Hugging Face's logo

text2font
/
text2svg_summarization-2-epochs-28-step-367000

Transformers
PyTorch
mt5
text2text-generation
Model card Files
xet
Community
pain commited on
Commit
a2160ad
·
1 Parent(s): 9665bef

Upload 12 files

Browse files
Files changed (13) hide show
  1. .gitattributes +1 -0
  2. config.json +32 -0
  3. generation_config.json +7 -0
  4. optimizer.pt +3 -0
  5. pytorch_model.bin +3 -0
  6. rng_state.pth +3 -0
  7. scheduler.pt +3 -0
  8. special_tokens_map.json +5 -0
  9. spiece.model +3 -0
  10. tokenizer.json +3 -0
  11. tokenizer_config.json +12 -0
  12. trainer_state.json +4420 -0
  13. training_args.bin +3 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "google/mt5-large",
3
+ "architectures": [
4
+ "MT5ForConditionalGeneration"
5
+ ],
6
+ "d_ff": 2816,
7
+ "d_kv": 64,
8
+ "d_model": 1024,
9
+ "decoder_start_token_id": 0,
10
+ "dense_act_fn": "gelu_new",
11
+ "dropout_rate": 0.1,
12
+ "eos_token_id": 1,
13
+ "feed_forward_proj": "gated-gelu",
14
+ "initializer_factor": 1.0,
15
+ "is_encoder_decoder": true,
16
+ "is_gated_act": true,
17
+ "layer_norm_epsilon": 1e-06,
18
+ "model_type": "mt5",
19
+ "num_decoder_layers": 24,
20
+ "num_heads": 16,
21
+ "num_layers": 24,
22
+ "output_past": true,
23
+ "pad_token_id": 0,
24
+ "relative_attention_max_distance": 128,
25
+ "relative_attention_num_buckets": 32,
26
+ "tie_word_embeddings": false,
27
+ "tokenizer_class": "T5Tokenizer",
28
+ "torch_dtype": "float32",
29
+ "transformers_version": "4.32.0.dev0",
30
+ "use_cache": true,
31
+ "vocab_size": 250112
32
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "decoder_start_token_id": 0,
4
+ "eos_token_id": 1,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.32.0.dev0"
7
+ }
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de3a28d17a0014d5330a65cad99528749887bf0d5ffd0ac03714274b4ef8e5af
3
+ size 9836995421
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2cc4946d764972b4927c6ef877e0d39ae2d8d32b5249d986a64210a827a0f91
3
+ size 4918519193
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d4de11ca89f079e723f0f73b1fe90a35f6e2f59c9aed72831885c46bd114cd9
3
+ size 14575
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f6e025c759e7c7df9227f233ca943ab1612a0ec235968076ef0e978b9f4b013
3
+ size 627
special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "eos_token": "</s>",
3
+ "pad_token": "<pad>",
4
+ "unk_token": "<unk>"
5
+ }
spiece.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef78f86560d809067d12bac6c09f19a462cb3af3f54d2b8acbba26e1433125d6
3
+ size 4309802
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93c3578052e1605d8332eb961bc08d72e246071974e4cc54aa6991826b802aa5
3
+ size 16330369
tokenizer_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": null,
3
+ "clean_up_tokenization_spaces": true,
4
+ "eos_token": "</s>",
5
+ "extra_ids": 0,
6
+ "legacy": true,
7
+ "model_max_length": 1000000000000000019884624838656,
8
+ "pad_token": "<pad>",
9
+ "sp_model_kwargs": {},
10
+ "tokenizer_class": "T5Tokenizer",
11
+ "unk_token": "<unk>"
12
+ }
trainer_state.json ADDED
@@ -0,0 +1,4420 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 27.923609525983412,
5
+ "global_step": 367000,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.04,
12
+ "learning_rate": 4.9993659489208456e-05,
13
+ "loss": 8.4894,
14
+ "step": 500
15
+ },
16
+ {
17
+ "epoch": 0.08,
18
+ "learning_rate": 4.99873189784169e-05,
19
+ "loss": 1.9027,
20
+ "step": 1000
21
+ },
22
+ {
23
+ "epoch": 0.11,
24
+ "learning_rate": 4.9980978467625356e-05,
25
+ "loss": 1.4486,
26
+ "step": 1500
27
+ },
28
+ {
29
+ "epoch": 0.15,
30
+ "learning_rate": 4.99746379568338e-05,
31
+ "loss": 1.2042,
32
+ "step": 2000
33
+ },
34
+ {
35
+ "epoch": 0.19,
36
+ "learning_rate": 4.996829744604226e-05,
37
+ "loss": 1.0918,
38
+ "step": 2500
39
+ },
40
+ {
41
+ "epoch": 0.23,
42
+ "learning_rate": 4.996195693525071e-05,
43
+ "loss": 0.943,
44
+ "step": 3000
45
+ },
46
+ {
47
+ "epoch": 0.27,
48
+ "learning_rate": 4.995561642445916e-05,
49
+ "loss": 0.8502,
50
+ "step": 3500
51
+ },
52
+ {
53
+ "epoch": 0.3,
54
+ "learning_rate": 4.994927591366761e-05,
55
+ "loss": 0.7086,
56
+ "step": 4000
57
+ },
58
+ {
59
+ "epoch": 0.34,
60
+ "learning_rate": 4.994293540287606e-05,
61
+ "loss": 0.7026,
62
+ "step": 4500
63
+ },
64
+ {
65
+ "epoch": 0.38,
66
+ "learning_rate": 4.993659489208451e-05,
67
+ "loss": 0.6089,
68
+ "step": 5000
69
+ },
70
+ {
71
+ "epoch": 0.42,
72
+ "learning_rate": 4.993025438129296e-05,
73
+ "loss": 0.5424,
74
+ "step": 5500
75
+ },
76
+ {
77
+ "epoch": 0.46,
78
+ "learning_rate": 4.992391387050141e-05,
79
+ "loss": 0.5609,
80
+ "step": 6000
81
+ },
82
+ {
83
+ "epoch": 0.49,
84
+ "learning_rate": 4.9917573359709865e-05,
85
+ "loss": 0.542,
86
+ "step": 6500
87
+ },
88
+ {
89
+ "epoch": 0.53,
90
+ "learning_rate": 4.991123284891831e-05,
91
+ "loss": 0.4546,
92
+ "step": 7000
93
+ },
94
+ {
95
+ "epoch": 0.57,
96
+ "learning_rate": 4.9904892338126766e-05,
97
+ "loss": 0.4209,
98
+ "step": 7500
99
+ },
100
+ {
101
+ "epoch": 0.61,
102
+ "learning_rate": 4.989855182733521e-05,
103
+ "loss": 0.408,
104
+ "step": 8000
105
+ },
106
+ {
107
+ "epoch": 0.65,
108
+ "learning_rate": 4.9892211316543666e-05,
109
+ "loss": 0.4211,
110
+ "step": 8500
111
+ },
112
+ {
113
+ "epoch": 0.68,
114
+ "learning_rate": 4.988587080575211e-05,
115
+ "loss": 0.3948,
116
+ "step": 9000
117
+ },
118
+ {
119
+ "epoch": 0.72,
120
+ "learning_rate": 4.9879530294960567e-05,
121
+ "loss": 0.3775,
122
+ "step": 9500
123
+ },
124
+ {
125
+ "epoch": 0.76,
126
+ "learning_rate": 4.9873189784169013e-05,
127
+ "loss": 0.3339,
128
+ "step": 10000
129
+ },
130
+ {
131
+ "epoch": 0.8,
132
+ "learning_rate": 4.986684927337747e-05,
133
+ "loss": 0.3237,
134
+ "step": 10500
135
+ },
136
+ {
137
+ "epoch": 0.84,
138
+ "learning_rate": 4.9860508762585914e-05,
139
+ "loss": 0.39,
140
+ "step": 11000
141
+ },
142
+ {
143
+ "epoch": 0.87,
144
+ "learning_rate": 4.985416825179437e-05,
145
+ "loss": 0.3066,
146
+ "step": 11500
147
+ },
148
+ {
149
+ "epoch": 0.91,
150
+ "learning_rate": 4.9847827741002814e-05,
151
+ "loss": 0.2733,
152
+ "step": 12000
153
+ },
154
+ {
155
+ "epoch": 0.95,
156
+ "learning_rate": 4.984148723021127e-05,
157
+ "loss": 0.3175,
158
+ "step": 12500
159
+ },
160
+ {
161
+ "epoch": 0.99,
162
+ "learning_rate": 4.9835146719419715e-05,
163
+ "loss": 0.2387,
164
+ "step": 13000
165
+ },
166
+ {
167
+ "epoch": 1.03,
168
+ "learning_rate": 4.982880620862817e-05,
169
+ "loss": 0.2686,
170
+ "step": 13500
171
+ },
172
+ {
173
+ "epoch": 1.07,
174
+ "learning_rate": 4.9822465697836615e-05,
175
+ "loss": 0.2533,
176
+ "step": 14000
177
+ },
178
+ {
179
+ "epoch": 1.1,
180
+ "learning_rate": 4.981612518704507e-05,
181
+ "loss": 0.2573,
182
+ "step": 14500
183
+ },
184
+ {
185
+ "epoch": 1.14,
186
+ "learning_rate": 4.980978467625352e-05,
187
+ "loss": 0.2577,
188
+ "step": 15000
189
+ },
190
+ {
191
+ "epoch": 1.18,
192
+ "learning_rate": 4.980344416546197e-05,
193
+ "loss": 0.2206,
194
+ "step": 15500
195
+ },
196
+ {
197
+ "epoch": 1.22,
198
+ "learning_rate": 4.979710365467042e-05,
199
+ "loss": 0.2335,
200
+ "step": 16000
201
+ },
202
+ {
203
+ "epoch": 1.26,
204
+ "learning_rate": 4.979076314387887e-05,
205
+ "loss": 0.2241,
206
+ "step": 16500
207
+ },
208
+ {
209
+ "epoch": 1.29,
210
+ "learning_rate": 4.978442263308732e-05,
211
+ "loss": 0.1935,
212
+ "step": 17000
213
+ },
214
+ {
215
+ "epoch": 1.33,
216
+ "learning_rate": 4.977808212229578e-05,
217
+ "loss": 0.1706,
218
+ "step": 17500
219
+ },
220
+ {
221
+ "epoch": 1.37,
222
+ "learning_rate": 4.9771741611504224e-05,
223
+ "loss": 0.1869,
224
+ "step": 18000
225
+ },
226
+ {
227
+ "epoch": 1.41,
228
+ "learning_rate": 4.976540110071268e-05,
229
+ "loss": 0.1732,
230
+ "step": 18500
231
+ },
232
+ {
233
+ "epoch": 1.45,
234
+ "learning_rate": 4.9759060589921124e-05,
235
+ "loss": 0.1603,
236
+ "step": 19000
237
+ },
238
+ {
239
+ "epoch": 1.48,
240
+ "learning_rate": 4.975272007912958e-05,
241
+ "loss": 0.1848,
242
+ "step": 19500
243
+ },
244
+ {
245
+ "epoch": 1.52,
246
+ "learning_rate": 4.9746379568338024e-05,
247
+ "loss": 0.166,
248
+ "step": 20000
249
+ },
250
+ {
251
+ "epoch": 1.56,
252
+ "learning_rate": 4.974003905754648e-05,
253
+ "loss": 0.1645,
254
+ "step": 20500
255
+ },
256
+ {
257
+ "epoch": 1.6,
258
+ "learning_rate": 4.973369854675493e-05,
259
+ "loss": 0.1521,
260
+ "step": 21000
261
+ },
262
+ {
263
+ "epoch": 1.64,
264
+ "learning_rate": 4.972735803596338e-05,
265
+ "loss": 0.1712,
266
+ "step": 21500
267
+ },
268
+ {
269
+ "epoch": 1.67,
270
+ "learning_rate": 4.972101752517183e-05,
271
+ "loss": 0.1625,
272
+ "step": 22000
273
+ },
274
+ {
275
+ "epoch": 1.71,
276
+ "learning_rate": 4.971467701438028e-05,
277
+ "loss": 0.1382,
278
+ "step": 22500
279
+ },
280
+ {
281
+ "epoch": 1.75,
282
+ "learning_rate": 4.970833650358873e-05,
283
+ "loss": 0.128,
284
+ "step": 23000
285
+ },
286
+ {
287
+ "epoch": 1.79,
288
+ "learning_rate": 4.9701995992797186e-05,
289
+ "loss": 0.1351,
290
+ "step": 23500
291
+ },
292
+ {
293
+ "epoch": 1.83,
294
+ "learning_rate": 4.969565548200563e-05,
295
+ "loss": 0.1162,
296
+ "step": 24000
297
+ },
298
+ {
299
+ "epoch": 1.86,
300
+ "learning_rate": 4.9689314971214087e-05,
301
+ "loss": 0.119,
302
+ "step": 24500
303
+ },
304
+ {
305
+ "epoch": 1.9,
306
+ "learning_rate": 4.968297446042253e-05,
307
+ "loss": 0.1359,
308
+ "step": 25000
309
+ },
310
+ {
311
+ "epoch": 1.94,
312
+ "learning_rate": 4.967663394963099e-05,
313
+ "loss": 0.1243,
314
+ "step": 25500
315
+ },
316
+ {
317
+ "epoch": 1.98,
318
+ "learning_rate": 4.9670293438839434e-05,
319
+ "loss": 0.1108,
320
+ "step": 26000
321
+ },
322
+ {
323
+ "epoch": 2.02,
324
+ "learning_rate": 4.966395292804789e-05,
325
+ "loss": 0.1337,
326
+ "step": 26500
327
+ },
328
+ {
329
+ "epoch": 2.05,
330
+ "learning_rate": 4.965761241725634e-05,
331
+ "loss": 0.1246,
332
+ "step": 27000
333
+ },
334
+ {
335
+ "epoch": 2.09,
336
+ "learning_rate": 4.965127190646479e-05,
337
+ "loss": 0.1277,
338
+ "step": 27500
339
+ },
340
+ {
341
+ "epoch": 2.13,
342
+ "learning_rate": 4.964493139567324e-05,
343
+ "loss": 0.1268,
344
+ "step": 28000
345
+ },
346
+ {
347
+ "epoch": 2.17,
348
+ "learning_rate": 4.963859088488169e-05,
349
+ "loss": 0.1159,
350
+ "step": 28500
351
+ },
352
+ {
353
+ "epoch": 2.21,
354
+ "learning_rate": 4.963225037409014e-05,
355
+ "loss": 0.0991,
356
+ "step": 29000
357
+ },
358
+ {
359
+ "epoch": 2.24,
360
+ "learning_rate": 4.962590986329859e-05,
361
+ "loss": 0.0979,
362
+ "step": 29500
363
+ },
364
+ {
365
+ "epoch": 2.28,
366
+ "learning_rate": 4.961956935250704e-05,
367
+ "loss": 0.0987,
368
+ "step": 30000
369
+ },
370
+ {
371
+ "epoch": 2.32,
372
+ "learning_rate": 4.961322884171549e-05,
373
+ "loss": 0.0976,
374
+ "step": 30500
375
+ },
376
+ {
377
+ "epoch": 2.36,
378
+ "learning_rate": 4.960688833092394e-05,
379
+ "loss": 0.0878,
380
+ "step": 31000
381
+ },
382
+ {
383
+ "epoch": 2.4,
384
+ "learning_rate": 4.960054782013239e-05,
385
+ "loss": 0.115,
386
+ "step": 31500
387
+ },
388
+ {
389
+ "epoch": 2.43,
390
+ "learning_rate": 4.959420730934084e-05,
391
+ "loss": 0.0815,
392
+ "step": 32000
393
+ },
394
+ {
395
+ "epoch": 2.47,
396
+ "learning_rate": 4.958786679854929e-05,
397
+ "loss": 0.1181,
398
+ "step": 32500
399
+ },
400
+ {
401
+ "epoch": 2.51,
402
+ "learning_rate": 4.9581526287757744e-05,
403
+ "loss": 0.1064,
404
+ "step": 33000
405
+ },
406
+ {
407
+ "epoch": 2.55,
408
+ "learning_rate": 4.957518577696619e-05,
409
+ "loss": 0.0958,
410
+ "step": 33500
411
+ },
412
+ {
413
+ "epoch": 2.59,
414
+ "learning_rate": 4.9568845266174644e-05,
415
+ "loss": 0.0966,
416
+ "step": 34000
417
+ },
418
+ {
419
+ "epoch": 2.62,
420
+ "learning_rate": 4.956250475538309e-05,
421
+ "loss": 0.0879,
422
+ "step": 34500
423
+ },
424
+ {
425
+ "epoch": 2.66,
426
+ "learning_rate": 4.9556164244591544e-05,
427
+ "loss": 0.1014,
428
+ "step": 35000
429
+ },
430
+ {
431
+ "epoch": 2.7,
432
+ "learning_rate": 4.95498237338e-05,
433
+ "loss": 0.082,
434
+ "step": 35500
435
+ },
436
+ {
437
+ "epoch": 2.74,
438
+ "learning_rate": 4.9543483223008445e-05,
439
+ "loss": 0.0879,
440
+ "step": 36000
441
+ },
442
+ {
443
+ "epoch": 2.78,
444
+ "learning_rate": 4.95371427122169e-05,
445
+ "loss": 0.0887,
446
+ "step": 36500
447
+ },
448
+ {
449
+ "epoch": 2.82,
450
+ "learning_rate": 4.9530802201425345e-05,
451
+ "loss": 0.0937,
452
+ "step": 37000
453
+ },
454
+ {
455
+ "epoch": 2.85,
456
+ "learning_rate": 4.95244616906338e-05,
457
+ "loss": 0.0961,
458
+ "step": 37500
