Hugging Face's logo

Wanfq
/
Explore-LM-Ext-7B-Rewriting

Text Generation
Transformers
PyTorch
English
llama
text-generation-inference
Inference Endpoints
Model card Files Community
Wanfq commited on
Commit
0e4224d
•
1 Parent(s): 91938be

First model version

Browse files
Files changed (11) hide show
  1. added_tokens.json +3 -0
  2. config.json +23 -0
  3. generation_config.json +7 -0
  4. latest +1 -0
  5. pytorch_model.bin +3 -0
  6. special_tokens_map.json +6 -0
  7. tokenizer.model +3 -0
  8. tokenizer_config.json +9 -0
  9. trainer_state.json +4525 -0
  10. training_args.bin +3 -0
  11. zero_to_fp32.py +483 -0
added_tokens.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "[PAD]": 32000
3
+ }
config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "../../../PLMs/ft_local/LLAMA/hf/llama-7b",
3
+ "architectures": [
4
+ "LLaMAForCausalLM"
5
+ ],
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 4096,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 11008,
12
+ "max_sequence_length": 2048,
13
+ "model_type": "llama",
14
+ "num_attention_heads": 32,
15
+ "num_hidden_layers": 32,
16
+ "pad_token_id": 0,
17
+ "rms_norm_eps": 1e-06,
18
+ "tie_word_embeddings": false,
19
+ "torch_dtype": "float16",
20
+ "transformers_version": "4.27.0.dev0",
21
+ "use_cache": true,
22
+ "vocab_size": 32001
23
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.27.0.dev0"
7
+ }
latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step750
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2164bad2ed3e674fb3ba762a7656c963c0415808261788903d8bc4df028175d6
3
+ size 26953810889
special_tokens_map.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "</s>",
3
+ "eos_token": "</s>",
4
+ "pad_token": "[PAD]",
5
+ "unk_token": "</s>"
6
+ }
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
tokenizer_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "",
3
+ "eos_token": "",
4
+ "model_max_length": 512,
5
+ "padding_side": "right",
6
+ "special_tokens_map_file": "../../../PLMs/ft_local/LLAMA/hf/llama-7b/special_tokens_map.json",
7
+ "tokenizer_class": "LLaMATokenizer",
8
+ "unk_token": ""
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,4525 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "global_step": 750,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.0,
12
+ "learning_rate": 0,
13
+ "loss": 1.5272,
14
+ "step": 1
15
+ },
16
+ {
17
+ "epoch": 0.01,
18
+ "learning_rate": 0,
19
+ "loss": 1.4685,
20
+ "step": 2
21
+ },
22
+ {
23
+ "epoch": 0.01,
24
+ "learning_rate": 0,
25
+ "loss": 1.5206,
26
+ "step": 3
27
+ },
28
+ {
29
+ "epoch": 0.02,
30
+ "learning_rate": 0.0,
31
+ "loss": 1.4631,
32
+ "step": 4
33
+ },
34
+ {
35
+ "epoch": 0.02,
36
+ "learning_rate": 4.421294589150075e-06,
37
+ "loss": 0.9541,
38
+ "step": 5
39
+ },
40
+ {
41
+ "epoch": 0.02,
42
+ "learning_rate": 7.007586128444221e-06,
43
+ "loss": 0.974,
44
+ "step": 6
45
+ },
46
+ {
47
+ "epoch": 0.03,
48
+ "learning_rate": 8.84258917830015e-06,
49
+ "loss": 0.7967,
50
+ "step": 7
51
+ },
52
+ {
53
+ "epoch": 0.03,
54
+ "learning_rate": 1.0265928122321039e-05,
55
+ "loss": 0.769,
56
+ "step": 8
57
+ },
58
+ {
59
+ "epoch": 0.04,
60
+ "learning_rate": 1.1428880717594295e-05,
61
+ "loss": 0.8381,
62
+ "step": 9
63
+ },
64
+ {
65
+ "epoch": 0.04,
66
+ "learning_rate": 1.2412143126717115e-05,
67
+ "loss": 0.8255,
68
+ "step": 10
69
+ },
70
+ {
71
+ "epoch": 0.04,
72
+ "learning_rate": 1.3263883767450225e-05,
73
+ "loss": 0.79,
74
+ "step": 11
75
+ },
76
+ {
77
+ "epoch": 0.05,
78
+ "learning_rate": 1.4015172256888441e-05,
79
+ "loss": 0.8393,
80
+ "step": 12
81
+ },
82
+ {
83
+ "epoch": 0.05,
84
+ "learning_rate": 1.4687222711471113e-05,
85
+ "loss": 0.6941,
86
+ "step": 13
87
+ },
88
+ {
89
+ "epoch": 0.06,
90
+ "learning_rate": 1.529516629701577e-05,
91
+ "loss": 0.6893,
92
+ "step": 14
93
+ },
94
+ {
95
+ "epoch": 0.06,
96
+ "learning_rate": 1.585017530674437e-05,
97
+ "loss": 0.7529,
98
+ "step": 15
99
+ },
100
+ {
101
+ "epoch": 0.06,
102
+ "learning_rate": 1.6360734103293242e-05,
103
+ "loss": 0.7648,
104
+ "step": 16
105
+ },
106
+ {
107
+ "epoch": 0.07,
108
+ "learning_rate": 1.683343771586719e-05,
109
+ "loss": 0.7482,
110
+ "step": 17
111
+ },
112
+ {
113
+ "epoch": 0.07,
114
+ "learning_rate": 1.7273514250765258e-05,
115
+ "loss": 0.75,
116
+ "step": 18
117
+ },
118
+ {
119
+ "epoch": 0.08,
120
+ "learning_rate": 1.76851783566003e-05,
121
+ "loss": 0.7613,
122
+ "step": 19
123
+ },
124
+ {
125
+ "epoch": 0.08,
126
+ "learning_rate": 1.807187734337212e-05,
127
+ "loss": 0.7992,
128
+ "step": 20
129
+ },
130
+ {
131
+ "epoch": 0.08,
132
+ "learning_rate": 1.8436466846038513e-05,
133
+ "loss": 0.6974,
134
+ "step": 21
135
+ },
136
+ {
137
+ "epoch": 0.09,
138
+ "learning_rate": 1.8781338930289855e-05,
139
+ "loss": 0.7481,
140
+ "step": 22
141
+ },
142
+ {
143
+ "epoch": 0.09,
144
+ "learning_rate": 1.9108517300621187e-05,
145
+ "loss": 0.7528,
146
+ "step": 23
147
+ },
148
+ {
149
+ "epoch": 0.1,
150
+ "learning_rate": 1.9419729255161336e-05,
151
+ "loss": 0.6867,
152
+ "step": 24
153
+ },
154
+ {
155
+ "epoch": 0.1,
156
+ "learning_rate": 1.9716460886165846e-05,
157
+ "loss": 0.7166,
158
+ "step": 25
159
+ },
160
+ {
161
+ "epoch": 0.1,
162
+ "learning_rate": 2e-05,
163
+ "loss": 0.7035,
164
+ "step": 26
165
+ },
166
+ {
167
+ "epoch": 0.11,
168
+ "learning_rate": 2e-05,
169
+ "loss": 0.8703,
170
+ "step": 27
171
+ },
172
+ {
173
+ "epoch": 0.11,
174
+ "learning_rate": 1.9972489683631363e-05,
175
+ "loss": 0.7793,
176
+ "step": 28
177
+ },
178
+ {
179
+ "epoch": 0.12,
180
+ "learning_rate": 1.9944979367262724e-05,
181
+ "loss": 0.7478,
182
+ "step": 29
183
+ },
184
+ {
185
+ "epoch": 0.12,
186
+ "learning_rate": 1.991746905089409e-05,
187
+ "loss": 0.7143,
188
+ "step": 30
189
+ },
190
+ {
191
+ "epoch": 0.12,
192
+ "learning_rate": 1.988995873452545e-05,
193
+ "loss": 0.7491,
194
+ "step": 31
195
+ },
196
+ {
197
+ "epoch": 0.13,
198
+ "learning_rate": 1.986244841815681e-05,
199
+ "loss": 0.7789,
200
