+
360
+
361
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
362
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
363
+ return
364
+
365
+ if debug:
366
+ for i in range(world_size):
367
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
368
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
369
+
370
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
371
+ wanted_params = len(frozen_param_shapes)
372
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
373
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
374
+ print(f'Frozen params: Have {avail_numel} numels to process.')
375
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
376
+
377
+ total_params = 0
378
+ total_numel = 0
379
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
380
+ total_params += 1
381
+ unpartitioned_numel = shape.numel()
382
+ total_numel += unpartitioned_numel
383
+
384
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
385
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
386
+
387
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
388
+
389
+ if debug:
390
+ print(
391
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
392
+ )
393
+
394
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
395
+
396
+
397
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
398
+ param_shapes = zero_model_states[0].param_shapes
399
+ avail_numel = fp32_flat_groups[0].numel() * world_size
400
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
401
+ # param, re-consolidating each param, while dealing with padding if any
402
+
403
+ # merge list of dicts, preserving order
404
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
405
+
406
+ if debug:
407
+ for i in range(world_size):
408
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
409
+
410
+ wanted_params = len(param_shapes)
411
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
412
+ # not asserting if there is a mismatch due to possible padding
413
+ avail_numel = fp32_flat_groups[0].numel() * world_size
414
+ print(f"Trainable params: Have {avail_numel} numels to process.")
415
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
416
+
417
+ # params
418
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
419
+ # out-of-core computing solution
420
+ offset = 0
421
+ total_numel = 0
422
+ total_params = 0
423
+ for name, shape in param_shapes.items():
424
+
425
+ unpartitioned_numel = shape.numel()
426
+ total_numel += unpartitioned_numel
427
+ total_params += 1
428
+
429
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
430
+
431
+ if debug:
432
+ print(
433
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
434
+ )
435
+
436
+ # XXX: memory usage doubles here
437
+ state_dict[name] = torch.cat(
438
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
439
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
440
+ offset += partitioned_numel
441
+
442
+ offset *= world_size
443
+
444
+ # Sanity check
445
+ if offset != avail_numel:
446
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
447
+
448
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
449
+
450
+
451
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
452
+ exclude_frozen_parameters):
453
+ state_dict = OrderedDict()
454
+
455
+ # buffers
456
+ buffers = zero_model_states[0].buffers
457
+ state_dict.update(buffers)
458
+ if debug:
459
+ print(f"added {len(buffers)} buffers")
460
+
461
+ if not exclude_frozen_parameters:
462
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
463
+
464
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
465
+
466
+ # recover shared parameters
467
+ for pair in zero_model_states[0].shared_params:
468
+ if pair[1] in state_dict:
469
+ state_dict[pair[0]] = state_dict[pair[1]]
470
+
471
+ return state_dict
472
+
473
+
474
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
475
+ """
476
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
477
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
478
+ via a model hub.
479
+
480
+ Args:
481
+ - ``checkpoint_dir``: path to the desired checkpoint folder
482
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
483
+ - ``exclude_frozen_parameters``: exclude frozen parameters
484
+
485
+ Returns:
486
+ - pytorch ``state_dict``
487
+
488
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
489
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
490
+ the checkpoint.
491
+
492
+ A typical usage might be ::
493
+
494
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
495
+ # do the training and checkpoint saving
496
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
497
+ model = model.cpu() # move to cpu
498
+ model.load_state_dict(state_dict)
499
+ # submit to model hub or save the model to share with others
500
+
501
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
502
+ application. i.e. you will need to re-initialize the deepspeed engine, since
503
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
504
+
505
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
506
+
507
+ """
508
+ if tag is None:
509
+ latest_path = os.path.join(checkpoint_dir, 'latest')
510
+ if os.path.isfile(latest_path):
511
+ with open(latest_path, 'r') as fd:
512
+ tag = fd.read().strip()
513
+ else:
514
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
515
+
516
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
517
+
518
+ if not os.path.isdir(ds_checkpoint_dir):
519
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
520
+
521
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
522
+
523
+
524
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
525
+ """
526
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
527
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
528
+
529
+ Args:
530
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
531
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
532
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
533
+ - ``exclude_frozen_parameters``: exclude frozen parameters
534
+ """
535
+
536
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
537
+ print(f"Saving fp32 state dict to {output_file}")
538
+ torch.save(state_dict, output_file)
539
+
540
+
541
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
542
+ """
543
+ 1. Put the provided model to cpu
544
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
545
+ 3. Load it into the provided model
546
+
547
+ Args:
548
+ - ``model``: the model object to update
549
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
550
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
551
+
552
+ Returns:
553
+ - ``model`: modified model
554
+
555
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
556
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
557
+ conveniently placed for you in the checkpoint folder.
558
+
559
+ A typical usage might be ::
560
+
561
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
562
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
563
+ # submit to model hub or save the model to share with others
564
+
565
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
566
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
567
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
568
+
569
+ """
570
+ logger.info(f"Extracting fp32 weights")
571
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
572
+
573
+ logger.info(f"Overwriting model with fp32 weights")
574
+ model = model.cpu()
575
+ model.load_state_dict(state_dict, strict=False)
576
+
577
+ return model
578
+
579
+
580
+ if __name__ == "__main__":
581
+
582
+ parser = argparse.ArgumentParser()
583
+ parser.add_argument("checkpoint_dir",
584
+ type=str,
585
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
586
+ parser.add_argument(
587
+ "output_file",
588
+ type=str,
589
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
590
+ parser.add_argument("-t",
591
+ "--tag",
592
+ type=str,
593
+ default=None,
594
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
595
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
596
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
597
+ args = parser.parse_args()
598
+
599
+ debug = args.debug
600
+
601
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
602
+ args.output_file,
603
+ tag=args.tag,
604
+ exclude_frozen_parameters=args.exclude_frozen_parameters)