459
+ },
460
+ {
461
+ "epoch": 2.89,
462
+ "learning_rate": 4.951812117984225e-05,
463
+ "loss": 0.1007,
464
+ "step": 38000
465
+ },
466
+ {
467
+ "epoch": 2.93,
468
+ "learning_rate": 4.95117806690507e-05,
469
+ "loss": 0.0892,
470
+ "step": 38500
471
+ },
472
+ {
473
+ "epoch": 2.97,
474
+ "learning_rate": 4.950544015825915e-05,
475
+ "loss": 0.0982,
476
+ "step": 39000
477
+ },
478
+ {
479
+ "epoch": 3.01,
480
+ "learning_rate": 4.94990996474676e-05,
481
+ "loss": 0.0835,
482
+ "step": 39500
483
+ },
484
+ {
485
+ "epoch": 3.04,
486
+ "learning_rate": 4.949275913667605e-05,
487
+ "loss": 0.0873,
488
+ "step": 40000
489
+ },
490
+ {
491
+ "epoch": 3.08,
492
+ "learning_rate": 4.948641862588451e-05,
493
+ "loss": 0.078,
494
+ "step": 40500
495
+ },
496
+ {
497
+ "epoch": 3.12,
498
+ "learning_rate": 4.9480078115092954e-05,
499
+ "loss": 0.0774,
500
+ "step": 41000
501
+ },
502
+ {
503
+ "epoch": 3.16,
504
+ "learning_rate": 4.947373760430141e-05,
505
+ "loss": 0.0733,
506
+ "step": 41500
507
+ },
508
+ {
509
+ "epoch": 3.2,
510
+ "learning_rate": 4.9467397093509854e-05,
511
+ "loss": 0.0837,
512
+ "step": 42000
513
+ },
514
+ {
515
+ "epoch": 3.23,
516
+ "learning_rate": 4.946105658271831e-05,
517
+ "loss": 0.0786,
518
+ "step": 42500
519
+ },
520
+ {
521
+ "epoch": 3.27,
522
+ "learning_rate": 4.9454716071926755e-05,
523
+ "loss": 0.0808,
524
+ "step": 43000
525
+ },
526
+ {
527
+ "epoch": 3.31,
528
+ "learning_rate": 4.944837556113521e-05,
529
+ "loss": 0.083,
530
+ "step": 43500
531
+ },
532
+ {
533
+ "epoch": 3.35,
534
+ "learning_rate": 4.944203505034366e-05,
535
+ "loss": 0.0687,
536
+ "step": 44000
537
+ },
538
+ {
539
+ "epoch": 3.39,
540
+ "learning_rate": 4.943569453955211e-05,
541
+ "loss": 0.0714,
542
+ "step": 44500
543
+ },
544
+ {
545
+ "epoch": 3.42,
546
+ "learning_rate": 4.942935402876056e-05,
547
+ "loss": 0.0708,
548
+ "step": 45000
549
+ },
550
+ {
551
+ "epoch": 3.46,
552
+ "learning_rate": 4.942301351796901e-05,
553
+ "loss": 0.0776,
554
+ "step": 45500
555
+ },
556
+ {
557
+ "epoch": 3.5,
558
+ "learning_rate": 4.941667300717746e-05,
559
+ "loss": 0.0749,
560
+ "step": 46000
561
+ },
562
+ {
563
+ "epoch": 3.54,
564
+ "learning_rate": 4.9410332496385916e-05,
565
+ "loss": 0.0845,
566
+ "step": 46500
567
+ },
568
+ {
569
+ "epoch": 3.58,
570
+ "learning_rate": 4.940399198559436e-05,
571
+ "loss": 0.0696,
572
+ "step": 47000
573
+ },
574
+ {
575
+ "epoch": 3.61,
576
+ "learning_rate": 4.939765147480282e-05,
577
+ "loss": 0.0848,
578
+ "step": 47500
579
+ },
580
+ {
581
+ "epoch": 3.65,
582
+ "learning_rate": 4.9391310964011263e-05,
583
+ "loss": 0.0686,
584
+ "step": 48000
585
+ },
586
+ {
587
+ "epoch": 3.69,
588
+ "learning_rate": 4.938497045321972e-05,
589
+ "loss": 0.0744,
590
+ "step": 48500
591
+ },
592
+ {
593
+ "epoch": 3.73,
594
+ "learning_rate": 4.9378629942428164e-05,
595
+ "loss": 0.0652,
596
+ "step": 49000
597
+ },
598
+ {
599
+ "epoch": 3.77,
600
+ "learning_rate": 4.937228943163662e-05,
601
+ "loss": 0.08,
602
+ "step": 49500
603
+ },
604
+ {
605
+ "epoch": 3.8,
606
+ "learning_rate": 4.9365948920845064e-05,
607
+ "loss": 0.0767,
608
+ "step": 50000
609
+ },
610
+ {
611
+ "epoch": 3.84,
612
+ "learning_rate": 4.935960841005352e-05,
613
+ "loss": 0.0727,
614
+ "step": 50500
615
+ },
616
+ {
617
+ "epoch": 3.88,
618
+ "learning_rate": 4.9353267899261965e-05,
619
+ "loss": 0.0666,
620
+ "step": 51000
621
+ },
622
+ {
623
+ "epoch": 3.92,
624
+ "learning_rate": 4.934692738847042e-05,
625
+ "loss": 0.0803,
626
+ "step": 51500
627
+ },
628
+ {
629
+ "epoch": 3.96,
630
+ "learning_rate": 4.9340586877678865e-05,
631
+ "loss": 0.077,
632
+ "step": 52000
633
+ },
634
+ {
635
+ "epoch": 3.99,
636
+ "learning_rate": 4.933424636688732e-05,
637
+ "loss": 0.0654,
638
+ "step": 52500
639
+ },
640
+ {
641
+ "epoch": 4.03,
642
+ "learning_rate": 4.9327905856095766e-05,
643
+ "loss": 0.0676,
644
+ "step": 53000
645
+ },
646
+ {
647
+ "epoch": 4.07,
648
+ "learning_rate": 4.932156534530422e-05,
649
+ "loss": 0.0672,
650
+ "step": 53500
651
+ },
652
+ {
653
+ "epoch": 4.11,
654
+ "learning_rate": 4.9315224834512666e-05,
655
+ "loss": 0.0641,
656
+ "step": 54000
657
+ },
658
+ {
659
+ "epoch": 4.15,
660
+ "learning_rate": 4.930888432372112e-05,
661
+ "loss": 0.0615,
662
+ "step": 54500
663
+ },
664
+ {
665
+ "epoch": 4.18,
666
+ "learning_rate": 4.930254381292957e-05,
667
+ "loss": 0.064,
668
+ "step": 55000
669
+ },
670
+ {
671
+ "epoch": 4.22,
672
+ "learning_rate": 4.929620330213802e-05,
673
+ "loss": 0.0683,
674
+ "step": 55500
675
+ },
676
+ {
677
+ "epoch": 4.26,
678
+ "learning_rate": 4.9289862791346474e-05,
679
+ "loss": 0.0607,
680
+ "step": 56000
681
+ },
682
+ {
683
+ "epoch": 4.3,
684
+ "learning_rate": 4.928352228055492e-05,
685
+ "loss": 0.0649,
686
+ "step": 56500
687
+ },
688
+ {
689
+ "epoch": 4.34,
690
+ "learning_rate": 4.9277181769763374e-05,
691
+ "loss": 0.0605,
692
+ "step": 57000
693
+ },
694
+ {
695
+ "epoch": 4.37,
696
+ "learning_rate": 4.927084125897182e-05,
697
+ "loss": 0.0621,
698
+ "step": 57500
699
+ },
700
+ {
701
+ "epoch": 4.41,
702
+ "learning_rate": 4.9264500748180274e-05,
703
+ "loss": 0.0608,
704
+ "step": 58000
705
+ },
706
+ {
707
+ "epoch": 4.45,
708
+ "learning_rate": 4.925816023738873e-05,
709
+ "loss": 0.0596,
710
+ "step": 58500
711
+ },
712
+ {
713
+ "epoch": 4.49,
714
+ "learning_rate": 4.9251819726597175e-05,
715
+ "loss": 0.0595,
716
+ "step": 59000
717
+ },
718
+ {
719
+ "epoch": 4.53,
720
+ "learning_rate": 4.924547921580563e-05,
721
+ "loss": 0.0616,
722
+ "step": 59500
723
+ },
724
+ {
725
+ "epoch": 4.57,
726
+ "learning_rate": 4.9239138705014075e-05,
727
+ "loss": 0.0579,
728
+ "step": 60000
729
+ },
730
+ {
731
+ "epoch": 4.6,
732
+ "learning_rate": 4.923279819422253e-05,
733
+ "loss": 0.0578,
734
+ "step": 60500
735
+ },
736
+ {
737
+ "epoch": 4.64,
738
+ "learning_rate": 4.922645768343098e-05,
739
+ "loss": 0.061,
740
+ "step": 61000
741
+ },
742
+ {
743
+ "epoch": 4.68,
744
+ "learning_rate": 4.922011717263943e-05,
745
+ "loss": 0.0616,
746
+ "step": 61500
747
+ },
748
+ {
749
+ "epoch": 4.72,
750
+ "learning_rate": 4.921377666184788e-05,
751
+ "loss": 0.0555,
752
+ "step": 62000
753
+ },
754
+ {
755
+ "epoch": 4.76,
756
+ "learning_rate": 4.920743615105633e-05,
757
+ "loss": 0.0651,
758
+ "step": 62500
759
+ },
760
+ {
761
+ "epoch": 4.79,
762
+ "learning_rate": 4.920109564026478e-05,
763
+ "loss": 0.0603,
764
+ "step": 63000
765
+ },
766
+ {
767
+ "epoch": 4.83,
768
+ "learning_rate": 4.919475512947323e-05,
769
+ "loss": 0.0638,
770
+ "step": 63500
771
+ },
772
+ {
773
+ "epoch": 4.87,
774
+ "learning_rate": 4.9188414618681684e-05,
775
+ "loss": 0.0568,
776
+ "step": 64000
777
+ },
778
+ {
779
+ "epoch": 4.91,
780
+ "learning_rate": 4.918207410789014e-05,
781
+ "loss": 0.0612,
782
+ "step": 64500
783
+ },
784
+ {
785
+ "epoch": 4.95,
786
+ "learning_rate": 4.9175733597098584e-05,
787
+ "loss": 0.056,
788
+ "step": 65000
789
+ },
790
+ {
791
+ "epoch": 4.98,
792
+ "learning_rate": 4.916939308630704e-05,
793
+ "loss": 0.0664,
794
+ "step": 65500
795
+ },
796
+ {
797
+ "epoch": 5.02,
798
+ "learning_rate": 4.9163052575515485e-05,
799
+ "loss": 0.0596,
800
+ "step": 66000
801
+ },
802
+ {
803
+ "epoch": 5.06,
804
+ "learning_rate": 4.915671206472394e-05,
805
+ "loss": 0.0543,
806
+ "step": 66500
807
+ },
808
+ {
809
+ "epoch": 5.1,
810
+ "learning_rate": 4.915037155393239e-05,
811
+ "loss": 0.0533,
812
+ "step": 67000
813
+ },
814
+ {
815
+ "epoch": 5.14,
816
+ "learning_rate": 4.914403104314084e-05,
817
+ "loss": 0.0543,
818
+ "step": 67500
819
+ },
820
+ {
821
+ "epoch": 5.17,
822
+ "learning_rate": 4.913769053234929e-05,
823
+ "loss": 0.0552,
824
+ "step": 68000
825
+ },
826
+ {
827
+ "epoch": 5.21,
828
+ "learning_rate": 4.913135002155774e-05,
829
+ "loss": 0.0542,
830
+ "step": 68500
831
+ },
832
+ {
833
+ "epoch": 5.25,
834
+ "learning_rate": 4.912500951076619e-05,
835
+ "loss": 0.0569,
836
+ "step": 69000
837
+ },
838
+ {
839
+ "epoch": 5.29,
840
+ "learning_rate": 4.9118668999974646e-05,
841
+ "loss": 0.0554,
842
+ "step": 69500
843
+ },
844
+ {
845
+ "epoch": 5.33,
846
+ "learning_rate": 4.911232848918309e-05,
847
+ "loss": 0.053,
848
+ "step": 70000
849
+ },
850
+ {
851
+ "epoch": 5.36,
852
+ "learning_rate": 4.910598797839154e-05,
853
+ "loss": 0.0567,
854
+ "step": 70500
855
+ },
856
+ {
857
+ "epoch": 5.4,
858
+ "learning_rate": 4.9099647467599994e-05,
859
+ "loss": 0.0601,
860
+ "step": 71000
861
+ },
862
+ {
863
+ "epoch": 5.44,
864
+ "learning_rate": 4.909330695680844e-05,
865
+ "loss": 0.0527,
866
+ "step": 71500
867
+ },
868
+ {
869
+ "epoch": 5.48,
870
+ "learning_rate": 4.9086966446016894e-05,
871
+ "loss": 0.0496,
872
+ "step": 72000
873
+ },
874
+ {
875
+ "epoch": 5.52,
876
+ "learning_rate": 4.908062593522534e-05,
877
+ "loss": 0.048,
878
+ "step": 72500
879
+ },
880
+ {
881
+ "epoch": 5.55,
882
+ "learning_rate": 4.9074285424433794e-05,
883
+ "loss": 0.0579,
884
+ "step": 73000
885
+ },
886
+ {
887
+ "epoch": 5.59,
888
+ "learning_rate": 4.906794491364224e-05,
889
+ "loss": 0.0517,
890
+ "step": 73500
891
+ },
892
+ {
893
+ "epoch": 5.63,
894
+ "learning_rate": 4.9061604402850695e-05,
895
+ "loss": 0.0524,
896
+ "step": 74000
897
+ },
898
+ {
899
+ "epoch": 5.67,
900
+ "learning_rate": 4.905526389205914e-05,
901
+ "loss": 0.0543,
902
+ "step": 74500
903
+ },
904
+ {
905
+ "epoch": 5.71,
906
+ "learning_rate": 4.9048923381267595e-05,
907
+ "loss": 0.0526,
908
+ "step": 75000
909
+ },
910
+ {
911
+ "epoch": 5.74,
912
+ "learning_rate": 4.904258287047605e-05,
913
+ "loss": 0.0533,
914
+ "step": 75500
915
+ },
916
+ {
917
+ "epoch": 5.78,
918
+ "learning_rate": 4.9036242359684496e-05,
919
+ "loss": 0.051,
920
+ "step": 76000
921
+ },
922
+ {
923
+ "epoch": 5.82,
924
+ "learning_rate": 4.902990184889295e-05,
925
+ "loss": 0.0516,
926
+ "step": 76500
927
+ },
928
+ {
929
+ "epoch": 5.86,
930
+ "learning_rate": 4.9023561338101396e-05,
931
+ "loss": 0.0488,
932
+ "step": 77000
933
+ },
934
+ {
935
+ "epoch": 5.9,
936
+ "learning_rate": 4.901722082730985e-05,
937
+ "loss": 0.0485,
938
+ "step": 77500
939
+ },
940
+ {
941
+ "epoch": 5.93,
942
+ "learning_rate": 4.9010880316518297e-05,
943
+ "loss": 0.0452,
944
+ "step": 78000
945
+ },
946
+ {
947
+ "epoch": 5.97,
948
+ "learning_rate": 4.900453980572675e-05,
949
+ "loss": 0.0549,
950
+ "step": 78500
951
+ },
952
+ {
953
+ "epoch": 6.01,
954
+ "learning_rate": 4.8998199294935204e-05,
955
+ "loss": 0.0552,
956
+ "step": 79000
957
+ },
958
+ {
959
+ "epoch": 6.05,
960
+ "learning_rate": 4.899185878414365e-05,
961
+ "loss": 0.0461,
962
+ "step": 79500
963
+ },
964
+ {
965
+ "epoch": 6.09,
966
+ "learning_rate": 4.8985518273352104e-05,
967
+ "loss": 0.0501,
968
+ "step": 80000
969
+ },
970
+ {
971
+ "epoch": 6.12,
972
+ "learning_rate": 4.897917776256055e-05,
973
+ "loss": 0.0458,
974
+ "step": 80500
975
+ },
976
+ {
977
+ "epoch": 6.16,
978
+ "learning_rate": 4.8972837251769005e-05,
979
+ "loss": 0.0435,
980
+ "step": 81000
981
+ },
982
+ {
983
+ "epoch": 6.2,
984
+ "learning_rate": 4.896649674097746e-05,
985
+ "loss": 0.0444,
986
+ "step": 81500
987
+ },
988
+ {
989
+ "epoch": 6.24,
990
+ "learning_rate": 4.8960156230185905e-05,
991
+ "loss": 0.046,
992
+ "step": 82000
993
+ },
994
+ {
995
+ "epoch": 6.28,
996
+ "learning_rate": 4.895381571939436e-05,
997
+ "loss": 0.0441,
998
+ "step": 82500
999
+ },
1000
+ {
1001
+ "epoch": 6.32,
1002
+ "learning_rate": 4.8947475208602805e-05,
1003
+ "loss": 0.0483,
1004
+ "step": 83000
1005
+ },
1006
+ {
1007
+ "epoch": 6.35,
1008
+ "learning_rate": 4.894113469781126e-05,
1009
+ "loss": 0.0455,
1010
+ "step": 83500
1011
+ },
1012
+ {
1013
+ "epoch": 6.39,
1014
+ "learning_rate": 4.893479418701971e-05,
1015
+ "loss": 0.0552,
1016
+ "step": 84000
1017
+ },
1018
+ {
1019
+ "epoch": 6.43,
1020
+ "learning_rate": 4.892845367622816e-05,
1021
+ "loss": 0.0447,
1022
+ "step": 84500
1023
+ },
1024
+ {
1025
+ "epoch": 6.47,
1026
+ "learning_rate": 4.892211316543661e-05,
1027
+ "loss": 0.0519,
1028
+ "step": 85000
1029
+ },
1030
+ {
1031
+ "epoch": 6.51,
1032
+ "learning_rate": 4.891577265464506e-05,
1033
+ "loss": 0.0452,
1034
+ "step": 85500
1035
+ },
1036
+ {
1037
+ "epoch": 6.54,
1038
+ "learning_rate": 4.8909432143853513e-05,
1039
+ "loss": 0.0459,
1040
+ "step": 86000
1041
+ },
1042
+ {
1043
+ "epoch": 6.58,
1044
+ "learning_rate": 4.890309163306196e-05,
1045
+ "loss": 0.0484,
1046
+ "step": 86500
1047
+ },
1048
+ {
1049
+ "epoch": 6.62,
1050
+ "learning_rate": 4.8896751122270414e-05,
1051
+ "loss": 0.0459,
1052
+ "step": 87000