+ "step": 32
201
+ },
202
+ {
203
+ "epoch": 0.13,
204
+ "learning_rate": 1.9834938101788172e-05,
205
+ "loss": 0.7516,
206
+ "step": 33
207
+ },
208
+ {
209
+ "epoch": 0.14,
210
+ "learning_rate": 1.9807427785419533e-05,
211
+ "loss": 0.796,
212
+ "step": 34
213
+ },
214
+ {
215
+ "epoch": 0.14,
216
+ "learning_rate": 1.9779917469050895e-05,
217
+ "loss": 0.7726,
218
+ "step": 35
219
+ },
220
+ {
221
+ "epoch": 0.14,
222
+ "learning_rate": 1.975240715268226e-05,
223
+ "loss": 0.6836,
224
+ "step": 36
225
+ },
226
+ {
227
+ "epoch": 0.15,
228
+ "learning_rate": 1.9724896836313617e-05,
229
+ "loss": 0.7613,
230
+ "step": 37
231
+ },
232
+ {
233
+ "epoch": 0.15,
234
+ "learning_rate": 1.9697386519944982e-05,
235
+ "loss": 0.6996,
236
+ "step": 38
237
+ },
238
+ {
239
+ "epoch": 0.16,
240
+ "learning_rate": 1.9669876203576343e-05,
241
+ "loss": 0.6864,
242
+ "step": 39
243
+ },
244
+ {
245
+ "epoch": 0.16,
246
+ "learning_rate": 1.9642365887207704e-05,
247
+ "loss": 0.7336,
248
+ "step": 40
249
+ },
250
+ {
251
+ "epoch": 0.16,
252
+ "learning_rate": 1.9614855570839065e-05,
253
+ "loss": 0.6163,
254
+ "step": 41
255
+ },
256
+ {
257
+ "epoch": 0.17,
258
+ "learning_rate": 1.958734525447043e-05,
259
+ "loss": 0.7216,
260
+ "step": 42
261
+ },
262
+ {
263
+ "epoch": 0.17,
264
+ "learning_rate": 1.9559834938101788e-05,
265
+ "loss": 0.6686,
266
+ "step": 43
267
+ },
268
+ {
269
+ "epoch": 0.18,
270
+ "learning_rate": 1.9532324621733152e-05,
271
+ "loss": 0.7211,
272
+ "step": 44
273
+ },
274
+ {
275
+ "epoch": 0.18,
276
+ "learning_rate": 1.9504814305364514e-05,
277
+ "loss": 0.8343,
278
+ "step": 45
279
+ },
280
+ {
281
+ "epoch": 0.18,
282
+ "learning_rate": 1.9477303988995875e-05,
283
+ "loss": 0.8646,
284
+ "step": 46
285
+ },
286
+ {
287
+ "epoch": 0.19,
288
+ "learning_rate": 1.9449793672627236e-05,
289
+ "loss": 0.8046,
290
+ "step": 47
291
+ },
292
+ {
293
+ "epoch": 0.19,
294
+ "learning_rate": 1.9422283356258597e-05,
295
+ "loss": 0.7184,
296
+ "step": 48
297
+ },
298
+ {
299
+ "epoch": 0.2,
300
+ "learning_rate": 1.9394773039889962e-05,
301
+ "loss": 0.7182,
302
+ "step": 49
303
+ },
304
+ {
305
+ "epoch": 0.2,
306
+ "learning_rate": 1.9367262723521323e-05,
307
+ "loss": 0.8039,
308
+ "step": 50
309
+ },
310
+ {
311
+ "epoch": 0.2,
312
+ "learning_rate": 1.9339752407152684e-05,
313
+ "loss": 0.6945,
314
+ "step": 51
315
+ },
316
+ {
317
+ "epoch": 0.21,
318
+ "learning_rate": 1.9312242090784045e-05,
319
+ "loss": 0.746,
320
+ "step": 52
321
+ },
322
+ {
323
+ "epoch": 0.21,
324
+ "learning_rate": 1.928473177441541e-05,
325
+ "loss": 0.7949,
326
+ "step": 53
327
+ },
328
+ {
329
+ "epoch": 0.22,
330
+ "learning_rate": 1.9257221458046768e-05,
331
+ "loss": 0.7124,
332
+ "step": 54
333
+ },
334
+ {
335
+ "epoch": 0.22,
336
+ "learning_rate": 1.9229711141678132e-05,
337
+ "loss": 0.7368,
338
+ "step": 55
339
+ },
340
+ {
341
+ "epoch": 0.22,
342
+ "learning_rate": 1.920220082530949e-05,
343
+ "loss": 0.7541,
344
+ "step": 56
345
+ },
346
+ {
347
+ "epoch": 0.23,
348
+ "learning_rate": 1.9174690508940855e-05,
349
+ "loss": 0.8161,
350
+ "step": 57
351
+ },
352
+ {
353
+ "epoch": 0.23,
354
+ "learning_rate": 1.9147180192572216e-05,
355
+ "loss": 0.7764,
356
+ "step": 58
357
+ },
358
+ {
359
+ "epoch": 0.24,
360
+ "learning_rate": 1.9119669876203577e-05,
361
+ "loss": 0.7333,
362
+ "step": 59
363
+ },
364
+ {
365
+ "epoch": 0.24,
366
+ "learning_rate": 1.909215955983494e-05,
367
+ "loss": 0.7668,
368
+ "step": 60
369
+ },
370
+ {
371
+ "epoch": 0.24,
372
+ "learning_rate": 1.9064649243466303e-05,
373
+ "loss": 0.6952,
374
+ "step": 61
375
+ },
376
+ {
377
+ "epoch": 0.25,
378
+ "learning_rate": 1.903713892709766e-05,
379
+ "loss": 0.7621,
380
+ "step": 62
381
+ },
382
+ {
383
+ "epoch": 0.25,
384
+ "learning_rate": 1.9009628610729025e-05,
385
+ "loss": 0.7054,
386
+ "step": 63
387
+ },
388
+ {
389
+ "epoch": 0.26,
390
+ "learning_rate": 1.8982118294360387e-05,
391
+ "loss": 0.7501,
392
+ "step": 64
393
+ },
394
+ {
395
+ "epoch": 0.26,
396
+ "learning_rate": 1.8954607977991748e-05,
397
+ "loss": 0.7345,
398
+ "step": 65
399
+ },
400
+ {
401
+ "epoch": 0.26,
402
+ "learning_rate": 1.892709766162311e-05,
403
+ "loss": 0.7633,
404
+ "step": 66
405
+ },
406
+ {
407
+ "epoch": 0.27,
408
+ "learning_rate": 1.889958734525447e-05,
409
+ "loss": 0.7899,
410
+ "step": 67
411
+ },
412
+ {
413
+ "epoch": 0.27,
414
+ "learning_rate": 1.8872077028885835e-05,
415
+ "loss": 0.6917,
416
+ "step": 68
417
+ },
418
+ {
419
+ "epoch": 0.28,
420
+ "learning_rate": 1.8844566712517196e-05,
421
+ "loss": 0.8104,
422
+ "step": 69
423
+ },
424
+ {
425
+ "epoch": 0.28,
426
+ "learning_rate": 1.8817056396148557e-05,
427
+ "loss": 0.7783,
428
+ "step": 70
429
+ },
430
+ {
431
+ "epoch": 0.28,
432
+ "learning_rate": 1.878954607977992e-05,
433
+ "loss": 0.7156,
434
+ "step": 71
435
+ },
436
+ {
437
+ "epoch": 0.29,
438
+ "learning_rate": 1.8762035763411283e-05,
439
+ "loss": 0.6756,
440
+ "step": 72
441
+ },
442
+ {
443
+ "epoch": 0.29,
444
+ "learning_rate": 1.873452544704264e-05,
445
+ "loss": 0.7048,
446
+ "step": 73
447
+ },
448
+ {
449
+ "epoch": 0.3,
450
+ "learning_rate": 1.8707015130674006e-05,
451
+ "loss": 0.8007,
452
+ "step": 74
453
+ },
454
+ {
455
+ "epoch": 0.3,
456
+ "learning_rate": 1.8679504814305367e-05,
457
+ "loss": 0.8125,
458
+ "step": 75
459
+ },
460
+ {
461
+ "epoch": 0.3,
462
+ "learning_rate": 1.8651994497936728e-05,
463
+ "loss": 0.7305,
464
+ "step": 76
465
+ },
466
+ {
467
+ "epoch": 0.31,
468
+ "learning_rate": 1.862448418156809e-05,
469
+ "loss": 0.742,
470
+ "step": 77
471
+ },
472
+ {
473
+ "epoch": 0.31,
474
+ "learning_rate": 1.859697386519945e-05,
475
+ "loss": 0.7163,
476
+ "step": 78
477
+ },
478
+ {
479
+ "epoch": 0.32,
480
+ "learning_rate": 1.856946354883081e-05,
481
+ "loss": 0.7119,
482
+ "step": 79
483
+ },
484
+ {
485
+ "epoch": 0.32,
486
+ "learning_rate": 1.8541953232462176e-05,
487
+ "loss": 0.6405,
488
+ "step": 80
489
+ },
490
+ {
491
+ "epoch": 0.32,
492