1053
+ },
1054
+ {
1055
+ "epoch": 6.66,
1056
+ "learning_rate": 4.889041061147887e-05,
1057
+ "loss": 0.0465,
1058
+ "step": 87500
1059
+ },
1060
+ {
1061
+ "epoch": 6.7,
1062
+ "learning_rate": 4.8884070100687314e-05,
1063
+ "loss": 0.0461,
1064
+ "step": 88000
1065
+ },
1066
+ {
1067
+ "epoch": 6.73,
1068
+ "learning_rate": 4.887772958989577e-05,
1069
+ "loss": 0.0542,
1070
+ "step": 88500
1071
+ },
1072
+ {
1073
+ "epoch": 6.77,
1074
+ "learning_rate": 4.8871389079104215e-05,
1075
+ "loss": 0.0453,
1076
+ "step": 89000
1077
+ },
1078
+ {
1079
+ "epoch": 6.81,
1080
+ "learning_rate": 4.886504856831267e-05,
1081
+ "loss": 0.0443,
1082
+ "step": 89500
1083
+ },
1084
+ {
1085
+ "epoch": 6.85,
1086
+ "learning_rate": 4.885870805752112e-05,
1087
+ "loss": 0.0496,
1088
+ "step": 90000
1089
+ },
1090
+ {
1091
+ "epoch": 6.89,
1092
+ "learning_rate": 4.885236754672957e-05,
1093
+ "loss": 0.0452,
1094
+ "step": 90500
1095
+ },
1096
+ {
1097
+ "epoch": 6.92,
1098
+ "learning_rate": 4.8846027035938016e-05,
1099
+ "loss": 0.0492,
1100
+ "step": 91000
1101
+ },
1102
+ {
1103
+ "epoch": 6.96,
1104
+ "learning_rate": 4.883968652514647e-05,
1105
+ "loss": 0.0472,
1106
+ "step": 91500
1107
+ },
1108
+ {
1109
+ "epoch": 7.0,
1110
+ "learning_rate": 4.8833346014354916e-05,
1111
+ "loss": 0.0466,
1112
+ "step": 92000
1113
+ },
1114
+ {
1115
+ "epoch": 7.04,
1116
+ "learning_rate": 4.882700550356337e-05,
1117
+ "loss": 0.0406,
1118
+ "step": 92500
1119
+ },
1120
+ {
1121
+ "epoch": 7.08,
1122
+ "learning_rate": 4.8820664992771816e-05,
1123
+ "loss": 0.0395,
1124
+ "step": 93000
1125
+ },
1126
+ {
1127
+ "epoch": 7.11,
1128
+ "learning_rate": 4.881432448198027e-05,
1129
+ "loss": 0.0432,
1130
+ "step": 93500
1131
+ },
1132
+ {
1133
+ "epoch": 7.15,
1134
+ "learning_rate": 4.880798397118872e-05,
1135
+ "loss": 0.0415,
1136
+ "step": 94000
1137
+ },
1138
+ {
1139
+ "epoch": 7.19,
1140
+ "learning_rate": 4.880164346039717e-05,
1141
+ "loss": 0.0466,
1142
+ "step": 94500
1143
+ },
1144
+ {
1145
+ "epoch": 7.23,
1146
+ "learning_rate": 4.879530294960562e-05,
1147
+ "loss": 0.0429,
1148
+ "step": 95000
1149
+ },
1150
+ {
1151
+ "epoch": 7.27,
1152
+ "learning_rate": 4.878896243881407e-05,
1153
+ "loss": 0.0434,
1154
+ "step": 95500
1155
+ },
1156
+ {
1157
+ "epoch": 7.3,
1158
+ "learning_rate": 4.8782621928022525e-05,
1159
+ "loss": 0.0472,
1160
+ "step": 96000
1161
+ },
1162
+ {
1163
+ "epoch": 7.34,
1164
+ "learning_rate": 4.877628141723097e-05,
1165
+ "loss": 0.041,
1166
+ "step": 96500
1167
+ },
1168
+ {
1169
+ "epoch": 7.38,
1170
+ "learning_rate": 4.8769940906439425e-05,
1171
+ "loss": 0.0429,
1172
+ "step": 97000
1173
+ },
1174
+ {
1175
+ "epoch": 7.42,
1176
+ "learning_rate": 4.876360039564787e-05,
1177
+ "loss": 0.04,
1178
+ "step": 97500
1179
+ },
1180
+ {
1181
+ "epoch": 7.46,
1182
+ "learning_rate": 4.8757259884856325e-05,
1183
+ "loss": 0.0439,
1184
+ "step": 98000
1185
+ },
1186
+ {
1187
+ "epoch": 7.49,
1188
+ "learning_rate": 4.875091937406478e-05,
1189
+ "loss": 0.0372,
1190
+ "step": 98500
1191
+ },
1192
+ {
1193
+ "epoch": 7.53,
1194
+ "learning_rate": 4.8744578863273226e-05,
1195
+ "loss": 0.0406,
1196
+ "step": 99000
1197
+ },
1198
+ {
1199
+ "epoch": 7.57,
1200
+ "learning_rate": 4.873823835248168e-05,
1201
+ "loss": 0.0422,
1202
+ "step": 99500
1203
+ },
1204
+ {
1205
+ "epoch": 7.61,
1206
+ "learning_rate": 4.8731897841690126e-05,
1207
+ "loss": 0.0406,
1208
+ "step": 100000
1209
+ },
1210
+ {
1211
+ "epoch": 7.65,
1212
+ "learning_rate": 4.872555733089858e-05,
1213
+ "loss": 0.0429,
1214
+ "step": 100500
1215
+ },
1216
+ {
1217
+ "epoch": 7.68,
1218
+ "learning_rate": 4.871921682010703e-05,
1219
+ "loss": 0.0438,
1220
+ "step": 101000
1221
+ },
1222
+ {
1223
+ "epoch": 7.72,
1224
+ "learning_rate": 4.871287630931548e-05,
1225
+ "loss": 0.0384,
1226
+ "step": 101500
1227
+ },
1228
+ {
1229
+ "epoch": 7.76,
1230
+ "learning_rate": 4.8706535798523934e-05,
1231
+ "loss": 0.0438,
1232
+ "step": 102000
1233
+ },
1234
+ {
1235
+ "epoch": 7.8,
1236
+ "learning_rate": 4.870019528773238e-05,
1237
+ "loss": 0.0399,
1238
+ "step": 102500
1239
+ },
1240
+ {
1241
+ "epoch": 7.84,
1242
+ "learning_rate": 4.8693854776940834e-05,
1243
+ "loss": 0.0419,
1244
+ "step": 103000
1245
+ },
1246
+ {
1247
+ "epoch": 7.87,
1248
+ "learning_rate": 4.868751426614928e-05,
1249
+ "loss": 0.0397,
1250
+ "step": 103500
1251
+ },
1252
+ {
1253
+ "epoch": 7.91,
1254
+ "learning_rate": 4.8681173755357735e-05,
1255
+ "loss": 0.0399,
1256
+ "step": 104000
1257
+ },
1258
+ {
1259
+ "epoch": 7.95,
1260
+ "learning_rate": 4.867483324456619e-05,
1261
+ "loss": 0.0388,
1262
+ "step": 104500
1263
+ },
1264
+ {
1265
+ "epoch": 7.99,
1266
+ "learning_rate": 4.8668492733774635e-05,
1267
+ "loss": 0.0381,
1268
+ "step": 105000
1269
+ },
1270
+ {
1271
+ "epoch": 8.03,
1272
+ "learning_rate": 4.866215222298309e-05,
1273
+ "loss": 0.0361,
1274
+ "step": 105500
1275
+ },
1276
+ {
1277
+ "epoch": 8.07,
1278
+ "learning_rate": 4.8655811712191536e-05,
1279
+ "loss": 0.0368,
1280
+ "step": 106000
1281
+ },
1282
+ {
1283
+ "epoch": 8.1,
1284
+ "learning_rate": 4.864947120139999e-05,
1285
+ "loss": 0.0389,
1286
+ "step": 106500
1287
+ },
1288
+ {
1289
+ "epoch": 8.14,
1290
+ "learning_rate": 4.8643130690608436e-05,
1291
+ "loss": 0.0402,
1292
+ "step": 107000
1293
+ },
1294
+ {
1295
+ "epoch": 8.18,
1296
+ "learning_rate": 4.863679017981689e-05,
1297
+ "loss": 0.0402,
1298
+ "step": 107500
1299
+ },
1300
+ {
1301
+ "epoch": 8.22,
1302
+ "learning_rate": 4.863044966902534e-05,
1303
+ "loss": 0.0368,
1304
+ "step": 108000
1305
+ },
1306
+ {
1307
+ "epoch": 8.26,
1308
+ "learning_rate": 4.862410915823379e-05,
1309
+ "loss": 0.0369,
1310
+ "step": 108500
1311
+ },
1312
+ {
1313
+ "epoch": 8.29,
1314
+ "learning_rate": 4.8617768647442244e-05,
1315
+ "loss": 0.0351,
1316
+ "step": 109000
1317
+ },
1318
+ {
1319
+ "epoch": 8.33,
1320
+ "learning_rate": 4.861142813665069e-05,
1321
+ "loss": 0.0383,
1322
+ "step": 109500
1323
+ },
1324
+ {
1325
+ "epoch": 8.37,
1326
+ "learning_rate": 4.8605087625859144e-05,
1327
+ "loss": 0.0379,
1328
+ "step": 110000
1329
+ },
1330
+ {
1331
+ "epoch": 8.41,
1332
+ "learning_rate": 4.85987471150676e-05,
1333
+ "loss": 0.036,
1334
+ "step": 110500
1335
+ },
1336
+ {
1337
+ "epoch": 8.45,
1338
+ "learning_rate": 4.8592406604276044e-05,
1339
+ "loss": 0.0393,
1340
+ "step": 111000
1341
+ },
1342
+ {
1343
+ "epoch": 8.48,
1344
+ "learning_rate": 4.85860660934845e-05,
1345
+ "loss": 0.0386,
1346
+ "step": 111500
1347
+ },
1348
+ {
1349
+ "epoch": 8.52,
1350
+ "learning_rate": 4.8579725582692945e-05,
1351
+ "loss": 0.0362,
1352
+ "step": 112000
1353
+ },
1354
+ {
1355
+ "epoch": 8.56,
1356
+ "learning_rate": 4.857338507190139e-05,
1357
+ "loss": 0.037,
1358
+ "step": 112500
1359
+ },
1360
+ {
1361
+ "epoch": 8.6,
1362
+ "learning_rate": 4.8567044561109845e-05,
1363
+ "loss": 0.0396,
1364
+ "step": 113000
1365
+ },
1366
+ {
1367
+ "epoch": 8.64,
1368
+ "learning_rate": 4.856070405031829e-05,
1369
+ "loss": 0.0354,
1370
+ "step": 113500
1371
+ },
1372
+ {
1373
+ "epoch": 8.67,
1374
+ "learning_rate": 4.8554363539526746e-05,
1375
+ "loss": 0.0331,
1376
+ "step": 114000
1377
+ },
1378
+ {
1379
+ "epoch": 8.71,
1380
+ "learning_rate": 4.854802302873519e-05,
1381
+ "loss": 0.0388,
1382
+ "step": 114500
1383
+ },
1384
+ {
1385
+ "epoch": 8.75,
1386
+ "learning_rate": 4.8541682517943646e-05,
1387
+ "loss": 0.0347,
1388
+ "step": 115000
1389
+ },
1390
+ {
1391
+ "epoch": 8.79,
1392
+ "learning_rate": 4.853534200715209e-05,
1393
+ "loss": 0.0379,
1394
+ "step": 115500
1395
+ },
1396
+ {
1397
+ "epoch": 8.83,
1398
+ "learning_rate": 4.8529001496360547e-05,
1399
+ "loss": 0.0341,
1400
+ "step": 116000
1401
+ },
1402
+ {
1403
+ "epoch": 8.86,
1404
+ "learning_rate": 4.8522660985569e-05,
1405
+ "loss": 0.0344,
1406
+ "step": 116500
1407
+ },
1408
+ {
1409
+ "epoch": 8.9,
1410
+ "learning_rate": 4.851632047477745e-05,
1411
+ "loss": 0.0365,
1412
+ "step": 117000
1413
+ },
1414
+ {
1415
+ "epoch": 8.94,
1416
+ "learning_rate": 4.85099799639859e-05,
1417
+ "loss": 0.0327,
1418
+ "step": 117500
1419
+ },
1420
+ {
1421
+ "epoch": 8.98,
1422
+ "learning_rate": 4.850363945319435e-05,
1423
+ "loss": 0.034,
1424
+ "step": 118000
1425
+ },
1426
+ {
1427
+ "epoch": 9.02,
1428
+ "learning_rate": 4.84972989424028e-05,
1429
+ "loss": 0.0349,
1430
+ "step": 118500
1431
+ },
1432
+ {
1433
+ "epoch": 9.05,
1434
+ "learning_rate": 4.8490958431611255e-05,
1435
+ "loss": 0.0294,
1436
+ "step": 119000
1437
+ },
1438
+ {
1439
+ "epoch": 9.09,
1440
+ "learning_rate": 4.84846179208197e-05,
1441
+ "loss": 0.0319,
1442
+ "step": 119500
1443
+ },
1444
+ {
1445
+ "epoch": 9.13,
1446
+ "learning_rate": 4.8478277410028155e-05,
1447
+ "loss": 0.031,
1448
+ "step": 120000
1449
+ },
1450
+ {
1451
+ "epoch": 9.17,
1452
+ "learning_rate": 4.84719368992366e-05,
1453
+ "loss": 0.0344,
1454
+ "step": 120500
1455
+ },
1456
+ {
1457
+ "epoch": 9.21,
1458
+ "learning_rate": 4.8465596388445055e-05,
1459
+ "loss": 0.0309,
1460
+ "step": 121000
1461
+ },
1462
+ {
1463
+ "epoch": 9.24,
1464
+ "learning_rate": 4.845925587765351e-05,
1465
+ "loss": 0.0312,
1466
+ "step": 121500
1467
+ },
1468
+ {
1469
+ "epoch": 9.28,
1470
+ "learning_rate": 4.8452915366861956e-05,
1471
+ "loss": 0.0314,
1472
+ "step": 122000
1473
+ },
1474
+ {
1475
+ "epoch": 9.32,
1476
+ "learning_rate": 4.844657485607041e-05,
1477
+ "loss": 0.0306,
1478
+ "step": 122500
1479
+ },
1480
+ {
1481
+ "epoch": 9.36,
1482
+ "learning_rate": 4.8440234345278856e-05,
1483
+ "loss": 0.0313,
1484
+ "step": 123000
1485
+ },
1486
+ {
1487
+ "epoch": 9.4,
1488
+ "learning_rate": 4.843389383448731e-05,
1489
+ "loss": 0.0361,
1490
+ "step": 123500
1491
+ },
1492
+ {
1493
+ "epoch": 9.43,
1494
+ "learning_rate": 4.842755332369576e-05,
1495
+ "loss": 0.0355,
1496
+ "step": 124000
1497
+ },
1498
+ {
1499
+ "epoch": 9.47,
1500
+ "learning_rate": 4.842121281290421e-05,
1501
+ "loss": 0.0323,
1502
+ "step": 124500
1503
+ },
1504
+ {
1505
+ "epoch": 9.51,
1506
+ "learning_rate": 4.8414872302112664e-05,
1507
+ "loss": 0.0329,
1508
+ "step": 125000
1509
+ },
1510
+ {
1511
+ "epoch": 9.55,
1512
+ "learning_rate": 4.840853179132111e-05,
1513
+ "loss": 0.0324,
1514
+ "step": 125500
1515
+ },
1516
+ {
1517
+ "epoch": 9.59,
1518
+ "learning_rate": 4.8402191280529564e-05,
1519
+ "loss": 0.0354,
1520
+ "step": 126000
1521
+ },
1522
+ {
1523
+ "epoch": 9.62,
1524
+ "learning_rate": 4.839585076973801e-05,
1525
+ "loss": 0.0367,
1526
+ "step": 126500
1527
+ },
1528
+ {
1529
+ "epoch": 9.66,
1530
+ "learning_rate": 4.8389510258946465e-05,
1531
+ "loss": 0.0306,
1532
+ "step": 127000
1533
+ },
1534
+ {
1535
+ "epoch": 9.7,
1536
+ "learning_rate": 4.838316974815492e-05,
1537
+ "loss": 0.0338,
1538
+ "step": 127500
1539
+ },
1540
+ {
1541
+ "epoch": 9.74,
1542
+ "learning_rate": 4.8376829237363365e-05,
1543
+ "loss": 0.0328,
1544
+ "step": 128000
1545
+ },
1546
+ {
1547
+ "epoch": 9.78,
1548
+ "learning_rate": 4.837048872657182e-05,
1549
+ "loss": 0.0343,
1550
+ "step": 128500
1551
+ },
1552
+ {
1553
+ "epoch": 9.82,
1554
+ "learning_rate": 4.8364148215780266e-05,
1555
+ "loss": 0.0324,
1556
+ "step": 129000
1557
+ },
1558
+ {
1559
+ "epoch": 9.85,
1560
+ "learning_rate": 4.835780770498872e-05,
1561
+ "loss": 0.0309,
1562
+ "step": 129500
1563
+ },
1564
+ {
1565
+ "epoch": 9.89,
1566
+ "learning_rate": 4.8351467194197166e-05,
1567
+ "loss": 0.0312,
1568
+ "step": 130000
1569
+ },
1570
+ {
1571
+ "epoch": 9.93,
1572
+ "learning_rate": 4.834512668340562e-05,
1573
+ "loss": 0.034,
1574
+ "step": 130500
1575
+ },
1576
+ {
1577
+ "epoch": 9.97,
1578
+ "learning_rate": 4.833878617261407e-05,
1579
+ "loss": 0.0307,
1580
+ "step": 131000
1581
+ },
1582
+ {
1583
+ "epoch": 10.01,
1584
+ "learning_rate": 4.833244566182252e-05,
1585
+ "loss": 0.0328,
1586
+ "step": 131500
1587
+ },
1588
+ {
1589
+ "epoch": 10.04,
1590
+ "learning_rate": 4.8326105151030974e-05,
1591
+ "loss": 0.0305,
1592
+ "step": 132000
1593
+ },
1594
+ {
1595
+ "epoch": 10.08,
1596
+ "learning_rate": 4.831976464023942e-05,
1597
+ "loss": 0.0297,
1598
+ "step": 132500
1599
+ },
1600
+ {
1601
+ "epoch": 10.12,
1602
+ "learning_rate": 4.831342412944787e-05,
1603
+ "loss": 0.0272,
1604
+ "step": 133000
1605
+ },
1606
+ {
1607
+ "epoch": 10.16,
1608
+ "learning_rate": 4.830708361865632e-05,
1609
+ "loss": 0.0274,
1610
+ "step": 133500
1611
+ },
1612
+ {
1613
+ "epoch": 10.2,
1614
+ "learning_rate": 4.830074310786477e-05,
1615
+ "loss": 0.0272,
1616
+ "step": 134000
1617
+ },
1618
+ {
1619
+ "epoch": 10.23,
1620