+ "learning_rate": 1.8514442916093537e-05,
493
+ "loss": 0.8169,
494
+ "step": 81
495
+ },
496
+ {
497
+ "epoch": 0.33,
498
+ "learning_rate": 1.84869325997249e-05,
499
+ "loss": 0.7386,
500
+ "step": 82
501
+ },
502
+ {
503
+ "epoch": 0.33,
504
+ "learning_rate": 1.845942228335626e-05,
505
+ "loss": 0.8507,
506
+ "step": 83
507
+ },
508
+ {
509
+ "epoch": 0.34,
510
+ "learning_rate": 1.843191196698762e-05,
511
+ "loss": 0.6441,
512
+ "step": 84
513
+ },
514
+ {
515
+ "epoch": 0.34,
516
+ "learning_rate": 1.8404401650618982e-05,
517
+ "loss": 0.7248,
518
+ "step": 85
519
+ },
520
+ {
521
+ "epoch": 0.34,
522
+ "learning_rate": 1.8376891334250347e-05,
523
+ "loss": 0.8051,
524
+ "step": 86
525
+ },
526
+ {
527
+ "epoch": 0.35,
528
+ "learning_rate": 1.8349381017881708e-05,
529
+ "loss": 0.7272,
530
+ "step": 87
531
+ },
532
+ {
533
+ "epoch": 0.35,
534
+ "learning_rate": 1.832187070151307e-05,
535
+ "loss": 0.6879,
536
+ "step": 88
537
+ },
538
+ {
539
+ "epoch": 0.36,
540
+ "learning_rate": 1.829436038514443e-05,
541
+ "loss": 0.7123,
542
+ "step": 89
543
+ },
544
+ {
545
+ "epoch": 0.36,
546
+ "learning_rate": 1.826685006877579e-05,
547
+ "loss": 0.6777,
548
+ "step": 90
549
+ },
550
+ {
551
+ "epoch": 0.36,
552
+ "learning_rate": 1.8239339752407156e-05,
553
+ "loss": 0.6839,
554
+ "step": 91
555
+ },
556
+ {
557
+ "epoch": 0.37,
558
+ "learning_rate": 1.8211829436038514e-05,
559
+ "loss": 0.7528,
560
+ "step": 92
561
+ },
562
+ {
563
+ "epoch": 0.37,
564
+ "learning_rate": 1.818431911966988e-05,
565
+ "loss": 0.734,
566
+ "step": 93
567
+ },
568
+ {
569
+ "epoch": 0.38,
570
+ "learning_rate": 1.815680880330124e-05,
571
+ "loss": 0.6664,
572
+ "step": 94
573
+ },
574
+ {
575
+ "epoch": 0.38,
576
+ "learning_rate": 1.81292984869326e-05,
577
+ "loss": 0.6881,
578
+ "step": 95
579
+ },
580
+ {
581
+ "epoch": 0.38,
582
+ "learning_rate": 1.8101788170563962e-05,
583
+ "loss": 0.7597,
584
+ "step": 96
585
+ },
586
+ {
587
+ "epoch": 0.39,
588
+ "learning_rate": 1.8074277854195327e-05,
589
+ "loss": 0.7656,
590
+ "step": 97
591
+ },
592
+ {
593
+ "epoch": 0.39,
594
+ "learning_rate": 1.8046767537826685e-05,
595
+ "loss": 0.7762,
596
+ "step": 98
597
+ },
598
+ {
599
+ "epoch": 0.4,
600
+ "learning_rate": 1.801925722145805e-05,
601
+ "loss": 0.7182,
602
+ "step": 99
603
+ },
604
+ {
605
+ "epoch": 0.4,
606
+ "learning_rate": 1.799174690508941e-05,
607
+ "loss": 0.7527,
608
+ "step": 100
609
+ },
610
+ {
611
+ "epoch": 0.4,
612
+ "learning_rate": 1.7964236588720772e-05,
613
+ "loss": 0.7343,
614
+ "step": 101
615
+ },
616
+ {
617
+ "epoch": 0.41,
618
+ "learning_rate": 1.7936726272352133e-05,
619
+ "loss": 0.8163,
620
+ "step": 102
621
+ },
622
+ {
623
+ "epoch": 0.41,
624
+ "learning_rate": 1.7909215955983494e-05,
625
+ "loss": 0.721,
626
+ "step": 103
627
+ },
628
+ {
629
+ "epoch": 0.42,
630
+ "learning_rate": 1.7881705639614855e-05,
631
+ "loss": 0.7631,
632
+ "step": 104
633
+ },
634
+ {
635
+ "epoch": 0.42,
636
+ "learning_rate": 1.785419532324622e-05,
637
+ "loss": 0.7249,
638
+ "step": 105
639
+ },
640
+ {
641
+ "epoch": 0.42,
642
+ "learning_rate": 1.782668500687758e-05,
643
+ "loss": 0.6999,
644
+ "step": 106
645
+ },
646
+ {
647
+ "epoch": 0.43,
648
+ "learning_rate": 1.7799174690508942e-05,
649
+ "loss": 0.7409,
650
+ "step": 107
651
+ },
652
+ {
653
+ "epoch": 0.43,
654
+ "learning_rate": 1.7771664374140304e-05,
655
+ "loss": 0.7594,
656
+ "step": 108
657
+ },
658
+ {
659
+ "epoch": 0.44,
660
+ "learning_rate": 1.7744154057771665e-05,
661
+ "loss": 0.7373,
662
+ "step": 109
663
+ },
664
+ {
665
+ "epoch": 0.44,
666
+ "learning_rate": 1.771664374140303e-05,
667
+ "loss": 0.7606,
668
+ "step": 110
669
+ },
670
+ {
671
+ "epoch": 0.44,
672
+ "learning_rate": 1.7689133425034387e-05,
673
+ "loss": 0.7624,
674
+ "step": 111
675
+ },
676
+ {
677
+ "epoch": 0.45,
678
+ "learning_rate": 1.7661623108665752e-05,
679
+ "loss": 0.7627,
680
+ "step": 112
681
+ },
682
+ {
683
+ "epoch": 0.45,
684
+ "learning_rate": 1.7634112792297113e-05,
685
+ "loss": 0.717,
686
+ "step": 113
687
+ },
688
+ {
689
+ "epoch": 0.46,
690
+ "learning_rate": 1.7606602475928474e-05,
691
+ "loss": 0.7427,
692
+ "step": 114
693
+ },
694
+ {
695
+ "epoch": 0.46,
696
+ "learning_rate": 1.7579092159559835e-05,
697
+ "loss": 0.6982,
698
+ "step": 115
699
+ },
700
+ {
701
+ "epoch": 0.46,
702
+ "learning_rate": 1.75515818431912e-05,
703
+ "loss": 0.7009,
704
+ "step": 116
705
+ },
706
+ {
707
+ "epoch": 0.47,
708
+ "learning_rate": 1.7524071526822558e-05,
709
+ "loss": 0.6482,
710
+ "step": 117
711
+ },
712
+ {
713
+ "epoch": 0.47,
714
+ "learning_rate": 1.7496561210453922e-05,
715
+ "loss": 0.6886,
716
+ "step": 118
717
+ },
718
+ {
719
+ "epoch": 0.48,
720
+ "learning_rate": 1.7469050894085284e-05,
721
+ "loss": 0.7181,
722
+ "step": 119
723
+ },
724
+ {
725
+ "epoch": 0.48,
726
+ "learning_rate": 1.7441540577716645e-05,
727
+ "loss": 0.7489,
728
+ "step": 120
729
+ },
730
+ {
731
+ "epoch": 0.48,
732
+ "learning_rate": 1.7414030261348006e-05,
733
+ "loss": 0.7776,
734
+ "step": 121
735
+ },
736
+ {
737
+ "epoch": 0.49,
738
+ "learning_rate": 1.738651994497937e-05,
739
+ "loss": 0.7156,
740
+ "step": 122
741
+ },
742
+ {
743
+ "epoch": 0.49,
744
+ "learning_rate": 1.735900962861073e-05,
745
+ "loss": 0.7206,
746
+ "step": 123
747
+ },
748
+ {
749
+ "epoch": 0.5,
750
+ "learning_rate": 1.7331499312242093e-05,
751
+ "loss": 0.6703,
752
+ "step": 124
753
+ },
754
+ {
755
+ "epoch": 0.5,
756
+ "learning_rate": 1.7303988995873454e-05,
757
+ "loss": 0.7133,
758
+ "step": 125
759
+ },
760
+ {
761
+ "epoch": 0.5,
762
+ "learning_rate": 1.7276478679504815e-05,
763
+ "loss": 0.7194,
764
+ "step": 126
765
+ },
766
+ {
767
+ "epoch": 0.51,
768
+ "learning_rate": 1.7248968363136177e-05,
769
+ "loss": 0.737,
770
+ "step": 127
771
+ },
772
+ {
773
+ "epoch": 0.51,
774
+ "learning_rate": 1.7221458046767538e-05,
775
+ "loss": 0.6835,
776
+ "step": 128
777
+ },
778
+ {
779
+ "epoch": 0.52,
780
+ "learning_rate": 1.7193947730398902e-05,
781
+ "loss": 0.7644,
782
+ "step": 129
783
+ },
784
+ {
785
+ "epoch": 0.52,
786
+ "learning_rate": 1.7166437414030264e-05,
787
+ "loss": 0.7093,
788
+ "step": 130
789
+ },
790
+ {
791
+ "epoch": 0.52,
792
+ "learning_rate": 1.7138927097661625e-05,
793
+ "loss": 0.6793,
794
+ "step": 131
795
+ },
796
+ {
797
+ "epoch": 0.53,
798
+ "learning_rate": 1.7111416781292986e-05,
799
+ "loss": 0.712,
800
+ "step": 132
801
+ },
802
+ {
803
+ "epoch": 0.53,
804
+ "learning_rate": 1.708390646492435e-05,
805
+ "loss": 0.6982,
806
+ "step": 133
807
+ },
808
+ {
809
+ "epoch": 0.54,
810
+ "learning_rate": 1.705639614855571e-05,
811
+ "loss": 0.7428,
812
+ "step": 134
813
+ },
814
+ {
815
+ "epoch": 0.54,
816
+ "learning_rate": 1.7028885832187073e-05,
817
+ "loss": 0.739,
818
+ "step": 135
819
+ },
820
+ {
821
+ "epoch": 0.54,
822
+ "learning_rate": 1.700137551581843e-05,
823
+ "loss": 0.65,
824
+ "step": 136
825
+ },
826
+ {
827
+ "epoch": 0.55,
828
+ "learning_rate": 1.6973865199449796e-05,
829
+ "loss": 0.6436,
830
+ "step": 137
831
+ },
832
+ {
833
+ "epoch": 0.55,
834
+ "learning_rate": 1.6946354883081157e-05,
835
+ "loss": 0.7373,
836
+ "step": 138
837
+ },
838
+ {
839
+ "epoch": 0.56,
840
+ "learning_rate": 1.6918844566712518e-05,
841
+ "loss": 0.6389,
842
+ "step": 139
843
+ },
844
+ {
845
+ "epoch": 0.56,
846
+ "learning_rate": 1.689133425034388e-05,
847
+ "loss": 0.7067,
848
+ "step": 140
849
+ },
850
+ {
851
+ "epoch": 0.56,
852
+ "learning_rate": 1.6863823933975244e-05,
853
+ "loss": 0.7783,
854
+ "step": 141
855
+ },
856
+ {
857
+ "epoch": 0.57,
858
+ "learning_rate": 1.6836313617606605e-05,
859
+ "loss": 0.8668,
860
+ "step": 142
861
+ },
862
+ {
863
+ "epoch": 0.57,
864
+ "learning_rate": 1.6808803301237966e-05,
865
+ "loss": 0.7307,
866
+ "step": 143
867
+ },
868
+ {
869
+ "epoch": 0.58,
870
+ "learning_rate": 1.6781292984869327e-05,
871
+ "loss": 0.719,
872
+ "step": 144
873
+ },
874
+ {
875
+ "epoch": 0.58,
876
+ "learning_rate": 1.675378266850069e-05,
877
+ "loss": 0.8011,
878
+ "step": 145
879
+ },
880
+ {
881
+ "epoch": 0.58,
882
+ "learning_rate": 1.672627235213205e-05,
883
+ "loss": 0.7748,
884
+ "step": 146
885
+ },
886
+ {
887
+ "epoch": 0.59,
888
+ "learning_rate": 1.669876203576341e-05,
889
+ "loss": 0.702,
890
+ "step": 147
891
+ },
892
+ {
893
+ "epoch": 0.59,
894
+ "learning_rate": 1.6671251719394776e-05,
895
+ "loss": 0.638,
896
+ "step": 148
897
+ },
898
+ {
899
+ "epoch": 0.6,
900
+ "learning_rate": 1.6643741403026137e-05,
901
+ "loss": 0.7378,
902
+ "step": 149
903
+ },
904
+ {
905
+ "epoch": 0.6,
906
+ "learning_rate": 1.6616231086657498e-05,
907
+ "loss": 0.6801,
908
+ "step": 150
909
+ },
910
+ {
911
+ "epoch": 0.6,
912
+ "learning_rate": 1.658872077028886e-05,
913
+ "loss": 0.796,
914
+ "step": 151
915
+ },
916
+ {
917
+ "epoch": 0.61,
918
+ "learning_rate": 1.6561210453920224e-05,
919
+ "loss": 0.7391,
920
+ "step": 152
921
+ },
922
+ {
923
+ "epoch": 0.61,
924
+ "learning_rate": 1.653370013755158e-05,
925
+ "loss": 0.7216,
926
+ "step": 153
927
+ },
928
+ {
929
+ "epoch": 0.62,
930
+ "learning_rate": 1.6506189821182946e-05,
931
+ "loss": 0.7182,
932
+ "step": 154
933
+ },
934
+ {
935
+ "epoch": 0.62,
936
+ "learning_rate": 1.6478679504814307e-05,
937
+ "loss": 0.6991,
938
+ "step": 155
939
+ },
940
+ {
941
+ "epoch": 0.62,
942
+ "learning_rate": 1.645116918844567e-05,
943
+ "loss": 0.6555,
944
+ "step": 156
945
+ },
946
+ {
947
+ "epoch": 0.63,
948
+ "learning_rate": 1.642365887207703e-05,
949
+ "loss": 0.6516,
950
+ "step": 157
951
+ },
952
+ {
953
+ "epoch": 0.63,
954
+ "learning_rate": 1.639614855570839e-05,
955
+ "loss": 0.728,
956
+ "step": 158
957
+ },
958
+ {
959
+ "epoch": 0.64,
960
+ "learning_rate": 1.6368638239339752e-05,
961
+ "loss": 0.7357,
962
+ "step": 159
963
+ },
964
+ {
965
+ "epoch": 0.64,
966
+ "learning_rate": 1.6341127922971117e-05,
967
+ "loss": 0.6942,
968
+ "step": 160
969
+ },
970
+ {
971
+ "epoch": 0.64,
972
+ "learning_rate": 1.6313617606602478e-05,
973
+ "loss": 0.7153,
974
+ "step": 161
975
+ },
976
+ {
977
+ "epoch": 0.65,
978
+ "learning_rate": 1.628610729023384e-05,
979
+ "loss": 0.7583,
980
+ "step": 162
981
+ },
982
+ {
983
+ "epoch": 0.65,
984
+ "learning_rate": 1.62585969738652e-05,
985
+ "loss": 0.6917,
986
+ "step": 163
987
+ },
988
+ {
989
+ "epoch": 0.66,
990
+ "learning_rate": 1.6231086657496562e-05,
991
+ "loss": 0.6655,
992
+ "step": 164
993
+ },
994
+ {
995
+ "epoch": 0.66,
996
+ "learning_rate": 1.6203576341127923e-05,
997
+ "loss": 0.7071,
998
+ "step": 165
999
+ },
1000
+ {
1001
+ "epoch": 0.66,
1002
+ "learning_rate": 1.6176066024759288e-05,
1003
+ "loss": 0.6545,
1004
+ "step": 166
1005
+ },
1006
+ {
1007
+ "epoch": 0.67,
1008
+ "learning_rate": 1.614855570839065e-05,
1009
+ "loss": 0.663,
1010
+ "step": 167
1011
+ },
1012
+ {
1013
+ "epoch": 0.67,
1014
+ "learning_rate": 1.612104539202201e-05,
1015
+ "loss": 0.7543,
1016
+ "step": 168
1017
+ },
1018
+ {
1019
+ "epoch": 0.68,
1020
+ "learning_rate": 1.609353507565337e-05,
1021
+ "loss": 0.6838,
1022
+ "step": 169
1023
+ },
1024
+ {
1025
+ "epoch": 0.68,
1026
+ "learning_rate": 1.6066024759284732e-05,
1027
+ "loss": 0.6506,
1028
+ "step": 170
1029
+ },
1030
+ {
1031
+ "epoch": 0.68,
1032
+ "learning_rate": 1.6038514442916097e-05,
1033
+ "loss": 0.6964,
1034
+ "step": 171
1035
+ },
1036
+ {
1037
+ "epoch": 0.69,
1038
+ "learning_rate": 1.6011004126547455e-05,
1039
+ "loss": 0.7175,
1040
+ "step": 172
1041
+ },
1042
+ {
1043
+ "epoch": 0.69,
1044
+ "learning_rate": 1.598349381017882e-05,
1045
+ "loss": 0.68,
1046
+ "step": 173
1047
+ },
1048
+ {
1049
+ "epoch": 0.7,
1050
+ "learning_rate": 1.595598349381018e-05,
1051
+ "loss": 0.6289,
1052
+ "step": 174
1053
+ },
1054
+ {
1055
+ "epoch": 0.7,
1056
+ "learning_rate": 1.5928473177441542e-05,
1057
+ "loss": 0.6606,
1058
+ "step": 175
1059
+ },
1060
+ {
1061
+ "epoch": 0.7,
1062
+ "learning_rate": 1.5900962861072903e-05,
1063
+ "loss": 0.7564,
1064
+ "step": 176
1065
+ },
1066
+ {
1067
+ "epoch": 0.71,
1068