+ "learning_rate": 4.829440259707322e-05,
1621
+ "loss": 0.0261,
1622
+ "step": 134500
1623
+ },
1624
+ {
1625
+ "epoch": 10.27,
1626
+ "learning_rate": 4.828806208628167e-05,
1627
+ "loss": 0.0332,
1628
+ "step": 135000
1629
+ },
1630
+ {
1631
+ "epoch": 10.31,
1632
+ "learning_rate": 4.828172157549012e-05,
1633
+ "loss": 0.0265,
1634
+ "step": 135500
1635
+ },
1636
+ {
1637
+ "epoch": 10.35,
1638
+ "learning_rate": 4.8275381064698575e-05,
1639
+ "loss": 0.0269,
1640
+ "step": 136000
1641
+ },
1642
+ {
1643
+ "epoch": 10.39,
1644
+ "learning_rate": 4.826904055390702e-05,
1645
+ "loss": 0.0302,
1646
+ "step": 136500
1647
+ },
1648
+ {
1649
+ "epoch": 10.42,
1650
+ "learning_rate": 4.8262700043115476e-05,
1651
+ "loss": 0.0284,
1652
+ "step": 137000
1653
+ },
1654
+ {
1655
+ "epoch": 10.46,
1656
+ "learning_rate": 4.825635953232392e-05,
1657
+ "loss": 0.028,
1658
+ "step": 137500
1659
+ },
1660
+ {
1661
+ "epoch": 10.5,
1662
+ "learning_rate": 4.8250019021532376e-05,
1663
+ "loss": 0.0258,
1664
+ "step": 138000
1665
+ },
1666
+ {
1667
+ "epoch": 10.54,
1668
+ "learning_rate": 4.824367851074082e-05,
1669
+ "loss": 0.0273,
1670
+ "step": 138500
1671
+ },
1672
+ {
1673
+ "epoch": 10.58,
1674
+ "learning_rate": 4.823733799994928e-05,
1675
+ "loss": 0.0295,
1676
+ "step": 139000
1677
+ },
1678
+ {
1679
+ "epoch": 10.61,
1680
+ "learning_rate": 4.823099748915773e-05,
1681
+ "loss": 0.028,
1682
+ "step": 139500
1683
+ },
1684
+ {
1685
+ "epoch": 10.65,
1686
+ "learning_rate": 4.822465697836618e-05,
1687
+ "loss": 0.0291,
1688
+ "step": 140000
1689
+ },
1690
+ {
1691
+ "epoch": 10.69,
1692
+ "learning_rate": 4.821831646757463e-05,
1693
+ "loss": 0.0296,
1694
+ "step": 140500
1695
+ },
1696
+ {
1697
+ "epoch": 10.73,
1698
+ "learning_rate": 4.821197595678308e-05,
1699
+ "loss": 0.0283,
1700
+ "step": 141000
1701
+ },
1702
+ {
1703
+ "epoch": 10.77,
1704
+ "learning_rate": 4.820563544599153e-05,
1705
+ "loss": 0.0283,
1706
+ "step": 141500
1707
+ },
1708
+ {
1709
+ "epoch": 10.8,
1710
+ "learning_rate": 4.8199294935199985e-05,
1711
+ "loss": 0.0275,
1712
+ "step": 142000
1713
+ },
1714
+ {
1715
+ "epoch": 10.84,
1716
+ "learning_rate": 4.819295442440843e-05,
1717
+ "loss": 0.0284,
1718
+ "step": 142500
1719
+ },
1720
+ {
1721
+ "epoch": 10.88,
1722
+ "learning_rate": 4.8186613913616885e-05,
1723
+ "loss": 0.0274,
1724
+ "step": 143000
1725
+ },
1726
+ {
1727
+ "epoch": 10.92,
1728
+ "learning_rate": 4.818027340282533e-05,
1729
+ "loss": 0.0282,
1730
+ "step": 143500
1731
+ },
1732
+ {
1733
+ "epoch": 10.96,
1734
+ "learning_rate": 4.8173932892033786e-05,
1735
+ "loss": 0.0313,
1736
+ "step": 144000
1737
+ },
1738
+ {
1739
+ "epoch": 10.99,
1740
+ "learning_rate": 4.816759238124223e-05,
1741
+ "loss": 0.0283,
1742
+ "step": 144500
1743
+ },
1744
+ {
1745
+ "epoch": 11.03,
1746
+ "learning_rate": 4.8161251870450686e-05,
1747
+ "loss": 0.0231,
1748
+ "step": 145000
1749
+ },
1750
+ {
1751
+ "epoch": 11.07,
1752
+ "learning_rate": 4.815491135965914e-05,
1753
+ "loss": 0.0246,
1754
+ "step": 145500
1755
+ },
1756
+ {
1757
+ "epoch": 11.11,
1758
+ "learning_rate": 4.8148570848867586e-05,
1759
+ "loss": 0.0244,
1760
+ "step": 146000
1761
+ },
1762
+ {
1763
+ "epoch": 11.15,
1764
+ "learning_rate": 4.814223033807604e-05,
1765
+ "loss": 0.0245,
1766
+ "step": 146500
1767
+ },
1768
+ {
1769
+ "epoch": 11.18,
1770
+ "learning_rate": 4.813588982728449e-05,
1771
+ "loss": 0.0241,
1772
+ "step": 147000
1773
+ },
1774
+ {
1775
+ "epoch": 11.22,
1776
+ "learning_rate": 4.812954931649294e-05,
1777
+ "loss": 0.0262,
1778
+ "step": 147500
1779
+ },
1780
+ {
1781
+ "epoch": 11.26,
1782
+ "learning_rate": 4.8123208805701394e-05,
1783
+ "loss": 0.0247,
1784
+ "step": 148000
1785
+ },
1786
+ {
1787
+ "epoch": 11.3,
1788
+ "learning_rate": 4.811686829490984e-05,
1789
+ "loss": 0.0265,
1790
+ "step": 148500
1791
+ },
1792
+ {
1793
+ "epoch": 11.34,
1794
+ "learning_rate": 4.8110527784118294e-05,
1795
+ "loss": 0.0284,
1796
+ "step": 149000
1797
+ },
1798
+ {
1799
+ "epoch": 11.37,
1800
+ "learning_rate": 4.810418727332674e-05,
1801
+ "loss": 0.0252,
1802
+ "step": 149500
1803
+ },
1804
+ {
1805
+ "epoch": 11.41,
1806
+ "learning_rate": 4.8097846762535195e-05,
1807
+ "loss": 0.0251,
1808
+ "step": 150000
1809
+ },
1810
+ {
1811
+ "epoch": 11.45,
1812
+ "learning_rate": 4.809150625174365e-05,
1813
+ "loss": 0.0262,
1814
+ "step": 150500
1815
+ },
1816
+ {
1817
+ "epoch": 11.49,
1818
+ "learning_rate": 4.8085165740952095e-05,
1819
+ "loss": 0.0256,
1820
+ "step": 151000
1821
+ },
1822
+ {
1823
+ "epoch": 11.53,
1824
+ "learning_rate": 4.807882523016055e-05,
1825
+ "loss": 0.025,
1826
+ "step": 151500
1827
+ },
1828
+ {
1829
+ "epoch": 11.57,
1830
+ "learning_rate": 4.8072484719368996e-05,
1831
+ "loss": 0.0265,
1832
+ "step": 152000
1833
+ },
1834
+ {
1835
+ "epoch": 11.6,
1836
+ "learning_rate": 4.806614420857745e-05,
1837
+ "loss": 0.0268,
1838
+ "step": 152500
1839
+ },
1840
+ {
1841
+ "epoch": 11.64,
1842
+ "learning_rate": 4.8059803697785896e-05,
1843
+ "loss": 0.0277,
1844
+ "step": 153000
1845
+ },
1846
+ {
1847
+ "epoch": 11.68,
1848
+ "learning_rate": 4.805346318699434e-05,
1849
+ "loss": 0.0238,
1850
+ "step": 153500
1851
+ },
1852
+ {
1853
+ "epoch": 11.72,
1854
+ "learning_rate": 4.8047122676202797e-05,
1855
+ "loss": 0.0243,
1856
+ "step": 154000
1857
+ },
1858
+ {
1859
+ "epoch": 11.76,
1860
+ "learning_rate": 4.8040782165411243e-05,
1861
+ "loss": 0.0279,
1862
+ "step": 154500
1863
+ },
1864
+ {
1865
+ "epoch": 11.79,
1866
+ "learning_rate": 4.80344416546197e-05,
1867
+ "loss": 0.0269,
1868
+ "step": 155000
1869
+ },
1870
+ {
1871
+ "epoch": 11.83,
1872
+ "learning_rate": 4.8028101143828144e-05,
1873
+ "loss": 0.0249,
1874
+ "step": 155500
1875
+ },
1876
+ {
1877
+ "epoch": 11.87,
1878
+ "learning_rate": 4.80217606330366e-05,
1879
+ "loss": 0.0256,
1880
+ "step": 156000
1881
+ },
1882
+ {
1883
+ "epoch": 11.91,
1884
+ "learning_rate": 4.801542012224505e-05,
1885
+ "loss": 0.025,
1886
+ "step": 156500
1887
+ },
1888
+ {
1889
+ "epoch": 11.95,
1890
+ "learning_rate": 4.80090796114535e-05,
1891
+ "loss": 0.0254,
1892
+ "step": 157000
1893
+ },
1894
+ {
1895
+ "epoch": 11.98,
1896
+ "learning_rate": 4.800273910066195e-05,
1897
+ "loss": 0.0232,
1898
+ "step": 157500
1899
+ },
1900
+ {
1901
+ "epoch": 12.02,
1902
+ "learning_rate": 4.79963985898704e-05,
1903
+ "loss": 0.0217,
1904
+ "step": 158000
1905
+ },
1906
+ {
1907
+ "epoch": 12.06,
1908
+ "learning_rate": 4.799005807907885e-05,
1909
+ "loss": 0.0212,
1910
+ "step": 158500
1911
+ },
1912
+ {
1913
+ "epoch": 12.1,
1914
+ "learning_rate": 4.79837175682873e-05,
1915
+ "loss": 0.0232,
1916
+ "step": 159000
1917
+ },
1918
+ {
1919
+ "epoch": 12.14,
1920
+ "learning_rate": 4.797737705749575e-05,
1921
+ "loss": 0.0203,
1922
+ "step": 159500
1923
+ },
1924
+ {
1925
+ "epoch": 12.17,
1926
+ "learning_rate": 4.7971036546704206e-05,
1927
+ "loss": 0.0221,
1928
+ "step": 160000
1929
+ },
1930
+ {
1931
+ "epoch": 12.21,
1932
+ "learning_rate": 4.796469603591265e-05,
1933
+ "loss": 0.024,
1934
+ "step": 160500
1935
+ },
1936
+ {
1937
+ "epoch": 12.25,
1938
+ "learning_rate": 4.7958355525121106e-05,
1939
+ "loss": 0.0228,
1940
+ "step": 161000
1941
+ },
1942
+ {
1943
+ "epoch": 12.29,
1944
+ "learning_rate": 4.795201501432955e-05,
1945
+ "loss": 0.0241,
1946
+ "step": 161500
1947
+ },
1948
+ {
1949
+ "epoch": 12.33,
1950
+ "learning_rate": 4.794567450353801e-05,
1951
+ "loss": 0.0225,
1952
+ "step": 162000
1953
+ },
1954
+ {
1955
+ "epoch": 12.36,
1956
+ "learning_rate": 4.793933399274646e-05,
1957
+ "loss": 0.0227,
1958
+ "step": 162500
1959
+ },
1960
+ {
1961
+ "epoch": 12.4,
1962
+ "learning_rate": 4.793299348195491e-05,
1963
+ "loss": 0.0229,
1964
+ "step": 163000
1965
+ },
1966
+ {
1967
+ "epoch": 12.44,
1968
+ "learning_rate": 4.792665297116336e-05,
1969
+ "loss": 0.0253,
1970
+ "step": 163500
1971
+ },
1972
+ {
1973
+ "epoch": 12.48,
1974
+ "learning_rate": 4.792031246037181e-05,
1975
+ "loss": 0.0224,
1976
+ "step": 164000
1977
+ },
1978
+ {
1979
+ "epoch": 12.52,
1980
+ "learning_rate": 4.791397194958026e-05,
1981
+ "loss": 0.0265,
1982
+ "step": 164500
1983
+ },
1984
+ {
1985
+ "epoch": 12.55,
1986
+ "learning_rate": 4.7907631438788715e-05,
1987
+ "loss": 0.0216,
1988
+ "step": 165000
1989
+ },
1990
+ {
1991
+ "epoch": 12.59,
1992
+ "learning_rate": 4.790129092799716e-05,
1993
+ "loss": 0.0229,
1994
+ "step": 165500
1995
+ },
1996
+ {
1997
+ "epoch": 12.63,
1998
+ "learning_rate": 4.7894950417205615e-05,
1999
+ "loss": 0.0237,
2000
+ "step": 166000
2001
+ },
2002
+ {
2003
+ "epoch": 12.67,
2004
+ "learning_rate": 4.788860990641406e-05,
2005
+ "loss": 0.0233,
2006
+ "step": 166500
2007
+ },
2008
+ {
2009
+ "epoch": 12.71,
2010
+ "learning_rate": 4.7882269395622516e-05,
2011
+ "loss": 0.0246,
2012
+ "step": 167000
2013
+ },
2014
+ {
2015
+ "epoch": 12.74,
2016
+ "learning_rate": 4.787592888483096e-05,
2017
+ "loss": 0.0233,
2018
+ "step": 167500
2019
+ },
2020
+ {
2021
+ "epoch": 12.78,
2022
+ "learning_rate": 4.7869588374039416e-05,
2023
+ "loss": 0.0228,
2024
+ "step": 168000
2025
+ },
2026
+ {
2027
+ "epoch": 12.82,
2028
+ "learning_rate": 4.786324786324787e-05,
2029
+ "loss": 0.0232,
2030
+ "step": 168500
2031
+ },
2032
+ {
2033
+ "epoch": 12.86,
2034
+ "learning_rate": 4.7856907352456317e-05,
2035
+ "loss": 0.021,
2036
+ "step": 169000
2037
+ },
2038
+ {
2039
+ "epoch": 12.9,
2040
+ "learning_rate": 4.785056684166477e-05,
2041
+ "loss": 0.0236,
2042
+ "step": 169500
2043
+ },
2044
+ {
2045
+ "epoch": 12.93,
2046
+ "learning_rate": 4.784422633087322e-05,
2047
+ "loss": 0.0225,
2048
+ "step": 170000
2049
+ },
2050
+ {
2051
+ "epoch": 12.97,
2052
+ "learning_rate": 4.783788582008167e-05,
2053
+ "loss": 0.0209,
2054
+ "step": 170500
2055
+ },
2056
+ {
2057
+ "epoch": 13.01,
2058
+ "learning_rate": 4.7831545309290124e-05,
2059
+ "loss": 0.0216,
2060
+ "step": 171000
2061
+ },
2062
+ {
2063
+ "epoch": 13.05,
2064
+ "learning_rate": 4.782520479849857e-05,
2065
+ "loss": 0.0201,
2066
+ "step": 171500
2067
+ },
2068
+ {
2069
+ "epoch": 13.09,
2070
+ "learning_rate": 4.7818864287707025e-05,
2071
+ "loss": 0.0203,
2072
+ "step": 172000
2073
+ },
2074
+ {
2075
+ "epoch": 13.12,
2076
+ "learning_rate": 4.781252377691547e-05,
2077
+ "loss": 0.0186,
2078
+ "step": 172500
2079
+ },
2080
+ {
2081
+ "epoch": 13.16,
2082
+ "learning_rate": 4.7806183266123925e-05,
2083
+ "loss": 0.0206,
2084
+ "step": 173000
2085
+ },
2086
+ {
2087
+ "epoch": 13.2,
2088
+ "learning_rate": 4.779984275533237e-05,
2089
+ "loss": 0.0198,
2090
+ "step": 173500
2091
+ },
2092
+ {
2093
+ "epoch": 13.24,
2094
+ "learning_rate": 4.779350224454082e-05,
2095
+ "loss": 0.02,
2096
+ "step": 174000
2097
+ },
2098
+ {
2099
+ "epoch": 13.28,
2100
+ "learning_rate": 4.778716173374927e-05,
2101
+ "loss": 0.0224,
2102
+ "step": 174500
2103
+ },
2104
+ {
2105
+ "epoch": 13.32,
2106
+ "learning_rate": 4.778082122295772e-05,
2107
+ "loss": 0.0206,
2108
+ "step": 175000
2109
+ },
2110
+ {
2111
+ "epoch": 13.35,
2112
+ "learning_rate": 4.777448071216617e-05,
2113
+ "loss": 0.0207,
2114
+ "step": 175500
2115
+ },
2116
+ {
2117
+ "epoch": 13.39,
2118
+ "learning_rate": 4.776814020137462e-05,
2119
+ "loss": 0.0214,
2120
+ "step": 176000
2121
+ },
2122
+ {
2123
+ "epoch": 13.43,
2124
+ "learning_rate": 4.776179969058307e-05,
2125
+ "loss": 0.0196,
2126
+ "step": 176500
2127
+ },
2128
+ {
2129
+ "epoch": 13.47,
2130
+ "learning_rate": 4.775545917979153e-05,
2131
+ "loss": 0.0202,
2132
+ "step": 177000
2133
+ },
2134
+ {
2135
+ "epoch": 13.51,
2136
+ "learning_rate": 4.7749118668999974e-05,
2137
+ "loss": 0.0193,
2138
+ "step": 177500
2139
+ },
2140
+ {
2141
+ "epoch": 13.54,
2142
+ "learning_rate": 4.774277815820843e-05,
2143
+ "loss": 0.0205,
2144
+ "step": 178000
2145
+ },
2146
+ {
2147
+ "epoch": 13.58,
2148
+ "learning_rate": 4.7736437647416874e-05,
2149
+ "loss": 0.0204,
2150
+ "step": 178500
2151
+ },
2152
+ {
2153
+ "epoch": 13.62,
2154
+ "learning_rate": 4.773009713662533e-05,
2155
+ "loss": 0.0193,
2156
+ "step": 179000
2157
+ },
2158
+ {
2159
+ "epoch": 13.66,
2160
+ "learning_rate": 4.772375662583378e-05,
2161
+ "loss": 0.0224,
2162
+ "step": 179500
2163
+ },
2164
+ {
2165
+ "epoch": 13.7,
2166
+ "learning_rate": 4.771741611504223e-05,
2167
+ "loss": 0.0193,
2168
+ "step": 180000
2169
+ },
2170
+ {
2171
+ "epoch": 13.73,
2172
+ "learning_rate": 4.771107560425068e-05,
2173
+ "loss": 0.0194,
2174
+ "step": 180500
2175
+ },
2176
+ {
2177
+ "epoch": 13.77,
2178
+ "learning_rate": 4.770473509345913e-05,
2179
+ "loss": 0.0202,
2180
+ "step": 181000