+ "learning_rate": 1.5873452544704268e-05,
1069
+ "loss": 0.6964,
1070
+ "step": 177
1071
+ },
1072
+ {
1073
+ "epoch": 0.71,
1074
+ "learning_rate": 1.5845942228335625e-05,
1075
+ "loss": 0.6616,
1076
+ "step": 178
1077
+ },
1078
+ {
1079
+ "epoch": 0.72,
1080
+ "learning_rate": 1.581843191196699e-05,
1081
+ "loss": 0.7258,
1082
+ "step": 179
1083
+ },
1084
+ {
1085
+ "epoch": 0.72,
1086
+ "learning_rate": 1.579092159559835e-05,
1087
+ "loss": 0.7294,
1088
+ "step": 180
1089
+ },
1090
+ {
1091
+ "epoch": 0.72,
1092
+ "learning_rate": 1.5763411279229712e-05,
1093
+ "loss": 0.7013,
1094
+ "step": 181
1095
+ },
1096
+ {
1097
+ "epoch": 0.73,
1098
+ "learning_rate": 1.5735900962861074e-05,
1099
+ "loss": 0.6723,
1100
+ "step": 182
1101
+ },
1102
+ {
1103
+ "epoch": 0.73,
1104
+ "learning_rate": 1.5708390646492435e-05,
1105
+ "loss": 0.6372,
1106
+ "step": 183
1107
+ },
1108
+ {
1109
+ "epoch": 0.74,
1110
+ "learning_rate": 1.5680880330123796e-05,
1111
+ "loss": 0.7439,
1112
+ "step": 184
1113
+ },
1114
+ {
1115
+ "epoch": 0.74,
1116
+ "learning_rate": 1.565337001375516e-05,
1117
+ "loss": 0.6167,
1118
+ "step": 185
1119
+ },
1120
+ {
1121
+ "epoch": 0.74,
1122
+ "learning_rate": 1.5625859697386522e-05,
1123
+ "loss": 0.6366,
1124
+ "step": 186
1125
+ },
1126
+ {
1127
+ "epoch": 0.75,
1128
+ "learning_rate": 1.5598349381017883e-05,
1129
+ "loss": 0.7574,
1130
+ "step": 187
1131
+ },
1132
+ {
1133
+ "epoch": 0.75,
1134
+ "learning_rate": 1.5570839064649244e-05,
1135
+ "loss": 0.7614,
1136
+ "step": 188
1137
+ },
1138
+ {
1139
+ "epoch": 0.76,
1140
+ "learning_rate": 1.5543328748280606e-05,
1141
+ "loss": 0.7012,
1142
+ "step": 189
1143
+ },
1144
+ {
1145
+ "epoch": 0.76,
1146
+ "learning_rate": 1.551581843191197e-05,
1147
+ "loss": 0.65,
1148
+ "step": 190
1149
+ },
1150
+ {
1151
+ "epoch": 0.76,
1152
+ "learning_rate": 1.5488308115543328e-05,
1153
+ "loss": 0.7321,
1154
+ "step": 191
1155
+ },
1156
+ {
1157
+ "epoch": 0.77,
1158
+ "learning_rate": 1.5460797799174693e-05,
1159
+ "loss": 0.6749,
1160
+ "step": 192
1161
+ },
1162
+ {
1163
+ "epoch": 0.77,
1164
+ "learning_rate": 1.5433287482806054e-05,
1165
+ "loss": 0.7115,
1166
+ "step": 193
1167
+ },
1168
+ {
1169
+ "epoch": 0.78,
1170
+ "learning_rate": 1.5405777166437415e-05,
1171
+ "loss": 0.7541,
1172
+ "step": 194
1173
+ },
1174
+ {
1175
+ "epoch": 0.78,
1176
+ "learning_rate": 1.5378266850068776e-05,
1177
+ "loss": 0.7499,
1178
+ "step": 195
1179
+ },
1180
+ {
1181
+ "epoch": 0.78,
1182
+ "learning_rate": 1.535075653370014e-05,
1183
+ "loss": 0.7071,
1184
+ "step": 196
1185
+ },
1186
+ {
1187
+ "epoch": 0.79,
1188
+ "learning_rate": 1.53232462173315e-05,
1189
+ "loss": 0.7087,
1190
+ "step": 197
1191
+ },
1192
+ {
1193
+ "epoch": 0.79,
1194
+ "learning_rate": 1.5295735900962863e-05,
1195
+ "loss": 0.7065,
1196
+ "step": 198
1197
+ },
1198
+ {
1199
+ "epoch": 0.8,
1200
+ "learning_rate": 1.5268225584594224e-05,
1201
+ "loss": 0.7079,
1202
+ "step": 199
1203
+ },
1204
+ {
1205
+ "epoch": 0.8,
1206
+ "learning_rate": 1.5240715268225586e-05,
1207
+ "loss": 0.7535,
1208
+ "step": 200
1209
+ },
1210
+ {
1211
+ "epoch": 0.8,
1212
+ "learning_rate": 1.5213204951856948e-05,
1213
+ "loss": 0.7611,
1214
+ "step": 201
1215
+ },
1216
+ {
1217
+ "epoch": 0.81,
1218
+ "learning_rate": 1.5185694635488308e-05,
1219
+ "loss": 0.7247,
1220
+ "step": 202
1221
+ },
1222
+ {
1223
+ "epoch": 0.81,
1224
+ "learning_rate": 1.5158184319119671e-05,
1225
+ "loss": 0.7209,
1226
+ "step": 203
1227
+ },
1228
+ {
1229
+ "epoch": 0.82,
1230
+ "learning_rate": 1.5130674002751034e-05,
1231
+ "loss": 0.7039,
1232
+ "step": 204
1233
+ },
1234
+ {
1235
+ "epoch": 0.82,
1236
+ "learning_rate": 1.5103163686382393e-05,
1237
+ "loss": 0.7522,
1238
+ "step": 205
1239
+ },
1240
+ {
1241
+ "epoch": 0.82,
1242
+ "learning_rate": 1.5075653370013756e-05,
1243
+ "loss": 0.6941,
1244
+ "step": 206
1245
+ },
1246
+ {
1247
+ "epoch": 0.83,
1248
+ "learning_rate": 1.5048143053645119e-05,
1249
+ "loss": 0.6611,
1250
+ "step": 207
1251
+ },
1252
+ {
1253
+ "epoch": 0.83,
1254
+ "learning_rate": 1.5020632737276479e-05,
1255
+ "loss": 0.7778,
1256
+ "step": 208
1257
+ },
1258
+ {
1259
+ "epoch": 0.84,
1260
+ "learning_rate": 1.4993122420907842e-05,
1261
+ "loss": 0.6705,
1262
+ "step": 209
1263
+ },
1264
+ {
1265
+ "epoch": 0.84,
1266
+ "learning_rate": 1.4965612104539204e-05,
1267
+ "loss": 0.6179,
1268
+ "step": 210
1269
+ },
1270
+ {
1271
+ "epoch": 0.84,
1272
+ "learning_rate": 1.4938101788170564e-05,
1273
+ "loss": 0.675,
1274
+ "step": 211
1275
+ },
1276
+ {
1277
+ "epoch": 0.85,
1278
+ "learning_rate": 1.4910591471801927e-05,
1279
+ "loss": 0.6672,
1280
+ "step": 212
1281
+ },
1282
+ {
1283
+ "epoch": 0.85,
1284
+ "learning_rate": 1.4883081155433288e-05,
1285
+ "loss": 0.7322,
1286
+ "step": 213
1287
+ },
1288
+ {
1289
+ "epoch": 0.86,
1290
+ "learning_rate": 1.4855570839064651e-05,
1291
+ "loss": 0.7011,
1292
+ "step": 214
1293
+ },
1294
+ {
1295
+ "epoch": 0.86,
1296
+ "learning_rate": 1.4828060522696012e-05,
1297
+ "loss": 0.7106,
1298
+ "step": 215
1299
+ },
1300
+ {
1301
+ "epoch": 0.86,
1302
+ "learning_rate": 1.4800550206327373e-05,
1303
+ "loss": 0.7264,
1304
+ "step": 216
1305
+ },
1306
+ {
1307
+ "epoch": 0.87,
1308
+ "learning_rate": 1.4773039889958736e-05,
1309
+ "loss": 0.7744,
1310
+ "step": 217
1311
+ },
1312
+ {
1313
+ "epoch": 0.87,
1314
+ "learning_rate": 1.4745529573590097e-05,
1315
+ "loss": 0.684,
1316
+ "step": 218
1317
+ },
1318
+ {
1319
+ "epoch": 0.88,
1320
+ "learning_rate": 1.4718019257221459e-05,
1321
+ "loss": 0.6544,
1322
+ "step": 219
1323
+ },
1324
+ {
1325
+ "epoch": 0.88,
1326
+ "learning_rate": 1.4690508940852822e-05,
1327
+ "loss": 0.7054,
1328
+ "step": 220
1329
+ },
1330
+ {
1331
+ "epoch": 0.88,
1332
+ "learning_rate": 1.4662998624484185e-05,
1333
+ "loss": 0.6471,
1334
+ "step": 221
1335
+ },
1336
+ {
1337
+ "epoch": 0.89,
1338
+ "learning_rate": 1.4635488308115544e-05,
1339
+ "loss": 0.6562,
1340
+ "step": 222
1341
+ },
1342
+ {
1343
+ "epoch": 0.89,
1344
+ "learning_rate": 1.4607977991746907e-05,
1345
+ "loss": 0.7364,
1346
+ "step": 223
1347
+ },
1348
+ {
1349
+ "epoch": 0.9,
1350
+ "learning_rate": 1.4580467675378266e-05,
1351
+ "loss": 0.752,
1352
+ "step": 224
1353
+ },
1354
+ {
1355
+ "epoch": 0.9,
1356
+ "learning_rate": 1.455295735900963e-05,
1357
+ "loss": 0.6894,
1358
+ "step": 225
1359
+ },
1360
+ {
1361
+ "epoch": 0.9,
1362
+ "learning_rate": 1.4525447042640992e-05,
1363
+ "loss": 0.7072,
1364
+ "step": 226
1365
+ },
1366
+ {
1367
+ "epoch": 0.91,
1368
+ "learning_rate": 1.4497936726272352e-05,
1369
+ "loss": 0.6649,
1370
+ "step": 227
1371
+ },
1372
+ {
1373
+ "epoch": 0.91,
1374
+ "learning_rate": 1.4470426409903715e-05,
1375
+ "loss": 0.7411,
1376
+ "step": 228
1377
+ },
1378
+ {
1379
+ "epoch": 0.92,
1380
+ "learning_rate": 1.4442916093535078e-05,
1381
+ "loss": 0.7079,
1382
+ "step": 229
1383
+ },
1384
+ {
1385
+ "epoch": 0.92,
1386
+ "learning_rate": 1.4415405777166439e-05,
1387
+ "loss": 0.6625,
1388
+ "step": 230
1389
+ },
1390
+ {
1391
+ "epoch": 0.92,
1392
+ "learning_rate": 1.43878954607978e-05,
1393
+ "loss": 0.7616,
1394
+ "step": 231
1395
+ },
1396
+ {
1397
+ "epoch": 0.93,
1398
+ "learning_rate": 1.4360385144429163e-05,
1399
+ "loss": 0.7437,
1400
+ "step": 232
1401
+ },
1402
+ {
1403
+ "epoch": 0.93,
1404
+ "learning_rate": 1.4332874828060524e-05,
1405
+ "loss": 0.7377,
1406
+ "step": 233
1407
+ },
1408
+ {
1409
+ "epoch": 0.94,
1410
+ "learning_rate": 1.4305364511691885e-05,
1411
+ "loss": 0.6102,
1412
+ "step": 234
1413
+ },
1414
+ {
1415
+ "epoch": 0.94,
1416
+ "learning_rate": 1.4277854195323247e-05,
1417
+ "loss": 0.7094,
1418
+ "step": 235
1419
+ },
1420
+ {
1421
+ "epoch": 0.94,
1422
+ "learning_rate": 1.425034387895461e-05,
1423
+ "loss": 0.694,
1424
+ "step": 236
1425
+ },
1426
+ {
1427
+ "epoch": 0.95,
1428
+ "learning_rate": 1.4222833562585972e-05,
1429
+ "loss": 0.6074,
1430
+ "step": 237
1431
+ },
1432
+ {
1433
+ "epoch": 0.95,
1434
+ "learning_rate": 1.4195323246217332e-05,
1435
+ "loss": 0.7402,
1436
+ "step": 238
1437
+ },
1438
+ {
1439
+ "epoch": 0.96,
1440
+ "learning_rate": 1.4167812929848695e-05,
1441
+ "loss": 0.6544,
1442
+ "step": 239
1443
+ },
1444
+ {
1445
+ "epoch": 0.96,
1446
+ "learning_rate": 1.4140302613480058e-05,
1447
+ "loss": 0.736,
1448
+ "step": 240
1449
+ },
1450
+ {
1451
+ "epoch": 0.96,
1452
+ "learning_rate": 1.4112792297111417e-05,
1453
+ "loss": 0.8055,
1454
+ "step": 241
1455
+ },
1456
+ {
1457
+ "epoch": 0.97,
1458
+ "learning_rate": 1.408528198074278e-05,
1459
+ "loss": 0.7056,
1460
+ "step": 242
1461
+ },
1462
+ {
1463
+ "epoch": 0.97,
1464
+ "learning_rate": 1.4057771664374143e-05,
1465
+ "loss": 0.7534,
1466
+ "step": 243
1467
+ },
1468
+ {
1469
+ "epoch": 0.98,
1470
+ "learning_rate": 1.4030261348005502e-05,
1471
+ "loss": 0.6616,
1472
+ "step": 244
1473
+ },
1474
+ {
1475
+ "epoch": 0.98,
1476
+ "learning_rate": 1.4002751031636865e-05,
1477
+ "loss": 0.6677,
1478
+ "step": 245
1479
+ },
1480
+ {
1481
+ "epoch": 0.98,
1482
+ "learning_rate": 1.3975240715268225e-05,
1483
+ "loss": 0.6832,
1484
+ "step": 246
1485
+ },
1486
+ {
1487
+ "epoch": 0.99,
1488
+ "learning_rate": 1.3947730398899588e-05,
1489
+ "loss": 0.68,
1490
+ "step": 247
1491
+ },
1492
+ {
1493
+ "epoch": 0.99,
1494
+ "learning_rate": 1.392022008253095e-05,
1495
+ "loss": 0.6881,
1496
+ "step": 248
1497
+ },
1498
+ {
1499
+ "epoch": 1.0,
1500
+ "learning_rate": 1.3892709766162312e-05,
1501
+ "loss": 0.6452,
1502
+ "step": 249
1503
+ },
1504
+ {
1505
+ "epoch": 1.0,
1506
+ "learning_rate": 1.3865199449793673e-05,
1507
+ "loss": 0.6573,
1508
+ "step": 250
1509
+ },
1510
+ {
1511
+ "epoch": 1.0,
1512
+ "learning_rate": 1.3837689133425036e-05,
1513
+ "loss": 0.3832,
1514
+ "step": 251
1515
+ },
1516
+ {
1517
+ "epoch": 1.01,
1518
+ "learning_rate": 1.3810178817056397e-05,
1519
+ "loss": 0.4213,
1520
+ "step": 252
1521
+ },
1522
+ {
1523
+ "epoch": 1.01,
1524
+ "learning_rate": 1.3782668500687758e-05,
1525
+ "loss": 0.3994,
1526
+ "step": 253
1527
+ },
1528
+ {
1529
+ "epoch": 1.02,
1530
+ "learning_rate": 1.3755158184319121e-05,
1531
+ "loss": 0.3177,
1532
+ "step": 254
1533
+ },
1534
+ {
1535
+ "epoch": 1.02,
1536
+ "learning_rate": 1.3727647867950483e-05,
1537
+ "loss": 0.3202,
1538
+ "step": 255
1539
+ },
1540
+ {
1541
+ "epoch": 1.02,
1542
+ "learning_rate": 1.3700137551581845e-05,
1543
+ "loss": 0.3,
1544
+ "step": 256
1545
+ },
1546
+ {
1547
+ "epoch": 1.03,
1548
+ "learning_rate": 1.3672627235213205e-05,
1549
+ "loss": 0.3263,
1550
+ "step": 257
1551
+ },
1552
+ {
1553
+ "epoch": 1.03,
1554
+ "learning_rate": 1.3645116918844568e-05,
1555
+ "loss": 0.3288,
1556
+ "step": 258
1557
+ },
1558
+ {
1559
+ "epoch": 1.04,
1560
+ "learning_rate": 1.361760660247593e-05,
1561
+ "loss": 0.3797,
1562
+ "step": 259
1563
+ },
1564
+ {
1565
+ "epoch": 1.04,
1566
+ "learning_rate": 1.359009628610729e-05,
1567
+ "loss": 0.3363,
1568
+ "step": 260
1569
+ },
1570
+ {
1571
+ "epoch": 1.04,
1572
+ "learning_rate": 1.3562585969738653e-05,
1573
+ "loss": 0.3725,
1574
+ "step": 261
1575
+ },
1576
+ {
1577
+ "epoch": 1.05,
1578
+ "learning_rate": 1.3535075653370016e-05,
1579
+ "loss": 0.3074,
1580
+ "step": 262
1581
+ },
1582
+ {
1583
+ "epoch": 1.05,
1584
+ "learning_rate": 1.3507565337001376e-05,
1585
+ "loss": 0.3079,
1586
+ "step": 263
1587
+ },
1588
+ {
1589
+ "epoch": 1.06,
1590
+ "learning_rate": 1.3480055020632738e-05,
1591
+ "loss": 0.2891,
1592
+ "step": 264
1593
+ },
1594
+ {
1595
+ "epoch": 1.06,
1596
+ "learning_rate": 1.3452544704264101e-05,
1597
+ "loss": 0.3187,
1598
+ "step": 265
1599
+ },
1600
+ {
1601
+ "epoch": 1.06,
1602
+ "learning_rate": 1.3425034387895461e-05,
1603
+ "loss": 0.2769,
1604
+ "step": 266
1605
+ },
1606
+ {
1607
+ "epoch": 1.07,
1608
+ "learning_rate": 1.3397524071526824e-05,
1609
+ "loss": 0.2724,
1610
+ "step": 267
1611
+ },
1612
+ {
1613
+ "epoch": 1.07,
1614
+ "learning_rate": 1.3370013755158185e-05,
1615
+ "loss": 0.3177,
1616
+ "step": 268
1617
+ },
1618
+ {
1619