2181
+ },
2182
+ {
2183
+ "epoch": 13.81,
2184
+ "learning_rate": 4.769839458266758e-05,
2185
+ "loss": 0.0204,
2186
+ "step": 181500
2187
+ },
2188
+ {
2189
+ "epoch": 13.85,
2190
+ "learning_rate": 4.769205407187603e-05,
2191
+ "loss": 0.0183,
2192
+ "step": 182000
2193
+ },
2194
+ {
2195
+ "epoch": 13.89,
2196
+ "learning_rate": 4.768571356108448e-05,
2197
+ "loss": 0.0199,
2198
+ "step": 182500
2199
+ },
2200
+ {
2201
+ "epoch": 13.92,
2202
+ "learning_rate": 4.7679373050292936e-05,
2203
+ "loss": 0.0206,
2204
+ "step": 183000
2205
+ },
2206
+ {
2207
+ "epoch": 13.96,
2208
+ "learning_rate": 4.767303253950138e-05,
2209
+ "loss": 0.0205,
2210
+ "step": 183500
2211
+ },
2212
+ {
2213
+ "epoch": 14.0,
2214
+ "learning_rate": 4.7666692028709836e-05,
2215
+ "loss": 0.0195,
2216
+ "step": 184000
2217
+ },
2218
+ {
2219
+ "epoch": 14.04,
2220
+ "learning_rate": 4.766035151791828e-05,
2221
+ "loss": 0.0169,
2222
+ "step": 184500
2223
+ },
2224
+ {
2225
+ "epoch": 14.08,
2226
+ "learning_rate": 4.765401100712674e-05,
2227
+ "loss": 0.0175,
2228
+ "step": 185000
2229
+ },
2230
+ {
2231
+ "epoch": 14.11,
2232
+ "learning_rate": 4.764767049633519e-05,
2233
+ "loss": 0.0161,
2234
+ "step": 185500
2235
+ },
2236
+ {
2237
+ "epoch": 14.15,
2238
+ "learning_rate": 4.764132998554364e-05,
2239
+ "loss": 0.0173,
2240
+ "step": 186000
2241
+ },
2242
+ {
2243
+ "epoch": 14.19,
2244
+ "learning_rate": 4.763498947475209e-05,
2245
+ "loss": 0.0169,
2246
+ "step": 186500
2247
+ },
2248
+ {
2249
+ "epoch": 14.23,
2250
+ "learning_rate": 4.762864896396054e-05,
2251
+ "loss": 0.0153,
2252
+ "step": 187000
2253
+ },
2254
+ {
2255
+ "epoch": 14.27,
2256
+ "learning_rate": 4.762230845316899e-05,
2257
+ "loss": 0.0177,
2258
+ "step": 187500
2259
+ },
2260
+ {
2261
+ "epoch": 14.3,
2262
+ "learning_rate": 4.7615967942377445e-05,
2263
+ "loss": 0.0177,
2264
+ "step": 188000
2265
+ },
2266
+ {
2267
+ "epoch": 14.34,
2268
+ "learning_rate": 4.760962743158589e-05,
2269
+ "loss": 0.0169,
2270
+ "step": 188500
2271
+ },
2272
+ {
2273
+ "epoch": 14.38,
2274
+ "learning_rate": 4.7603286920794345e-05,
2275
+ "loss": 0.0181,
2276
+ "step": 189000
2277
+ },
2278
+ {
2279
+ "epoch": 14.42,
2280
+ "learning_rate": 4.759694641000279e-05,
2281
+ "loss": 0.0172,
2282
+ "step": 189500
2283
+ },
2284
+ {
2285
+ "epoch": 14.46,
2286
+ "learning_rate": 4.7590605899211246e-05,
2287
+ "loss": 0.0187,
2288
+ "step": 190000
2289
+ },
2290
+ {
2291
+ "epoch": 14.49,
2292
+ "learning_rate": 4.758426538841969e-05,
2293
+ "loss": 0.0196,
2294
+ "step": 190500
2295
+ },
2296
+ {
2297
+ "epoch": 14.53,
2298
+ "learning_rate": 4.7577924877628146e-05,
2299
+ "loss": 0.0175,
2300
+ "step": 191000
2301
+ },
2302
+ {
2303
+ "epoch": 14.57,
2304
+ "learning_rate": 4.75715843668366e-05,
2305
+ "loss": 0.0168,
2306
+ "step": 191500
2307
+ },
2308
+ {
2309
+ "epoch": 14.61,
2310
+ "learning_rate": 4.756524385604505e-05,
2311
+ "loss": 0.0182,
2312
+ "step": 192000
2313
+ },
2314
+ {
2315
+ "epoch": 14.65,
2316
+ "learning_rate": 4.75589033452535e-05,
2317
+ "loss": 0.0179,
2318
+ "step": 192500
2319
+ },
2320
+ {
2321
+ "epoch": 14.68,
2322
+ "learning_rate": 4.755256283446195e-05,
2323
+ "loss": 0.0182,
2324
+ "step": 193000
2325
+ },
2326
+ {
2327
+ "epoch": 14.72,
2328
+ "learning_rate": 4.75462223236704e-05,
2329
+ "loss": 0.0199,
2330
+ "step": 193500
2331
+ },
2332
+ {
2333
+ "epoch": 14.76,
2334
+ "learning_rate": 4.753988181287885e-05,
2335
+ "loss": 0.0199,
2336
+ "step": 194000
2337
+ },
2338
+ {
2339
+ "epoch": 14.8,
2340
+ "learning_rate": 4.75335413020873e-05,
2341
+ "loss": 0.018,
2342
+ "step": 194500
2343
+ },
2344
+ {
2345
+ "epoch": 14.84,
2346
+ "learning_rate": 4.752720079129575e-05,
2347
+ "loss": 0.0205,
2348
+ "step": 195000
2349
+ },
2350
+ {
2351
+ "epoch": 14.87,
2352
+ "learning_rate": 4.7520860280504195e-05,
2353
+ "loss": 0.0192,
2354
+ "step": 195500
2355
+ },
2356
+ {
2357
+ "epoch": 14.91,
2358
+ "learning_rate": 4.751451976971265e-05,
2359
+ "loss": 0.018,
2360
+ "step": 196000
2361
+ },
2362
+ {
2363
+ "epoch": 14.95,
2364
+ "learning_rate": 4.7508179258921095e-05,
2365
+ "loss": 0.0195,
2366
+ "step": 196500
2367
+ },
2368
+ {
2369
+ "epoch": 14.99,
2370
+ "learning_rate": 4.750183874812955e-05,
2371
+ "loss": 0.019,
2372
+ "step": 197000
2373
+ },
2374
+ {
2375
+ "epoch": 15.03,
2376
+ "learning_rate": 4.7495498237338e-05,
2377
+ "loss": 0.0162,
2378
+ "step": 197500
2379
+ },
2380
+ {
2381
+ "epoch": 15.07,
2382
+ "learning_rate": 4.748915772654645e-05,
2383
+ "loss": 0.0157,
2384
+ "step": 198000
2385
+ },
2386
+ {
2387
+ "epoch": 15.1,
2388
+ "learning_rate": 4.74828172157549e-05,
2389
+ "loss": 0.014,
2390
+ "step": 198500
2391
+ },
2392
+ {
2393
+ "epoch": 15.14,
2394
+ "learning_rate": 4.747647670496335e-05,
2395
+ "loss": 0.0147,
2396
+ "step": 199000
2397
+ },
2398
+ {
2399
+ "epoch": 15.18,
2400
+ "learning_rate": 4.74701361941718e-05,
2401
+ "loss": 0.0152,
2402
+ "step": 199500
2403
+ },
2404
+ {
2405
+ "epoch": 15.22,
2406
+ "learning_rate": 4.746379568338026e-05,
2407
+ "loss": 0.0155,
2408
+ "step": 200000
2409
+ },
2410
+ {
2411
+ "epoch": 15.26,
2412
+ "learning_rate": 4.7457455172588704e-05,
2413
+ "loss": 0.0163,
2414
+ "step": 200500
2415
+ },
2416
+ {
2417
+ "epoch": 15.29,
2418
+ "learning_rate": 4.745111466179716e-05,
2419
+ "loss": 0.0167,
2420
+ "step": 201000
2421
+ },
2422
+ {
2423
+ "epoch": 15.33,
2424
+ "learning_rate": 4.7444774151005604e-05,
2425
+ "loss": 0.0154,
2426
+ "step": 201500
2427
+ },
2428
+ {
2429
+ "epoch": 15.37,
2430
+ "learning_rate": 4.743843364021406e-05,
2431
+ "loss": 0.0157,
2432
+ "step": 202000
2433
+ },
2434
+ {
2435
+ "epoch": 15.41,
2436
+ "learning_rate": 4.743209312942251e-05,
2437
+ "loss": 0.0155,
2438
+ "step": 202500
2439
+ },
2440
+ {
2441
+ "epoch": 15.45,
2442
+ "learning_rate": 4.742575261863096e-05,
2443
+ "loss": 0.0161,
2444
+ "step": 203000
2445
+ },
2446
+ {
2447
+ "epoch": 15.48,
2448
+ "learning_rate": 4.741941210783941e-05,
2449
+ "loss": 0.014,
2450
+ "step": 203500
2451
+ },
2452
+ {
2453
+ "epoch": 15.52,
2454
+ "learning_rate": 4.741307159704786e-05,
2455
+ "loss": 0.0151,
2456
+ "step": 204000
2457
+ },
2458
+ {
2459
+ "epoch": 15.56,
2460
+ "learning_rate": 4.740673108625631e-05,
2461
+ "loss": 0.0177,
2462
+ "step": 204500
2463
+ },
2464
+ {
2465
+ "epoch": 15.6,
2466
+ "learning_rate": 4.740039057546476e-05,
2467
+ "loss": 0.0164,
2468
+ "step": 205000
2469
+ },
2470
+ {
2471
+ "epoch": 15.64,
2472
+ "learning_rate": 4.739405006467321e-05,
2473
+ "loss": 0.0172,
2474
+ "step": 205500
2475
+ },
2476
+ {
2477
+ "epoch": 15.67,
2478
+ "learning_rate": 4.7387709553881666e-05,
2479
+ "loss": 0.0156,
2480
+ "step": 206000
2481
+ },
2482
+ {
2483
+ "epoch": 15.71,
2484
+ "learning_rate": 4.738136904309011e-05,
2485
+ "loss": 0.0175,
2486
+ "step": 206500
2487
+ },
2488
+ {
2489
+ "epoch": 15.75,
2490
+ "learning_rate": 4.7375028532298567e-05,
2491
+ "loss": 0.016,
2492
+ "step": 207000
2493
+ },
2494
+ {
2495
+ "epoch": 15.79,
2496
+ "learning_rate": 4.7368688021507013e-05,
2497
+ "loss": 0.0168,
2498
+ "step": 207500
2499
+ },
2500
+ {
2501
+ "epoch": 15.83,
2502
+ "learning_rate": 4.736234751071547e-05,
2503
+ "loss": 0.0177,
2504
+ "step": 208000
2505
+ },
2506
+ {
2507
+ "epoch": 15.86,
2508
+ "learning_rate": 4.735600699992392e-05,
2509
+ "loss": 0.0165,
2510
+ "step": 208500
2511
+ },
2512
+ {
2513
+ "epoch": 15.9,
2514
+ "learning_rate": 4.734966648913237e-05,
2515
+ "loss": 0.0154,
2516
+ "step": 209000
2517
+ },
2518
+ {
2519
+ "epoch": 15.94,
2520
+ "learning_rate": 4.734332597834082e-05,
2521
+ "loss": 0.0165,
2522
+ "step": 209500
2523
+ },
2524
+ {
2525
+ "epoch": 15.98,
2526
+ "learning_rate": 4.733698546754927e-05,
2527
+ "loss": 0.0177,
2528
+ "step": 210000
2529
+ },
2530
+ {
2531
+ "epoch": 16.02,
2532
+ "learning_rate": 4.733064495675772e-05,
2533
+ "loss": 0.0163,
2534
+ "step": 210500
2535
+ },
2536
+ {
2537
+ "epoch": 16.05,
2538
+ "learning_rate": 4.732430444596617e-05,
2539
+ "loss": 0.0128,
2540
+ "step": 211000
2541
+ },
2542
+ {
2543
+ "epoch": 16.09,
2544
+ "learning_rate": 4.731796393517462e-05,
2545
+ "loss": 0.0127,
2546
+ "step": 211500
2547
+ },
2548
+ {
2549
+ "epoch": 16.13,
2550
+ "learning_rate": 4.7311623424383075e-05,
2551
+ "loss": 0.0128,
2552
+ "step": 212000
2553
+ },
2554
+ {
2555
+ "epoch": 16.17,
2556
+ "learning_rate": 4.730528291359152e-05,
2557
+ "loss": 0.0132,
2558
+ "step": 212500
2559
+ },
2560
+ {
2561
+ "epoch": 16.21,
2562
+ "learning_rate": 4.7298942402799976e-05,
2563
+ "loss": 0.0136,
2564
+ "step": 213000
2565
+ },
2566
+ {
2567
+ "epoch": 16.24,
2568
+ "learning_rate": 4.729260189200842e-05,
2569
+ "loss": 0.0126,
2570
+ "step": 213500
2571
+ },
2572
+ {
2573
+ "epoch": 16.28,
2574
+ "learning_rate": 4.7286261381216876e-05,
2575
+ "loss": 0.0139,
2576
+ "step": 214000
2577
+ },
2578
+ {
2579
+ "epoch": 16.32,
2580
+ "learning_rate": 4.727992087042532e-05,
2581
+ "loss": 0.012,
2582
+ "step": 214500
2583
+ },
2584
+ {
2585
+ "epoch": 16.36,
2586
+ "learning_rate": 4.727358035963378e-05,
2587
+ "loss": 0.0142,
2588
+ "step": 215000
2589
+ },
2590
+ {
2591
+ "epoch": 16.4,
2592
+ "learning_rate": 4.7267239848842224e-05,
2593
+ "loss": 0.0129,
2594
+ "step": 215500
2595
+ },
2596
+ {
2597
+ "epoch": 16.43,
2598
+ "learning_rate": 4.726089933805067e-05,
2599
+ "loss": 0.0134,
2600
+ "step": 216000
2601
+ },
2602
+ {
2603
+ "epoch": 16.47,
2604
+ "learning_rate": 4.7254558827259124e-05,
2605
+ "loss": 0.0138,
2606
+ "step": 216500
2607
+ },
2608
+ {
2609
+ "epoch": 16.51,
2610
+ "learning_rate": 4.724821831646758e-05,
2611
+ "loss": 0.014,
2612
+ "step": 217000
2613
+ },
2614
+ {
2615
+ "epoch": 16.55,
2616
+ "learning_rate": 4.7241877805676024e-05,
2617
+ "loss": 0.0131,
2618
+ "step": 217500
2619
+ },
2620
+ {
2621
+ "epoch": 16.59,
2622
+ "learning_rate": 4.723553729488448e-05,
2623
+ "loss": 0.0138,
2624
+ "step": 218000
2625
+ },
2626
+ {
2627
+ "epoch": 16.62,
2628
+ "learning_rate": 4.7229196784092925e-05,
2629
+ "loss": 0.0144,
2630
+ "step": 218500
2631
+ },
2632
+ {
2633
+ "epoch": 16.66,
2634
+ "learning_rate": 4.722285627330138e-05,
2635
+ "loss": 0.0141,
2636
+ "step": 219000
2637
+ },
2638
+ {
2639
+ "epoch": 16.7,
2640
+ "learning_rate": 4.7216515762509825e-05,
2641
+ "loss": 0.0158,
2642
+ "step": 219500
2643
+ },
2644
+ {
2645
+ "epoch": 16.74,
2646
+ "learning_rate": 4.721017525171828e-05,
2647
+ "loss": 0.0153,
2648
+ "step": 220000
2649
+ },
2650
+ {
2651
+ "epoch": 16.78,
2652
+ "learning_rate": 4.720383474092673e-05,
2653
+ "loss": 0.014,
2654
+ "step": 220500
2655
+ },
2656
+ {
2657
+ "epoch": 16.82,
2658
+ "learning_rate": 4.719749423013518e-05,
2659
+ "loss": 0.0144,
2660
+ "step": 221000
2661
+ },
2662
+ {
2663
+ "epoch": 16.85,
2664
+ "learning_rate": 4.719115371934363e-05,
2665
+ "loss": 0.0142,
2666
+ "step": 221500
2667
+ },
2668
+ {
2669
+ "epoch": 16.89,
2670
+ "learning_rate": 4.718481320855208e-05,
2671
+ "loss": 0.0148,
2672
+ "step": 222000
2673
+ },
2674
+ {
2675
+ "epoch": 16.93,
2676
+ "learning_rate": 4.717847269776053e-05,
2677
+ "loss": 0.0162,
2678
+ "step": 222500
2679
+ },
2680
+ {
2681
+ "epoch": 16.97,
2682
+ "learning_rate": 4.717213218696899e-05,
2683
+ "loss": 0.015,
2684
+ "step": 223000
2685
+ },
2686
+ {
2687
+ "epoch": 17.01,
2688
+ "learning_rate": 4.7165791676177434e-05,
2689
+ "loss": 0.0136,
2690
+ "step": 223500
2691
+ },
2692
+ {
2693
+ "epoch": 17.04,
2694
+ "learning_rate": 4.715945116538589e-05,
2695
+ "loss": 0.0124,
2696
+ "step": 224000
2697
+ },
2698
+ {
2699
+ "epoch": 17.08,
2700
+ "learning_rate": 4.7153110654594334e-05,
2701
+ "loss": 0.0107,
2702
+ "step": 224500
2703
+ },
2704
+ {
2705
+ "epoch": 17.12,
2706
+ "learning_rate": 4.714677014380279e-05,
2707
+ "loss": 0.0125,
2708
+ "step": 225000
2709
+ },
2710
+ {
2711
+ "epoch": 17.16,
2712
+ "learning_rate": 4.7140429633011235e-05,
2713
+ "loss": 0.012,
2714
+ "step": 225500
2715
+ },
2716
+ {
2717
+ "epoch": 17.2,
2718
+ "learning_rate": 4.713408912221969e-05,
2719
+ "loss": 0.0119,
2720
+ "step": 226000
2721
+ },
2722
+ {
2723
+ "epoch": 17.23,
2724
+ "learning_rate": 4.712774861142814e-05,
2725
+ "loss": 0.0117,
2726
+ "step": 226500
2727
+ },
2728
+ {
2729
+ "epoch": 17.27,
2730
+ "learning_rate": 4.712140810063659e-05,
2731
+ "loss": 0.0098,
2732
+ "step": 227000
2733
+ },
2734
+ {
2735
+ "epoch": 17.31,
2736
+ "learning_rate": 4.711506758984504e-05,
2737
+ "loss": 0.0114,
2738
+ "step": 227500
2739
+ },
2740
+ {
2741
+ "epoch": 17.35,
2742
+ "learning_rate": 4.710872707905349e-05,