+ "epoch": 1.08,
1620
+ "learning_rate": 1.3342503438789546e-05,
1621
+ "loss": 0.2724,
1622
+ "step": 269
1623
+ },
1624
+ {
1625
+ "epoch": 1.08,
1626
+ "learning_rate": 1.3314993122420909e-05,
1627
+ "loss": 0.2723,
1628
+ "step": 270
1629
+ },
1630
+ {
1631
+ "epoch": 1.08,
1632
+ "learning_rate": 1.328748280605227e-05,
1633
+ "loss": 0.3022,
1634
+ "step": 271
1635
+ },
1636
+ {
1637
+ "epoch": 1.09,
1638
+ "learning_rate": 1.3259972489683632e-05,
1639
+ "loss": 0.284,
1640
+ "step": 272
1641
+ },
1642
+ {
1643
+ "epoch": 1.09,
1644
+ "learning_rate": 1.3232462173314994e-05,
1645
+ "loss": 0.304,
1646
+ "step": 273
1647
+ },
1648
+ {
1649
+ "epoch": 1.1,
1650
+ "learning_rate": 1.3204951856946356e-05,
1651
+ "loss": 0.308,
1652
+ "step": 274
1653
+ },
1654
+ {
1655
+ "epoch": 1.1,
1656
+ "learning_rate": 1.3177441540577719e-05,
1657
+ "loss": 0.3128,
1658
+ "step": 275
1659
+ },
1660
+ {
1661
+ "epoch": 1.1,
1662
+ "learning_rate": 1.314993122420908e-05,
1663
+ "loss": 0.3224,
1664
+ "step": 276
1665
+ },
1666
+ {
1667
+ "epoch": 1.11,
1668
+ "learning_rate": 1.3122420907840441e-05,
1669
+ "loss": 0.3139,
1670
+ "step": 277
1671
+ },
1672
+ {
1673
+ "epoch": 1.11,
1674
+ "learning_rate": 1.3094910591471804e-05,
1675
+ "loss": 0.3198,
1676
+ "step": 278
1677
+ },
1678
+ {
1679
+ "epoch": 1.12,
1680
+ "learning_rate": 1.3067400275103163e-05,
1681
+ "loss": 0.2906,
1682
+ "step": 279
1683
+ },
1684
+ {
1685
+ "epoch": 1.12,
1686
+ "learning_rate": 1.3039889958734526e-05,
1687
+ "loss": 0.3141,
1688
+ "step": 280
1689
+ },
1690
+ {
1691
+ "epoch": 1.12,
1692
+ "learning_rate": 1.301237964236589e-05,
1693
+ "loss": 0.3022,
1694
+ "step": 281
1695
+ },
1696
+ {
1697
+ "epoch": 1.13,
1698
+ "learning_rate": 1.2984869325997249e-05,
1699
+ "loss": 0.3306,
1700
+ "step": 282
1701
+ },
1702
+ {
1703
+ "epoch": 1.13,
1704
+ "learning_rate": 1.2957359009628612e-05,
1705
+ "loss": 0.2575,
1706
+ "step": 283
1707
+ },
1708
+ {
1709
+ "epoch": 1.14,
1710
+ "learning_rate": 1.2929848693259975e-05,
1711
+ "loss": 0.3409,
1712
+ "step": 284
1713
+ },
1714
+ {
1715
+ "epoch": 1.14,
1716
+ "learning_rate": 1.2902338376891334e-05,
1717
+ "loss": 0.2825,
1718
+ "step": 285
1719
+ },
1720
+ {
1721
+ "epoch": 1.14,
1722
+ "learning_rate": 1.2874828060522697e-05,
1723
+ "loss": 0.3299,
1724
+ "step": 286
1725
+ },
1726
+ {
1727
+ "epoch": 1.15,
1728
+ "learning_rate": 1.284731774415406e-05,
1729
+ "loss": 0.2843,
1730
+ "step": 287
1731
+ },
1732
+ {
1733
+ "epoch": 1.15,
1734
+ "learning_rate": 1.281980742778542e-05,
1735
+ "loss": 0.3134,
1736
+ "step": 288
1737
+ },
1738
+ {
1739
+ "epoch": 1.16,
1740
+ "learning_rate": 1.2792297111416782e-05,
1741
+ "loss": 0.276,
1742
+ "step": 289
1743
+ },
1744
+ {
1745
+ "epoch": 1.16,
1746
+ "learning_rate": 1.2764786795048143e-05,
1747
+ "loss": 0.3417,
1748
+ "step": 290
1749
+ },
1750
+ {
1751
+ "epoch": 1.16,
1752
+ "learning_rate": 1.2737276478679506e-05,
1753
+ "loss": 0.2623,
1754
+ "step": 291
1755
+ },
1756
+ {
1757
+ "epoch": 1.17,
1758
+ "learning_rate": 1.2709766162310868e-05,
1759
+ "loss": 0.2778,
1760
+ "step": 292
1761
+ },
1762
+ {
1763
+ "epoch": 1.17,
1764
+ "learning_rate": 1.2682255845942229e-05,
1765
+ "loss": 0.3325,
1766
+ "step": 293
1767
+ },
1768
+ {
1769
+ "epoch": 1.18,
1770
+ "learning_rate": 1.2654745529573592e-05,
1771
+ "loss": 0.3026,
1772
+ "step": 294
1773
+ },
1774
+ {
1775
+ "epoch": 1.18,
1776
+ "learning_rate": 1.2627235213204953e-05,
1777
+ "loss": 0.3896,
1778
+ "step": 295
1779
+ },
1780
+ {
1781
+ "epoch": 1.18,
1782
+ "learning_rate": 1.2599724896836314e-05,
1783
+ "loss": 0.3385,
1784
+ "step": 296
1785
+ },
1786
+ {
1787
+ "epoch": 1.19,
1788
+ "learning_rate": 1.2572214580467677e-05,
1789
+ "loss": 0.3296,
1790
+ "step": 297
1791
+ },
1792
+ {
1793
+ "epoch": 1.19,
1794
+ "learning_rate": 1.254470426409904e-05,
1795
+ "loss": 0.2828,
1796
+ "step": 298
1797
+ },
1798
+ {
1799
+ "epoch": 1.2,
1800
+ "learning_rate": 1.25171939477304e-05,
1801
+ "loss": 0.3029,
1802
+ "step": 299
1803
+ },
1804
+ {
1805
+ "epoch": 1.2,
1806
+ "learning_rate": 1.2489683631361762e-05,
1807
+ "loss": 0.3288,
1808
+ "step": 300
1809
+ },
1810
+ {
1811
+ "epoch": 1.2,
1812
+ "learning_rate": 1.2462173314993125e-05,
1813
+ "loss": 0.2971,
1814
+ "step": 301
1815
+ },
1816
+ {
1817
+ "epoch": 1.21,
1818
+ "learning_rate": 1.2434662998624485e-05,
1819
+ "loss": 0.2642,
1820
+ "step": 302
1821
+ },
1822
+ {
1823
+ "epoch": 1.21,
1824
+ "learning_rate": 1.2407152682255848e-05,
1825
+ "loss": 0.2711,
1826
+ "step": 303
1827
+ },
1828
+ {
1829
+ "epoch": 1.22,
1830
+ "learning_rate": 1.2379642365887207e-05,
1831
+ "loss": 0.3037,
1832
+ "step": 304
1833
+ },
1834
+ {
1835
+ "epoch": 1.22,
1836
+ "learning_rate": 1.235213204951857e-05,
1837
+ "loss": 0.2917,
1838
+ "step": 305
1839
+ },
1840
+ {
1841
+ "epoch": 1.22,
1842
+ "learning_rate": 1.2324621733149933e-05,
1843
+ "loss": 0.32,
1844
+ "step": 306
1845
+ },
1846
+ {
1847
+ "epoch": 1.23,
1848
+ "learning_rate": 1.2297111416781292e-05,
1849
+ "loss": 0.2759,
1850
+ "step": 307
1851
+ },
1852
+ {
1853
+ "epoch": 1.23,
1854
+ "learning_rate": 1.2269601100412655e-05,
1855
+ "loss": 0.3036,
1856
+ "step": 308
1857
+ },
1858
+ {
1859
+ "epoch": 1.24,
1860
+ "learning_rate": 1.2242090784044018e-05,
1861
+ "loss": 0.3284,
1862
+ "step": 309
1863
+ },
1864
+ {
1865
+ "epoch": 1.24,
1866
+ "learning_rate": 1.221458046767538e-05,
1867
+ "loss": 0.2982,
1868
+ "step": 310
1869
+ },
1870
+ {
1871
+ "epoch": 1.24,
1872
+ "learning_rate": 1.218707015130674e-05,
1873
+ "loss": 0.3157,
1874
+ "step": 311
1875
+ },
1876
+ {
1877
+ "epoch": 1.25,
1878
+ "learning_rate": 1.2159559834938104e-05,
1879
+ "loss": 0.316,
1880
+ "step": 312
1881
+ },
1882
+ {
1883
+ "epoch": 1.25,
1884
+ "learning_rate": 1.2132049518569465e-05,
1885
+ "loss": 0.3031,
1886
+ "step": 313
1887
+ },
1888
+ {
1889
+ "epoch": 1.26,
1890
+ "learning_rate": 1.2104539202200826e-05,
1891
+ "loss": 0.3152,
1892
+ "step": 314
1893
+ },
1894
+ {