2743
+ "loss": 0.0121,
2744
+ "step": 228000
2745
+ },
2746
+ {
2747
+ "epoch": 17.39,
2748
+ "learning_rate": 4.710238656826194e-05,
2749
+ "loss": 0.0112,
2750
+ "step": 228500
2751
+ },
2752
+ {
2753
+ "epoch": 17.42,
2754
+ "learning_rate": 4.7096046057470396e-05,
2755
+ "loss": 0.0121,
2756
+ "step": 229000
2757
+ },
2758
+ {
2759
+ "epoch": 17.46,
2760
+ "learning_rate": 4.708970554667884e-05,
2761
+ "loss": 0.0119,
2762
+ "step": 229500
2763
+ },
2764
+ {
2765
+ "epoch": 17.5,
2766
+ "learning_rate": 4.70833650358873e-05,
2767
+ "loss": 0.0128,
2768
+ "step": 230000
2769
+ },
2770
+ {
2771
+ "epoch": 17.54,
2772
+ "learning_rate": 4.7077024525095744e-05,
2773
+ "loss": 0.0109,
2774
+ "step": 230500
2775
+ },
2776
+ {
2777
+ "epoch": 17.58,
2778
+ "learning_rate": 4.70706840143042e-05,
2779
+ "loss": 0.0118,
2780
+ "step": 231000
2781
+ },
2782
+ {
2783
+ "epoch": 17.61,
2784
+ "learning_rate": 4.706434350351265e-05,
2785
+ "loss": 0.0121,
2786
+ "step": 231500
2787
+ },
2788
+ {
2789
+ "epoch": 17.65,
2790
+ "learning_rate": 4.70580029927211e-05,
2791
+ "loss": 0.013,
2792
+ "step": 232000
2793
+ },
2794
+ {
2795
+ "epoch": 17.69,
2796
+ "learning_rate": 4.705166248192955e-05,
2797
+ "loss": 0.013,
2798
+ "step": 232500
2799
+ },
2800
+ {
2801
+ "epoch": 17.73,
2802
+ "learning_rate": 4.7045321971138e-05,
2803
+ "loss": 0.0119,
2804
+ "step": 233000
2805
+ },
2806
+ {
2807
+ "epoch": 17.77,
2808
+ "learning_rate": 4.703898146034645e-05,
2809
+ "loss": 0.0118,
2810
+ "step": 233500
2811
+ },
2812
+ {
2813
+ "epoch": 17.8,
2814
+ "learning_rate": 4.70326409495549e-05,
2815
+ "loss": 0.0134,
2816
+ "step": 234000
2817
+ },
2818
+ {
2819
+ "epoch": 17.84,
2820
+ "learning_rate": 4.702630043876335e-05,
2821
+ "loss": 0.0125,
2822
+ "step": 234500
2823
+ },
2824
+ {
2825
+ "epoch": 17.88,
2826
+ "learning_rate": 4.70199599279718e-05,
2827
+ "loss": 0.0122,
2828
+ "step": 235000
2829
+ },
2830
+ {
2831
+ "epoch": 17.92,
2832
+ "learning_rate": 4.701361941718025e-05,
2833
+ "loss": 0.0114,
2834
+ "step": 235500
2835
+ },
2836
+ {
2837
+ "epoch": 17.96,
2838
+ "learning_rate": 4.70072789063887e-05,
2839
+ "loss": 0.0122,
2840
+ "step": 236000
2841
+ },
2842
+ {
2843
+ "epoch": 17.99,
2844
+ "learning_rate": 4.7000938395597146e-05,
2845
+ "loss": 0.0126,
2846
+ "step": 236500
2847
+ },
2848
+ {
2849
+ "epoch": 18.03,
2850
+ "learning_rate": 4.69945978848056e-05,
2851
+ "loss": 0.0096,
2852
+ "step": 237000
2853
+ },
2854
+ {
2855
+ "epoch": 18.07,
2856
+ "learning_rate": 4.698825737401405e-05,
2857
+ "loss": 0.0101,
2858
+ "step": 237500
2859
+ },
2860
+ {
2861
+ "epoch": 18.11,
2862
+ "learning_rate": 4.69819168632225e-05,
2863
+ "loss": 0.0095,
2864
+ "step": 238000
2865
+ },
2866
+ {
2867
+ "epoch": 18.15,
2868
+ "learning_rate": 4.6975576352430954e-05,
2869
+ "loss": 0.01,
2870
+ "step": 238500
2871
+ },
2872
+ {
2873
+ "epoch": 18.18,
2874
+ "learning_rate": 4.69692358416394e-05,
2875
+ "loss": 0.0096,
2876
+ "step": 239000
2877
+ },
2878
+ {
2879
+ "epoch": 18.22,
2880
+ "learning_rate": 4.6962895330847854e-05,
2881
+ "loss": 0.0088,
2882
+ "step": 239500
2883
+ },
2884
+ {
2885
+ "epoch": 18.26,
2886
+ "learning_rate": 4.69565548200563e-05,
2887
+ "loss": 0.0112,
2888
+ "step": 240000
2889
+ },
2890
+ {
2891
+ "epoch": 18.3,
2892
+ "learning_rate": 4.6950214309264755e-05,
2893
+ "loss": 0.0097,
2894
+ "step": 240500
2895
+ },
2896
+ {
2897
+ "epoch": 18.34,
2898
+ "learning_rate": 4.694387379847321e-05,
2899
+ "loss": 0.0092,
2900
+ "step": 241000
2901
+ },
2902
+ {
2903
+ "epoch": 18.37,
2904
+ "learning_rate": 4.6937533287681655e-05,
2905
+ "loss": 0.0111,
2906
+ "step": 241500
2907
+ },
2908
+ {
2909
+ "epoch": 18.41,
2910
+ "learning_rate": 4.693119277689011e-05,
2911
+ "loss": 0.0099,
2912
+ "step": 242000
2913
+ },
2914
+ {
2915
+ "epoch": 18.45,
2916
+ "learning_rate": 4.6924852266098555e-05,
2917
+ "loss": 0.0098,
2918
+ "step": 242500
2919
+ },
2920
+ {
2921
+ "epoch": 18.49,
2922
+ "learning_rate": 4.691851175530701e-05,
2923
+ "loss": 0.0105,
2924
+ "step": 243000
2925
+ },
2926
+ {
2927
+ "epoch": 18.53,
2928
+ "learning_rate": 4.691217124451546e-05,
2929
+ "loss": 0.0109,
2930
+ "step": 243500
2931
+ },
2932
+ {
2933
+ "epoch": 18.57,
2934
+ "learning_rate": 4.690583073372391e-05,
2935
+ "loss": 0.0101,
2936
+ "step": 244000
2937
+ },
2938
+ {
2939
+ "epoch": 18.6,
2940
+ "learning_rate": 4.689949022293236e-05,
2941
+ "loss": 0.0117,
2942
+ "step": 244500
2943
+ },
2944
+ {
2945
+ "epoch": 18.64,
2946
+ "learning_rate": 4.689314971214081e-05,
2947
+ "loss": 0.0104,
2948
+ "step": 245000
2949
+ },
2950
+ {
2951
+ "epoch": 18.68,
2952
+ "learning_rate": 4.6886809201349263e-05,
2953
+ "loss": 0.0118,
2954
+ "step": 245500
2955
+ },
2956
+ {
2957
+ "epoch": 18.72,
2958
+ "learning_rate": 4.688046869055772e-05,
2959
+ "loss": 0.0112,
2960
+ "step": 246000
2961
+ },
2962
+ {
2963
+ "epoch": 18.76,
2964
+ "learning_rate": 4.6874128179766164e-05,
2965
+ "loss": 0.0104,
2966
+ "step": 246500
2967
+ },
2968
+ {
2969
+ "epoch": 18.79,
2970
+ "learning_rate": 4.686778766897462e-05,
2971
+ "loss": 0.0109,
2972
+ "step": 247000
2973
+ },
2974
+ {
2975
+ "epoch": 18.83,
2976
+ "learning_rate": 4.6861447158183064e-05,
2977
+ "loss": 0.0107,
2978
+ "step": 247500
2979
+ },
2980
+ {
2981
+ "epoch": 18.87,
2982
+ "learning_rate": 4.685510664739152e-05,
2983
+ "loss": 0.0111,
2984
+ "step": 248000
2985
+ },
2986
+ {
2987
+ "epoch": 18.91,
2988
+ "learning_rate": 4.6848766136599965e-05,
2989
+ "loss": 0.0102,
2990
+ "step": 248500
2991
+ },
2992
+ {
2993
+ "epoch": 18.95,
2994
+ "learning_rate": 4.684242562580842e-05,
2995
+ "loss": 0.0091,
2996
+ "step": 249000
2997
+ },
2998
+ {
2999
+ "epoch": 18.98,
3000
+ "learning_rate": 4.683608511501687e-05,
3001
+ "loss": 0.0124,
3002
+ "step": 249500
3003
+ },
3004
+ {
3005
+ "epoch": 19.02,
3006
+ "learning_rate": 4.682974460422532e-05,
3007
+ "loss": 0.0101,
3008
+ "step": 250000
3009
+ },
3010
+ {
3011
+ "epoch": 19.06,
3012
+ "learning_rate": 4.682340409343377e-05,
3013
+ "loss": 0.0073,
3014
+ "step": 250500
3015
+ },
3016
+ {
3017
+ "epoch": 19.1,
3018
+ "learning_rate": 4.681706358264222e-05,
3019
+ "loss": 0.0078,
3020
+ "step": 251000
3021
+ },
3022
+ {
3023
+ "epoch": 19.14,
3024
+ "learning_rate": 4.681072307185067e-05,
3025
+ "loss": 0.0088,
3026
+ "step": 251500
3027
+ },
3028
+ {
3029
+ "epoch": 19.17,
3030
+ "learning_rate": 4.6804382561059126e-05,
3031
+ "loss": 0.0078,
3032
+ "step": 252000
3033
+ },
3034
+ {
3035
+ "epoch": 19.21,
3036
+ "learning_rate": 4.679804205026757e-05,
3037
+ "loss": 0.0079,
3038
+ "step": 252500
3039
+ },
3040
+ {
3041
+ "epoch": 19.25,
3042
+ "learning_rate": 4.679170153947603e-05,
3043
+ "loss": 0.0085,
3044
+ "step": 253000
3045
+ },
3046
+ {
3047
+ "epoch": 19.29,
3048
+ "learning_rate": 4.6785361028684474e-05,
3049
+ "loss": 0.0086,
3050
+ "step": 253500
3051
+ },
3052
+ {
3053
+ "epoch": 19.33,
3054
+ "learning_rate": 4.677902051789293e-05,
3055
+ "loss": 0.0077,
3056
+ "step": 254000
3057
+ },
3058
+ {
3059
+ "epoch": 19.36,
3060
+ "learning_rate": 4.6772680007101374e-05,
3061
+ "loss": 0.0085,
3062
+ "step": 254500
3063
+ },
3064
+ {
3065
+ "epoch": 19.4,
3066
+ "learning_rate": 4.676633949630983e-05,
3067
+ "loss": 0.008,
3068
+ "step": 255000
3069
+ },
3070
+ {
3071
+ "epoch": 19.44,
3072
+ "learning_rate": 4.6759998985518274e-05,
3073
+ "loss": 0.0104,
3074
+ "step": 255500
3075
+ },
3076
+ {
3077
+ "epoch": 19.48,
3078
+ "learning_rate": 4.675365847472673e-05,
3079
+ "loss": 0.009,
3080
+ "step": 256000
3081
+ },
3082
+ {
3083
+ "epoch": 19.52,
3084
+ "learning_rate": 4.6747317963935175e-05,
3085
+ "loss": 0.0096,
3086
+ "step": 256500
3087
+ },
3088
+ {
3089
+ "epoch": 19.55,
3090
+ "learning_rate": 4.674097745314362e-05,
3091
+ "loss": 0.0086,
3092
+ "step": 257000
3093
+ },
3094
+ {
3095
+ "epoch": 19.59,
3096
+ "learning_rate": 4.6734636942352075e-05,
3097
+ "loss": 0.0089,
3098
+ "step": 257500
3099
+ },
3100
+ {
3101
+ "epoch": 19.63,
3102
+ "learning_rate": 4.672829643156053e-05,
3103
+ "loss": 0.0091,
3104
+ "step": 258000
3105
+ },
3106
+ {
3107
+ "epoch": 19.67,
3108
+ "learning_rate": 4.6721955920768976e-05,
3109
+ "loss": 0.009,
3110
+ "step": 258500
3111
+ },
3112
+ {
3113
+ "epoch": 19.71,
3114
+ "learning_rate": 4.671561540997743e-05,
3115
+ "loss": 0.009,
3116
+ "step": 259000
3117
+ },
3118
+ {
3119
+ "epoch": 19.74,
3120
+ "learning_rate": 4.6709274899185876e-05,
3121
+ "loss": 0.009,
3122
+ "step": 259500
3123
+ },
3124
+ {
3125
+ "epoch": 19.78,
3126
+ "learning_rate": 4.670293438839433e-05,
3127
+ "loss": 0.0102,
3128
+ "step": 260000
3129
+ },
3130
+ {
3131
+ "epoch": 19.82,
3132
+ "learning_rate": 4.669659387760278e-05,
3133
+ "loss": 0.0097,
3134
+ "step": 260500
3135
+ },
3136
+ {
3137
+ "epoch": 19.86,
3138
+ "learning_rate": 4.669025336681123e-05,
3139
+ "loss": 0.0093,
3140
+ "step": 261000
3141
+ },
3142
+ {
3143
+ "epoch": 19.9,
3144
+ "learning_rate": 4.6683912856019684e-05,
3145
+ "loss": 0.0098,
3146
+ "step": 261500
3147
+ },
3148
+ {
3149
+ "epoch": 19.93,
3150
+ "learning_rate": 4.667757234522813e-05,
3151
+ "loss": 0.0093,
3152
+ "step": 262000
3153
+ },
3154
+ {
3155
+ "epoch": 19.97,
3156
+ "learning_rate": 4.6671231834436584e-05,
3157
+ "loss": 0.0101,
3158
+ "step": 262500
3159
+ },
3160
+ {
3161
+ "epoch": 20.01,
3162
+ "learning_rate": 4.666489132364503e-05,
3163
+ "loss": 0.0088,
3164
+ "step": 263000
3165
+ },
3166
+ {
3167
+ "epoch": 20.05,
3168
+ "learning_rate": 4.6658550812853485e-05,
3169
+ "loss": 0.0065,
3170
+ "step": 263500
3171
+ },
3172
+ {
3173
+ "epoch": 20.09,
3174
+ "learning_rate": 4.665221030206194e-05,
3175
+ "loss": 0.0072,
3176
+ "step": 264000
3177
+ },
3178
+ {
3179
+ "epoch": 20.12,
3180
+ "learning_rate": 4.6645869791270385e-05,
3181
+ "loss": 0.0074,
3182
+ "step": 264500
3183
+ },
3184
+ {
3185
+ "epoch": 20.16,
3186
+ "learning_rate": 4.663952928047884e-05,
3187
+ "loss": 0.0079,
3188
+ "step": 265000
3189
+ },
3190
+ {
3191
+ "epoch": 20.2,
3192
+ "learning_rate": 4.6633188769687285e-05,
3193
+ "loss": 0.0079,
3194
+ "step": 265500
3195
+ },
3196
+ {
3197
+ "epoch": 20.24,
3198
+ "learning_rate": 4.662684825889574e-05,
3199
+ "loss": 0.008,
3200
+ "step": 266000
3201
+ },
3202
+ {
3203
+ "epoch": 20.28,
3204
+ "learning_rate": 4.662050774810419e-05,
3205
+ "loss": 0.0073,
3206
+ "step": 266500
3207
+ },
3208
+ {
3209
+ "epoch": 20.31,
3210
+ "learning_rate": 4.661416723731264e-05,
3211
+ "loss": 0.0069,
3212
+ "step": 267000
3213
+ },
3214
+ {
3215
+ "epoch": 20.35,
3216
+ "learning_rate": 4.660782672652109e-05,
3217
+ "loss": 0.0072,
3218
+ "step": 267500
3219
+ },
3220
+ {
3221
+ "epoch": 20.39,
3222
+ "learning_rate": 4.660148621572954e-05,
3223
+ "loss": 0.0073,
3224
+ "step": 268000
3225
+ },
3226
+ {
3227
+ "epoch": 20.43,
3228
+ "learning_rate": 4.6595145704937994e-05,
3229
+ "loss": 0.008,
3230
+ "step": 268500
3231
+ },
3232
+ {
3233
+ "epoch": 20.47,
3234
+ "learning_rate": 4.658880519414645e-05,
3235
+ "loss": 0.0092,
3236
+ "step": 269000
3237
+ },
3238
+ {
3239
+ "epoch": 20.51,
3240
+ "learning_rate": 4.6582464683354894e-05,
3241
+ "loss": 0.0072,
3242
+ "step": 269500
3243
+ },
3244
+ {
3245
+ "epoch": 20.54,
3246
+ "learning_rate": 4.657612417256335e-05,
3247
+ "loss": 0.0068,
3248
+ "step": 270000
3249
+ },
3250
+ {
3251
+ "epoch": 20.58,
3252
+ "learning_rate": 4.6569783661771794e-05,
3253
+ "loss": 0.0085,
3254
+ "step": 270500
3255
+ },
3256
+ {
3257
+ "epoch": 20.62,
3258
+ "learning_rate": 4.656344315098025e-05,
3259
+ "loss": 0.0072,
3260
+ "step": 271000
3261
+ },
3262
+ {
3263
+ "epoch": 20.66,
3264
+ "learning_rate": 4.6557102640188695e-05,
3265
+ "loss": 0.0081,
3266
+ "step": 271500
3267
+ },
3268
+ {
3269
+ "epoch": 20.7,
3270
+ "learning_rate": 4.655076212939715e-05,
3271
+ "loss": 0.0079,
3272
+ "step": 272000
3273
+ },
3274
+ {
3275
+ "epoch": 20.73,
3276
+ "learning_rate": 4.65444216186056e-05,
3277
+ "loss": 0.0092,
3278
+ "step": 272500
3279
+ },
3280
+ {
3281
+ "epoch": 20.77,
3282
+ "learning_rate": 4.653808110781405e-05,
3283
+ "loss": 0.0092,
3284
+ "step": 273000
3285
+ },
3286
+ {
3287
+ "epoch": 20.81,
3288
+ "learning_rate": 4.65317405970225e-05,
3289
+ "loss": 0.0088,
3290
+ "step": 273500
3291
+ },
3292
+ {
3293
+ "epoch": 20.85,
3294
+ "learning_rate": 4.652540008623095e-05,
3295
+ "loss": 0.0078,
3296
+ "step": 274000
3297
+ },
3298
+ {
3299
+ "epoch": 20.89,
3300
+ "learning_rate": 4.65190595754394e-05,
3301
+ "loss": 0.0093,
3302
+ "step": 274500
3303
+ },
3304
+ {
3305
+ "epoch": 20.92,
3306
+ "learning_rate": 4.651271906464785e-05,
3307
+ "loss": 0.0074,
3308
+ "step": 275000
3309
+ },
3310
+ {
3311
+ "epoch": 20.96,
3312
+ "learning_rate": 4.65063785538563e-05,
3313
+ "loss": 0.008,
3314
+ "step": 275500
3315
+ },
3316
+ {
3317
+ "epoch": 21.0,
3318
+ "learning_rate": 4.650003804306475e-05,
3319
+ "loss": 0.0073,
3320
+ "step": 276000
3321
+ },
3322
+ {
3323
+ "epoch": 21.04,
3324
+ "learning_rate": 4.6493697532273204e-05,
3325
+ "loss": 0.0055,
3326
+ "step": 276500
3327
+ },
3328
+ {
3329
+ "epoch": 21.08,
3330
+ "learning_rate": 4.648735702148165e-05,
3331
+ "loss": 0.0065,
3332
+ "step": 277000
3333
+ },
3334
+ {
3335
+ "epoch": 21.11,
3336
+ "learning_rate": 4.6481016510690104e-05,
3337
+ "loss": 0.0053,
3338
+ "step": 277500
3339
+ },
3340
+ {
3341
+ "epoch": 21.15,
3342
+ "learning_rate": 4.647467599989855e-05,
3343
+ "loss": 0.0059,
3344
+ "step": 278000
3345
+ },
3346
+ {
3347
+ "epoch": 21.19,
3348
+ "learning_rate": 4.6468335489107005e-05,
3349
+ "loss": 0.0067,
3350
+ "step": 278500
3351
+ },
3352
+ {
3353
+ "epoch": 21.23,
3354
+ "learning_rate": 4.646199497831545e-05,
3355
+ "loss": 0.0056,
3356
+ "step": 279000
3357
+ },
3358
+ {
3359
+ "epoch": 21.27,
3360
+ "learning_rate": 4.6455654467523905e-05,
3361
+ "loss": 0.0059,
3362
+ "step": 279500
3363
+ },
3364
+ {
3365
+ "epoch": 21.3,
3366
+ "learning_rate": 4.644931395673235e-05,
3367
+ "loss": 0.0072,
3368
+ "step": 280000
3369
+ },
3370
+ {
3371
+ "epoch": 21.34,
3372
+ "learning_rate": 4.6442973445940805e-05,
3373
+ "loss": 0.0071,
3374
+ "step": 280500
3375
+ },
3376
+ {
3377
+ "epoch": 21.38,
3378
+ "learning_rate": 4.643663293514926e-05,
3379
+ "loss": 0.0063,
3380
+ "step": 281000
3381
+ },
3382
+ {
3383
+ "epoch": 21.42,
3384
+ "learning_rate": 4.6430292424357706e-05,
3385
+ "loss": 0.0067,
3386
+ "step": 281500
3387
+ },
3388
+ {
3389
+ "epoch": 21.46,
3390
+ "learning_rate": 4.642395191356616e-05,
3391
+ "loss": 0.0059,
3392
+ "step": 282000
3393
+ },
3394
+ {
3395
+ "epoch": 21.49,
3396
+ "learning_rate": 4.6417611402774606e-05,
3397
+ "loss": 0.0066,
3398
+ "step": 282500
3399
+ },
3400
+ {
3401
+ "epoch": 21.53,
3402
+ "learning_rate": 4.641127089198306e-05,
3403
+ "loss": 0.0066,
3404
+ "step": 283000
3405
+ },
3406
+ {
3407
+ "epoch": 21.57,
3408
+ "learning_rate": 4.6404930381191513e-05,
3409
+ "loss": 0.0067,
3410
+ "step": 283500
3411
+ },
3412
+ {
3413
+ "epoch": 21.61,
3414
+ "learning_rate": 4.639858987039996e-05,
3415
+ "loss": 0.0065,
3416
+ "step": 284000
3417
+ },
3418
+ {
3419
+ "epoch": 21.65,
3420
+ "learning_rate": 4.6392249359608414e-05,
3421
+ "loss": 0.0073,
3422
+ "step": 284500
3423
+ },
3424
+ {
3425
+ "epoch": 21.68,
3426
+ "learning_rate": 4.638590884881686e-05,
3427
+ "loss": 0.0069,
3428
+ "step": 285000
3429
+ },
3430
+ {
3431
+ "epoch": 21.72,
3432
+ "learning_rate": 4.6379568338025314e-05,
3433
+ "loss": 0.0071,
3434
+ "step": 285500
3435
+ },
3436
+ {
3437
+ "epoch": 21.76,
3438
+ "learning_rate": 4.637322782723376e-05,
3439
+ "loss": 0.0072,
3440
+ "step": 286000
3441
+ },
3442
+ {
3443
+ "epoch": 21.8,
3444
+ "learning_rate": 4.6366887316442215e-05,
3445
+ "loss": 0.0079,
3446
+ "step": 286500
3447
+ },
3448
+ {
3449
+ "epoch": 21.84,
3450
+ "learning_rate": 4.636054680565067e-05,
3451
+ "loss": 0.0075,
3452
+ "step": 287000
3453
+ },
3454
+ {
3455
+ "epoch": 21.87,
3456
+ "learning_rate": 4.6354206294859115e-05,
3457
+ "loss": 0.006,
3458
+ "step": 287500
3459
+ },
3460
+ {
3461
+ "epoch": 21.91,
3462
+ "learning_rate": 4.634786578406757e-05,
3463
+ "loss": 0.007,
3464
+ "step": 288000
3465
+ },
3466
+ {
3467
+ "epoch": 21.95,
3468
+ "learning_rate": 4.6341525273276016e-05,
3469
+ "loss": 0.0078,
3470
+ "step": 288500
3471
+ },
3472
+ {
3473
+ "epoch": 21.99,
3474
+ "learning_rate": 4.633518476248447e-05,
3475
+ "loss": 0.0068,
3476
+ "step": 289000
3477
+ },
3478
+ {
3479
+ "epoch": 22.03,
3480
+ "learning_rate": 4.632884425169292e-05,
3481
+ "loss": 0.0059,
3482
+ "step": 289500
3483
+ },
3484
+ {
3485
+ "epoch": 22.06,
3486
+ "learning_rate": 4.632250374090137e-05,
3487
+ "loss": 0.0038,
3488
+ "step": 290000
3489
+ },
3490
+ {
3491
+ "epoch": 22.1,
3492
+ "learning_rate": 4.631616323010982e-05,
3493
+ "loss": 0.0049,
3494
+ "step": 290500
3495
+ },
3496
+ {
3497
+ "epoch": 22.14,
3498
+ "learning_rate": 4.630982271931827e-05,
3499
+ "loss": 0.0052,
3500
+ "step": 291000
3501
+ },
3502
+ {
3503
+ "epoch": 22.18,
3504
+ "learning_rate": 4.6303482208526724e-05,
3505
+ "loss": 0.0059,
3506
+ "step": 291500
3507
+ },
3508
+ {
3509
+ "epoch": 22.22,
3510
+ "learning_rate": 4.629714169773517e-05,
3511
+ "loss": 0.0049,
3512
+ "step": 292000
3513
+ },
3514
+ {
3515
+ "epoch": 22.26,
3516
+ "learning_rate": 4.6290801186943624e-05,
3517
+ "loss": 0.0055,
3518
+ "step": 292500
3519
+ },
3520
+ {
3521
+ "epoch": 22.29,
3522
+ "learning_rate": 4.628446067615208e-05,
3523
+ "loss": 0.0056,
3524
+ "step": 293000
3525
+ },
3526
+ {
3527
+ "epoch": 22.33,
3528
+ "learning_rate": 4.6278120165360524e-05,
3529
+ "loss": 0.0057,
3530
+ "step": 293500
3531
+ },
3532
+ {
3533
+ "epoch": 22.37,
3534
+ "learning_rate": 4.627177965456898e-05,
3535
+ "loss": 0.0054,
3536
+ "step": 294000
3537
+ },
3538
+ {
3539
+ "epoch": 22.41,
3540
+ "learning_rate": 4.6265439143777425e-05,
3541
+ "loss": 0.0061,
3542
+ "step": 294500
3543
+ },
3544
+ {
3545
+ "epoch": 22.45,
3546
+ "learning_rate": 4.625909863298588e-05,
3547
+ "loss": 0.0057,
3548
+ "step": 295000
3549
+ },
3550
+ {
3551
+ "epoch": 22.48,
3552
+ "learning_rate": 4.6252758122194325e-05,
3553
+ "loss": 0.0054,
3554
+ "step": 295500
3555
+ },
3556
+ {
3557
+ "epoch": 22.52,
3558
+ "learning_rate": 4.624641761140278e-05,
3559
+ "loss": 0.0068,
3560
+ "step": 296000
3561
+ },
3562
+ {
3563
+ "epoch": 22.56,
3564
+ "learning_rate": 4.6240077100611226e-05,
3565
+ "loss": 0.0054,
3566
+ "step": 296500
3567
+ },
3568
+ {
3569
+ "epoch": 22.6,
3570
+ "learning_rate": 4.623373658981968e-05,
3571
+ "loss": 0.0056,
3572
+ "step": 297000
3573
+ },
3574
+ {
3575
+ "epoch": 22.64,
3576
+ "learning_rate": 4.6227396079028126e-05,
3577
+ "loss": 0.0059,
3578
+ "step": 297500
3579
+ },
3580
+ {
3581
+ "epoch": 22.67,
3582
+ "learning_rate": 4.622105556823658e-05,
3583
+ "loss": 0.0052,
3584
+ "step": 298000
3585
+ },
3586
+ {
3587
+ "epoch": 22.71,
3588
+ "learning_rate": 4.621471505744503e-05,
3589
+ "loss": 0.0066,
3590
+ "step": 298500
3591
+ },
3592
+ {
3593
+ "epoch": 22.75,
3594
+ "learning_rate": 4.620837454665348e-05,
3595
+ "loss": 0.0065,
3596
+ "step": 299000
3597
+ },
3598
+ {
3599
+ "epoch": 22.79,
3600
+ "learning_rate": 4.620203403586193e-05,
3601
+ "loss": 0.0054,
3602
+ "step": 299500
3603
+ },
3604
+ {
3605
+ "epoch": 22.83,
3606
+ "learning_rate": 4.619569352507038e-05,
3607
+ "loss": 0.0056,
3608
+ "step": 300000
3609
+ },
3610
+ {
3611
+ "epoch": 22.86,
3612
+ "learning_rate": 4.618935301427883e-05,
3613
+ "loss": 0.0057,
3614
+ "step": 300500
3615
+ },
3616
+ {
3617
+ "epoch": 22.9,
3618
+ "learning_rate": 4.618301250348728e-05,
3619
+ "loss": 0.0051,
3620
+ "step": 301000
3621
+ },
3622
+ {
3623
+ "epoch": 22.94,
3624
+ "learning_rate": 4.6176671992695735e-05,
3625
+ "loss": 0.0065,
3626
+ "step": 301500
3627
+ },
3628
+ {
3629
+ "epoch": 22.98,
3630
+ "learning_rate": 4.617033148190418e-05,
3631
+ "loss": 0.0058,
3632
+ "step": 302000
3633
+ },
3634
+ {
3635
+ "epoch": 23.02,
3636
+ "learning_rate": 4.6163990971112635e-05,
3637
+ "loss": 0.0061,
3638
+ "step": 302500
3639
+ },
3640
+ {
3641
+ "epoch": 23.05,
3642
+ "learning_rate": 4.615765046032108e-05,
3643
+ "loss": 0.0038,
3644
+ "step": 303000
3645
+ },
3646
+ {
3647
+ "epoch": 23.09,
3648
+ "learning_rate": 4.6151309949529536e-05,
3649
+ "loss": 0.0045,
3650
+ "step": 303500
3651
+ },
3652
+ {
3653
+ "epoch": 23.13,
3654
+ "learning_rate": 4.614496943873799e-05,
3655
+ "loss": 0.004,
3656
+ "step": 304000
3657
+ },
3658
+ {
3659
+ "epoch": 23.17,
3660
+ "learning_rate": 4.6138628927946436e-05,
3661
+ "loss": 0.0042,
3662
+ "step": 304500
3663
+ },
3664
+ {
3665
+ "epoch": 23.21,
3666
+ "learning_rate": 4.613228841715489e-05,
3667
+ "loss": 0.0046,
3668
+ "step": 305000
3669
+ },
3670
+ {
3671
+ "epoch": 23.24,
3672
+ "learning_rate": 4.6125947906363336e-05,
3673
+ "loss": 0.0046,
3674
+ "step": 305500
3675
+ },
3676
+ {
3677
+ "epoch": 23.28,
3678
+ "learning_rate": 4.611960739557179e-05,
3679
+ "loss": 0.0044,
3680
+ "step": 306000
3681
+ },
3682
+ {
3683
+ "epoch": 23.32,
3684
+ "learning_rate": 4.611326688478024e-05,
3685
+ "loss": 0.0045,
3686
+ "step": 306500
3687
+ },
3688
+ {
3689
+ "epoch": 23.36,
3690
+ "learning_rate": 4.610692637398869e-05,
3691
+ "loss": 0.0053,
3692
+ "step": 307000
3693
+ },
3694
+ {
3695
+ "epoch": 23.4,
3696
+ "learning_rate": 4.6100585863197144e-05,
3697
+ "loss": 0.0046,
3698
+ "step": 307500
3699
+ },
3700
+ {
3701
+ "epoch": 23.43,
3702
+ "learning_rate": 4.609424535240559e-05,
3703
+ "loss": 0.0049,
3704
+ "step": 308000
3705
+ },
3706
+ {
3707
+ "epoch": 23.47,
3708
+ "learning_rate": 4.6087904841614044e-05,
3709
+ "loss": 0.0066,
3710
+ "step": 308500
3711
+ },
3712
+ {
3713
+ "epoch": 23.51,
3714
+ "learning_rate": 4.608156433082249e-05,
3715
+ "loss": 0.0049,
3716
+ "step": 309000
3717
+ },
3718
+ {
3719
+ "epoch": 23.55,
3720
+ "learning_rate": 4.6075223820030945e-05,
3721
+ "loss": 0.0055,
3722
+ "step": 309500
3723
+ },
3724
+ {
3725
+ "epoch": 23.59,
3726
+ "learning_rate": 4.60688833092394e-05,
3727
+ "loss": 0.005,
3728
+ "step": 310000
3729
+ },
3730
+ {
3731
+ "epoch": 23.62,
3732
+ "learning_rate": 4.6062542798447845e-05,
3733
+ "loss": 0.0053,
3734
+ "step": 310500
3735
+ },
3736
+ {
3737
+ "epoch": 23.66,
3738
+ "learning_rate": 4.60562022876563e-05,
3739
+ "loss": 0.0048,
3740
+ "step": 311000
3741
+ },
3742
+ {
3743
+ "epoch": 23.7,
3744
+ "learning_rate": 4.6049861776864746e-05,
3745
+ "loss": 0.0046,
3746
+ "step": 311500
3747
+ },
3748
+ {
3749
+ "epoch": 23.74,
3750
+ "learning_rate": 4.60435212660732e-05,
3751
+ "loss": 0.0055,
3752
+ "step": 312000
3753
+ },
3754
+ {
3755
+ "epoch": 23.78,
3756
+ "learning_rate": 4.603718075528165e-05,
3757
+ "loss": 0.0057,
3758
+ "step": 312500
3759
+ },
3760
+ {
3761
+ "epoch": 23.81,
3762
+ "learning_rate": 4.60308402444901e-05,
3763
+ "loss": 0.0055,
3764
+ "step": 313000
3765
+ },
3766
+ {
3767
+ "epoch": 23.85,
3768
+ "learning_rate": 4.602449973369855e-05,
3769
+ "loss": 0.0049,
3770
+ "step": 313500
3771
+ },
3772
+ {
3773
+ "epoch": 23.89,
3774
+ "learning_rate": 4.6018159222907e-05,
3775
+ "loss": 0.0062,
3776
+ "step": 314000
3777
+ },
3778
+ {
3779
+ "epoch": 23.93,
3780
+ "learning_rate": 4.6011818712115454e-05,
3781
+ "loss": 0.0053,
3782
+ "step": 314500
3783
+ },
3784
+ {
3785
+ "epoch": 23.97,
3786
+ "learning_rate": 4.60054782013239e-05,
3787
+ "loss": 0.0047,
3788
+ "step": 315000
3789
+ },
3790
+ {
3791
+ "epoch": 24.01,
3792
+ "learning_rate": 4.5999137690532354e-05,
3793
+ "loss": 0.0051,
3794
+ "step": 315500
3795
+ },
3796
+ {
3797
+ "epoch": 24.04,
3798
+ "learning_rate": 4.59927971797408e-05,
3799
+ "loss": 0.0034,
3800
+ "step": 316000
3801
+ },
3802
+ {
3803
+ "epoch": 24.08,
3804
+ "learning_rate": 4.5986456668949255e-05,
3805
+ "loss": 0.0037,
3806
+ "step": 316500
3807
+ },
3808
+ {
3809
+ "epoch": 24.12,
3810
+ "learning_rate": 4.59801161581577e-05,
3811
+ "loss": 0.0031,
3812
+ "step": 317000
3813
+ },
3814
+ {
3815
+ "epoch": 24.16,
3816
+ "learning_rate": 4.5973775647366155e-05,
3817
+ "loss": 0.0035,
3818
+ "step": 317500
3819
+ },
3820
+ {
3821
+ "epoch": 24.2,
3822
+ "learning_rate": 4.59674351365746e-05,
3823
+ "loss": 0.0047,
3824
+ "step": 318000
3825
+ },
3826
+ {
3827
+ "epoch": 24.23,
3828
+ "learning_rate": 4.5961094625783055e-05,
3829
+ "loss": 0.0041,
3830
+ "step": 318500
3831
+ },
3832
+ {
3833
+ "epoch": 24.27,
3834
+ "learning_rate": 4.59547541149915e-05,
3835
+ "loss": 0.0034,
3836
+ "step": 319000
3837
+ },
3838
+ {
3839
+ "epoch": 24.31,
3840
+ "learning_rate": 4.5948413604199956e-05,
3841
+ "loss": 0.0043,
3842
+ "step": 319500
3843
+ },
3844
+ {
3845
+ "epoch": 24.35,
3846
+ "learning_rate": 4.59420730934084e-05,
3847
+ "loss": 0.0042,
3848
+ "step": 320000
3849
+ },
3850
+ {
3851
+ "epoch": 24.39,
3852
+ "learning_rate": 4.5935732582616856e-05,
3853
+ "loss": 0.0043,
3854
+ "step": 320500
3855
+ },
3856
+ {
3857
+ "epoch": 24.42,
3858
+ "learning_rate": 4.59293920718253e-05,
3859
+ "loss": 0.004,
3860
+ "step": 321000
3861
+ },
3862
+ {
3863
+ "epoch": 24.46,
3864
+ "learning_rate": 4.592305156103376e-05,
3865
+ "loss": 0.0043,
3866
+ "step": 321500
3867
+ },
3868
+ {
3869
+ "epoch": 24.5,
3870
+ "learning_rate": 4.591671105024221e-05,
3871
+ "loss": 0.004,
3872
+ "step": 322000
3873
+ },
3874
+ {
3875
+ "epoch": 24.54,
3876
+ "learning_rate": 4.591037053945066e-05,
3877
+ "loss": 0.0054,
3878
+ "step": 322500
3879
+ },
3880
+ {
3881
+ "epoch": 24.58,
3882
+ "learning_rate": 4.590403002865911e-05,
3883
+ "loss": 0.0039,
3884
+ "step": 323000
3885
+ },
3886
+ {
3887
+ "epoch": 24.61,
3888
+ "learning_rate": 4.589768951786756e-05,
3889
+ "loss": 0.0051,
3890
+ "step": 323500
3891
+ },
3892
+ {
3893
+ "epoch": 24.65,
3894
+ "learning_rate": 4.589134900707601e-05,
3895
+ "loss": 0.0044,
3896
+ "step": 324000
3897
+ },
3898
+ {
3899
+ "epoch": 24.69,
3900
+ "learning_rate": 4.5885008496284465e-05,
3901
+ "loss": 0.0042,
3902
+ "step": 324500
3903
+ },
3904
+ {
3905
+ "epoch": 24.73,
3906
+ "learning_rate": 4.587866798549291e-05,
3907
+ "loss": 0.0049,
3908
+ "step": 325000
3909
+ },
3910
+ {
3911
+ "epoch": 24.77,
3912
+ "learning_rate": 4.5872327474701365e-05,
3913
+ "loss": 0.0048,
3914
+ "step": 325500
3915
+ },
3916
+ {
3917
+ "epoch": 24.8,
3918
+ "learning_rate": 4.586598696390981e-05,
3919
+ "loss": 0.0047,
3920
+ "step": 326000
3921
+ },
3922
+ {
3923
+ "epoch": 24.84,
3924
+ "learning_rate": 4.5859646453118266e-05,
3925
+ "loss": 0.0049,
3926
+ "step": 326500
3927
+ },
3928
+ {
3929
+ "epoch": 24.88,
3930
+ "learning_rate": 4.585330594232672e-05,
3931
+ "loss": 0.0048,
3932
+ "step": 327000
3933
+ },
3934
+ {
3935
+ "epoch": 24.92,
3936
+ "learning_rate": 4.5846965431535166e-05,
3937
+ "loss": 0.0044,
3938
+ "step": 327500
3939
+ },
3940
+ {
3941
+ "epoch": 24.96,
3942
+ "learning_rate": 4.584062492074362e-05,
3943
+ "loss": 0.0044,
3944
+ "step": 328000
3945
+ },
3946
+ {
3947
+ "epoch": 24.99,
3948
+ "learning_rate": 4.5834284409952066e-05,
3949
+ "loss": 0.005,
3950
+ "step": 328500
3951
+ },
3952
+ {
3953
+ "epoch": 25.03,
3954
+ "learning_rate": 4.582794389916052e-05,
3955
+ "loss": 0.0033,
3956
+ "step": 329000
3957
+ },
3958
+ {
3959
+ "epoch": 25.07,
3960
+ "learning_rate": 4.582160338836897e-05,
3961
+ "loss": 0.0032,
3962
+ "step": 329500
3963
+ },
3964
+ {
3965
+ "epoch": 25.11,
3966
+ "learning_rate": 4.581526287757742e-05,
3967
+ "loss": 0.0032,
3968
+ "step": 330000
3969
+ },
3970
+ {
3971
+ "epoch": 25.15,
3972
+ "learning_rate": 4.5808922366785874e-05,
3973
+ "loss": 0.0038,
3974
+ "step": 330500
3975
+ },
3976
+ {
3977
+ "epoch": 25.18,
3978
+ "learning_rate": 4.580258185599432e-05,
3979
+ "loss": 0.0031,
3980
+ "step": 331000
3981
+ },
3982
+ {
3983
+ "epoch": 25.22,
3984
+ "learning_rate": 4.5796241345202775e-05,
3985
+ "loss": 0.0039,
3986
+ "step": 331500
3987
+ },
3988
+ {
3989
+ "epoch": 25.26,
3990
+ "learning_rate": 4.578990083441122e-05,
3991
+ "loss": 0.0036,
3992
+ "step": 332000
3993
+ },
3994
+ {
3995
+ "epoch": 25.3,
3996
+ "learning_rate": 4.5783560323619675e-05,
3997
+ "loss": 0.0031,
3998
+ "step": 332500
3999
+ },
4000
+ {
4001
+ "epoch": 25.34,
4002
+ "learning_rate": 4.577721981282813e-05,
4003
+ "loss": 0.0035,
4004
+ "step": 333000
4005
+ },
4006
+ {
4007
+ "epoch": 25.37,
4008
+ "learning_rate": 4.5770879302036575e-05,
4009
+ "loss": 0.0041,
4010
+ "step": 333500
4011
+ },
4012
+ {
4013
+ "epoch": 25.41,
4014
+ "learning_rate": 4.576453879124503e-05,
4015
+ "loss": 0.0044,
4016
+ "step": 334000
4017
+ },
4018
+ {
4019
+ "epoch": 25.45,
4020
+ "learning_rate": 4.5758198280453476e-05,
4021
+ "loss": 0.0033,
4022
+ "step": 334500
4023
+ },
4024
+ {
4025
+ "epoch": 25.49,
4026
+ "learning_rate": 4.575185776966193e-05,
4027
+ "loss": 0.0041,
4028
+ "step": 335000
4029
+ },
4030
+ {
4031
+ "epoch": 25.53,
4032
+ "learning_rate": 4.5745517258870376e-05,
4033
+ "loss": 0.004,
4034
+ "step": 335500
4035
+ },
4036
+ {
4037
+ "epoch": 25.56,
4038
+ "learning_rate": 4.573917674807883e-05,
4039
+ "loss": 0.0037,
4040
+ "step": 336000
4041
+ },
4042
+ {
4043
+ "epoch": 25.6,
4044
+ "learning_rate": 4.573283623728728e-05,
4045
+ "loss": 0.0046,
4046
+ "step": 336500
4047
+ },
4048
+ {
4049
+ "epoch": 25.64,
4050
+ "learning_rate": 4.572649572649573e-05,
4051
+ "loss": 0.0045,
4052
+ "step": 337000
4053
+ },
4054
+ {
4055
+ "epoch": 25.68,
4056
+ "learning_rate": 4.572015521570418e-05,
4057
+ "loss": 0.0043,
4058
+ "step": 337500
4059
+ },
4060
+ {
4061
+ "epoch": 25.72,
4062
+ "learning_rate": 4.571381470491263e-05,
4063
+ "loss": 0.0037,
4064
+ "step": 338000
4065
+ },
4066
+ {
4067
+ "epoch": 25.76,
4068
+ "learning_rate": 4.570747419412108e-05,
4069
+ "loss": 0.004,
4070
+ "step": 338500
4071
+ },
4072
+ {
4073
+ "epoch": 25.79,
4074
+ "learning_rate": 4.570113368332953e-05,
4075
+ "loss": 0.0038,
4076
+ "step": 339000
4077
+ },
4078
+ {
4079
+ "epoch": 25.83,
4080
+ "learning_rate": 4.569479317253798e-05,
4081
+ "loss": 0.0039,
4082
+ "step": 339500
4083
+ },
4084
+ {
4085
+ "epoch": 25.87,
4086
+ "learning_rate": 4.568845266174643e-05,
4087
+ "loss": 0.0046,
4088
+ "step": 340000
4089
+ },
4090
+ {
4091
+ "epoch": 25.91,
4092
+ "learning_rate": 4.568211215095488e-05,
4093
+ "loss": 0.0042,
4094
+ "step": 340500
4095
+ },
4096
+ {
4097
+ "epoch": 25.95,
4098
+ "learning_rate": 4.567577164016333e-05,
4099
+ "loss": 0.004,
4100
+ "step": 341000
4101
+ },
4102
+ {
4103
+ "epoch": 25.98,
4104
+ "learning_rate": 4.5669431129371786e-05,
4105
+ "loss": 0.0037,
4106
+ "step": 341500
4107
+ },
4108
+ {
4109
+ "epoch": 26.02,
4110
+ "learning_rate": 4.566309061858023e-05,
4111
+ "loss": 0.0033,
4112
+ "step": 342000
4113
+ },
4114
+ {
4115
+ "epoch": 26.06,
4116
+ "learning_rate": 4.5656750107788686e-05,
4117
+ "loss": 0.0031,
4118
+ "step": 342500
4119
+ },
4120
+ {
4121
+ "epoch": 26.1,
4122
+ "learning_rate": 4.565040959699713e-05,
4123
+ "loss": 0.0031,
4124
+ "step": 343000
4125
+ },
4126
+ {
4127
+ "epoch": 26.14,
4128
+ "learning_rate": 4.5644069086205586e-05,
4129
+ "loss": 0.0036,
4130
+ "step": 343500
4131
+ },
4132
+ {
4133
+ "epoch": 26.17,
4134
+ "learning_rate": 4.563772857541403e-05,
4135
+ "loss": 0.0028,
4136
+ "step": 344000
4137
+ },
4138
+ {
4139
+ "epoch": 26.21,
4140
+ "learning_rate": 4.563138806462249e-05,
4141
+ "loss": 0.0029,
4142
+ "step": 344500
4143
+ },
4144
+ {
4145
+ "epoch": 26.25,
4146
+ "learning_rate": 4.562504755383094e-05,
4147
+ "loss": 0.0031,
4148
+ "step": 345000
4149
+ },
4150
+ {
4151
+ "epoch": 26.29,
4152
+ "learning_rate": 4.561870704303939e-05,
4153
+ "loss": 0.003,
4154
+ "step": 345500
4155
+ },
4156
+ {
4157
+ "epoch": 26.33,
4158
+ "learning_rate": 4.561236653224784e-05,
4159
+ "loss": 0.0028,
4160
+ "step": 346000
4161
+ },
4162
+ {
4163
+ "epoch": 26.36,
4164
+ "learning_rate": 4.560602602145629e-05,
4165
+ "loss": 0.0041,
4166
+ "step": 346500
4167
+ },
4168
+ {
4169
+ "epoch": 26.4,
4170
+ "learning_rate": 4.559968551066474e-05,
4171
+ "loss": 0.0041,
4172
+ "step": 347000
4173
+ },
4174
+ {
4175
+ "epoch": 26.44,
4176
+ "learning_rate": 4.5593344999873195e-05,
4177
+ "loss": 0.0034,
4178
+ "step": 347500
4179
+ },
4180
+ {
4181
+ "epoch": 26.48,
4182
+ "learning_rate": 4.558700448908164e-05,
4183
+ "loss": 0.0032,
4184
+ "step": 348000
4185
+ },
4186
+ {
4187
+ "epoch": 26.52,
4188
+ "learning_rate": 4.5580663978290095e-05,
4189
+ "loss": 0.0034,
4190
+ "step": 348500
4191
+ },
4192
+ {
4193
+ "epoch": 26.55,
4194
+ "learning_rate": 4.557432346749854e-05,
4195
+ "loss": 0.003,
4196
+ "step": 349000
4197
+ },
4198
+ {
4199
+ "epoch": 26.59,
4200
+ "learning_rate": 4.5567982956706996e-05,
4201
+ "loss": 0.0032,
4202
+ "step": 349500
4203
+ },
4204
+ {
4205
+ "epoch": 26.63,
4206
+ "learning_rate": 4.556164244591545e-05,
4207
+ "loss": 0.0029,
4208
+ "step": 350000
4209
+ },
4210
+ {
4211
+ "epoch": 26.67,
4212
+ "learning_rate": 4.5555301935123896e-05,
4213
+ "loss": 0.0037,
4214
+ "step": 350500
4215
+ },
4216
+ {
4217
+ "epoch": 26.71,
4218
+ "learning_rate": 4.554896142433235e-05,
4219
+ "loss": 0.0037,
4220
+ "step": 351000
4221
+ },
4222
+ {
4223
+ "epoch": 26.74,
4224
+ "learning_rate": 4.5542620913540797e-05,
4225
+ "loss": 0.0032,
4226
+ "step": 351500
4227
+ },
4228
+ {
4229
+ "epoch": 26.78,
4230
+ "learning_rate": 4.553628040274925e-05,
4231
+ "loss": 0.0034,
4232
+ "step": 352000
4233
+ },
4234
+ {
4235
+ "epoch": 26.82,
4236
+ "learning_rate": 4.55299398919577e-05,
4237
+ "loss": 0.0041,
4238
+ "step": 352500
4239
+ },
4240
+ {
4241
+ "epoch": 26.86,
4242
+ "learning_rate": 4.552359938116615e-05,
4243
+ "loss": 0.0037,
4244
+ "step": 353000
4245
+ },
4246
+ {
4247
+ "epoch": 26.9,
4248
+ "learning_rate": 4.5517258870374604e-05,
4249
+ "loss": 0.0042,
4250
+ "step": 353500
4251
+ },
4252
+ {
4253
+ "epoch": 26.93,
4254
+ "learning_rate": 4.551091835958305e-05,
4255
+ "loss": 0.0032,
4256
+ "step": 354000
4257
+ },
4258
+ {
4259
+ "epoch": 26.97,
4260
+ "learning_rate": 4.5504577848791505e-05,
4261
+ "loss": 0.0039,
4262
+ "step": 354500
4263
+ },
4264
+ {
4265
+ "epoch": 27.01,
4266
+ "learning_rate": 4.549823733799995e-05,
4267
+ "loss": 0.0039,
4268
+ "step": 355000
4269
+ },
4270
+ {
4271
+ "epoch": 27.05,
4272
+ "learning_rate": 4.5491896827208405e-05,
4273
+ "loss": 0.0028,
4274
+ "step": 355500
4275
+ },
4276
+ {
4277
+ "epoch": 27.09,
4278
+ "learning_rate": 4.548555631641685e-05,
4279
+ "loss": 0.0032,
4280
+ "step": 356000
4281
+ },
4282
+ {
4283
+ "epoch": 27.12,
4284
+ "learning_rate": 4.5479215805625305e-05,
4285
+ "loss": 0.0033,
4286
+ "step": 356500
4287
+ },
4288
+ {
4289
+ "epoch": 27.16,
4290
+ "learning_rate": 4.547287529483375e-05,
4291
+ "loss": 0.003,
4292
+ "step": 357000
4293
+ },
4294
+ {
4295
+ "epoch": 27.2,
4296
+ "learning_rate": 4.5466534784042206e-05,
4297
+ "loss": 0.0026,
4298
+ "step": 357500
4299
+ },
4300
+ {
4301
+ "epoch": 27.24,
4302
+ "learning_rate": 4.546019427325065e-05,
4303
+ "loss": 0.0028,
4304
+ "step": 358000
4305
+ },
4306
+ {
4307
+ "epoch": 27.28,
4308
+ "learning_rate": 4.5453853762459106e-05,
4309
+ "loss": 0.0034,
4310
+ "step": 358500
4311
+ },
4312
+ {
4313
+ "epoch": 27.31,
4314
+ "learning_rate": 4.544751325166755e-05,
4315
+ "loss": 0.0036,
4316
+ "step": 359000
4317
+ },
4318
+ {
4319
+ "epoch": 27.35,
4320
+ "learning_rate": 4.544117274087601e-05,
4321
+ "loss": 0.0026,
4322
+ "step": 359500
4323
+ },
4324
+ {
4325
+ "epoch": 27.39,
4326
+ "learning_rate": 4.5434832230084454e-05,
4327
+ "loss": 0.0032,
4328
+ "step": 360000
4329
+ },
4330
+ {
4331
+ "epoch": 27.43,
4332
+ "learning_rate": 4.542849171929291e-05,
4333
+ "loss": 0.003,
4334
+ "step": 360500
4335
+ },
4336
+ {
4337
+ "epoch": 27.47,
4338
+ "learning_rate": 4.5422151208501354e-05,
4339
+ "loss": 0.0029,
4340
+ "step": 361000
4341
+ },
4342
+ {
4343
+ "epoch": 27.51,
4344
+ "learning_rate": 4.541581069770981e-05,
4345
+ "loss": 0.0032,
4346
+ "step": 361500
4347
+ },
4348
+ {
4349
+ "epoch": 27.54,
4350
+ "learning_rate": 4.540947018691826e-05,
4351
+ "loss": 0.0037,
4352
+ "step": 362000
4353
+ },
4354
+ {
4355
+ "epoch": 27.58,
4356
+ "learning_rate": 4.540312967612671e-05,
4357
+ "loss": 0.0028,
4358
+ "step": 362500
4359
+ },
4360
+ {
4361
+ "epoch": 27.62,
4362
+ "learning_rate": 4.539678916533516e-05,
4363
+ "loss": 0.0025,
4364
+ "step": 363000
4365
+ },
4366
+ {
4367
+ "epoch": 27.66,
4368
+ "learning_rate": 4.539044865454361e-05,
4369
+ "loss": 0.0027,
4370
+ "step": 363500
4371
+ },
4372
+ {
4373
+ "epoch": 27.7,
4374
+ "learning_rate": 4.538410814375206e-05,
4375
+ "loss": 0.0033,
4376
+ "step": 364000
4377
+ },
4378
+ {
4379
+ "epoch": 27.73,
4380
+ "learning_rate": 4.5377767632960516e-05,
4381
+ "loss": 0.003,
4382
+ "step": 364500
4383
+ },
4384
+ {
4385
+ "epoch": 27.77,
4386
+ "learning_rate": 4.537142712216896e-05,
4387
+ "loss": 0.0035,
4388
+ "step": 365000
4389
+ },
4390
+ {
4391
+ "epoch": 27.81,
4392
+ "learning_rate": 4.5365086611377416e-05,
4393
+ "loss": 0.0035,
4394
+ "step": 365500
4395
+ },
4396
+ {
4397
+ "epoch": 27.85,
4398
+ "learning_rate": 4.535874610058586e-05,
4399
+ "loss": 0.0038,
4400
+ "step": 366000
4401
+ },
4402
+ {
4403
+ "epoch": 27.89,
4404
+ "learning_rate": 4.5352405589794317e-05,
4405
+ "loss": 0.004,
4406
+ "step": 366500
4407
+ },
4408
+ {
4409
+ "epoch": 27.92,
4410
+ "learning_rate": 4.534606507900276e-05,
4411
+ "loss": 0.0029,
4412
+ "step": 367000
4413
+ }
4414
+ ],
4415
+ "max_steps": 3942900,
4416
+ "num_train_epochs": 300,
4417
+ "total_flos": 6.4695900758016e+16,
4418
+ "trial_name": null,
4419
+ "trial_params": null
4420
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d68df231715e5a47b66c5ebe8633c945aedeca845d8b46c0a2d6bf844ac6e436
3
+ size 4411