1895
+ "epoch": 1.26,
1896
+ "learning_rate": 1.2077028885832187e-05,
1897
+ "loss": 0.2929,
1898
+ "step": 315
1899
+ },
1900
+ {
1901
+ "epoch": 1.26,
1902
+ "learning_rate": 1.204951856946355e-05,
1903
+ "loss": 0.2616,
1904
+ "step": 316
1905
+ },
1906
+ {
1907
+ "epoch": 1.27,
1908
+ "learning_rate": 1.2022008253094913e-05,
1909
+ "loss": 0.3035,
1910
+ "step": 317
1911
+ },
1912
+ {
1913
+ "epoch": 1.27,
1914
+ "learning_rate": 1.1994497936726273e-05,
1915
+ "loss": 0.2955,
1916
+ "step": 318
1917
+ },
1918
+ {
1919
+ "epoch": 1.28,
1920
+ "learning_rate": 1.1966987620357635e-05,
1921
+ "loss": 0.2996,
1922
+ "step": 319
1923
+ },
1924
+ {
1925
+ "epoch": 1.28,
1926
+ "learning_rate": 1.1939477303988998e-05,
1927
+ "loss": 0.3157,
1928
+ "step": 320
1929
+ },
1930
+ {
1931
+ "epoch": 1.28,
1932
+ "learning_rate": 1.1911966987620358e-05,
1933
+ "loss": 0.3141,
1934
+ "step": 321
1935
+ },
1936
+ {
1937
+ "epoch": 1.29,
1938
+ "learning_rate": 1.188445667125172e-05,
1939
+ "loss": 0.3236,
1940
+ "step": 322
1941
+ },
1942
+ {
1943
+ "epoch": 1.29,
1944
+ "learning_rate": 1.1856946354883084e-05,
1945
+ "loss": 0.3105,
1946
+ "step": 323
1947
+ },
1948
+ {
1949
+ "epoch": 1.3,
1950
+ "learning_rate": 1.1829436038514443e-05,
1951
+ "loss": 0.3012,
1952
+ "step": 324
1953
+ },
1954
+ {
1955
+ "epoch": 1.3,
1956
+ "learning_rate": 1.1801925722145806e-05,
1957
+ "loss": 0.2917,
1958
+ "step": 325
1959
+ },
1960
+ {
1961
+ "epoch": 1.3,
1962
+ "learning_rate": 1.1774415405777167e-05,
1963
+ "loss": 0.3089,
1964
+ "step": 326
1965
+ },
1966
+ {
1967
+ "epoch": 1.31,
1968
+ "learning_rate": 1.1746905089408529e-05,
1969
+ "loss": 0.3057,
1970
+ "step": 327
1971
+ },
1972
+ {
1973
+ "epoch": 1.31,
1974
+ "learning_rate": 1.1719394773039891e-05,
1975
+ "loss": 0.314,
1976
+ "step": 328
1977
+ },
1978
+ {
1979
+ "epoch": 1.32,
1980
+ "learning_rate": 1.1691884456671253e-05,
1981
+ "loss": 0.3231,
1982
+ "step": 329
1983
+ },
1984
+ {
1985
+ "epoch": 1.32,
1986
+ "learning_rate": 1.1664374140302614e-05,
1987
+ "loss": 0.3227,
1988
+ "step": 330
1989
+ },
1990
+ {
1991
+ "epoch": 1.32,
1992
+ "learning_rate": 1.1636863823933977e-05,
1993
+ "loss": 0.2869,
1994
+ "step": 331
1995
+ },
1996
+ {
1997
+ "epoch": 1.33,
1998
+ "learning_rate": 1.1609353507565338e-05,
1999
+ "loss": 0.3323,
2000
+ "step": 332
2001
+ },
2002
+ {
2003
+ "epoch": 1.33,
2004
+ "learning_rate": 1.1581843191196699e-05,
2005
+ "loss": 0.264,
2006
+ "step": 333
2007
+ },
2008
+ {
2009
+ "epoch": 1.34,
2010
+ "learning_rate": 1.1554332874828062e-05,
2011
+ "loss": 0.2834,
2012
+ "step": 334
2013
+ },
2014
+ {
2015
+ "epoch": 1.34,
2016
+ "learning_rate": 1.1526822558459423e-05,
2017
+ "loss": 0.3327,
2018
+ "step": 335
2019
+ },
2020
+ {
2021
+ "epoch": 1.34,
2022
+ "learning_rate": 1.1499312242090786e-05,
2023
+ "loss": 0.3061,
2024
+ "step": 336
2025
+ },
2026
+ {
2027
+ "epoch": 1.35,
2028
+ "learning_rate": 1.1471801925722146e-05,
2029
+ "loss": 0.29,
2030
+ "step": 337
2031
+ },
2032
+ {
2033
+ "epoch": 1.35,
2034
+ "learning_rate": 1.1444291609353509e-05,
2035
+ "loss": 0.3232,
2036
+ "step": 338
2037
+ },
2038
+ {
2039
+ "epoch": 1.36,
2040
+ "learning_rate": 1.1416781292984871e-05,
2041
+ "loss": 0.3349,
2042
+ "step": 339
2043
+ },
2044
+ {
2045
+ "epoch": 1.36,
2046
+ "learning_rate": 1.1389270976616231e-05,
2047
+ "loss": 0.2967,
2048
+ "step": 340
2049
+ },
2050
+ {
2051
+ "epoch": 1.36,
2052
+ "learning_rate": 1.1361760660247594e-05,
2053
+ "loss": 0.322,
2054
+ "step": 341
2055
+ },
2056
+ {
2057
+ "epoch": 1.37,
2058
+ "learning_rate": 1.1334250343878957e-05,
2059
+ "loss": 0.3212,
2060
+ "step": 342
2061
+ },
2062
+ {
2063
+ "epoch": 1.37,
2064
+ "learning_rate": 1.1306740027510316e-05,
2065
+ "loss": 0.3236,
2066
+ "step": 343
2067
+ },
2068
+ {
2069
+ "epoch": 1.38,
2070
+ "learning_rate": 1.127922971114168e-05,
2071
+ "loss": 0.3043,
2072
+ "step": 344
2073
+ },
2074
+ {
2075
+ "epoch": 1.38,
2076
+ "learning_rate": 1.1251719394773042e-05,
2077
+ "loss": 0.3347,
2078
+ "step": 345
2079
+ },
2080
+ {
2081
+ "epoch": 1.38,
2082
+ "learning_rate": 1.1224209078404402e-05,
2083
+ "loss": 0.335,
2084
+ "step": 346
2085
+ },
2086
+ {
2087
+ "epoch": 1.39,
2088
+ "learning_rate": 1.1196698762035765e-05,
2089
+ "loss": 0.3274,
2090
+ "step": 347
2091
+ },
2092
+ {
2093
+ "epoch": 1.39,
2094
+ "learning_rate": 1.1169188445667126e-05,
2095
+ "loss": 0.2917,
2096
+ "step": 348
2097
+ },
2098
+ {
2099
+ "epoch": 1.4,
2100
+ "learning_rate": 1.1141678129298487e-05,
2101
+ "loss": 0.3259,
2102
+ "step": 349
2103
+ },
2104
+ {
2105
+ "epoch": 1.4,
2106
+ "learning_rate": 1.111416781292985e-05,
2107
+ "loss": 0.2635,
2108
+ "step": 350
2109
+ },
2110
+ {
2111
+ "epoch": 1.4,
2112
+ "learning_rate": 1.1086657496561211e-05,
2113
+ "loss": 0.2856,
2114
+ "step": 351
2115
+ },
2116
+ {
2117
+ "epoch": 1.41,
2118
+ "learning_rate": 1.1059147180192574e-05,
2119
+ "loss": 0.329,
2120
+ "step": 352
2121
+ },
2122
+ {
2123
+ "epoch": 1.41,
2124
+ "learning_rate": 1.1031636863823935e-05,
2125
+ "loss": 0.31,
2126
+ "step": 353
2127
+ },
2128
+ {
2129
+ "epoch": 1.42,
2130
+ "learning_rate": 1.1004126547455296e-05,
2131
+ "loss": 0.2703,
2132
+ "step": 354
2133
+ },
2134
+ {
2135
+ "epoch": 1.42,
2136
+ "learning_rate": 1.097661623108666e-05,
2137
+ "loss": 0.3278,
2138
+ "step": 355
2139
+ },
2140
+ {
2141
+ "epoch": 1.42,
2142
+ "learning_rate": 1.094910591471802e-05,
2143
+ "loss": 0.3352,
2144
+ "step": 356
2145
+ },
2146
+ {
2147
+ "epoch": 1.43,
2148
+ "learning_rate": 1.0921595598349382e-05,
2149
+ "loss": 0.2798,
2150
+ "step": 357
2151
+ },
2152
+ {
2153
+ "epoch": 1.43,
2154
+ "learning_rate": 1.0894085281980745e-05,
2155
+ "loss": 0.3067,
2156
+ "step": 358
2157
+ },
2158
+ {
2159
+ "epoch": 1.44,
2160
+ "learning_rate": 1.0866574965612104e-05,
2161
+ "loss": 0.2838,
2162
+ "step": 359
2163
+ },
2164
+ {
2165
+ "epoch": 1.44,
2166
+ "learning_rate": 1.0839064649243467e-05,
2167
+ "loss": 0.3022,
2168
+ "step": 360
2169
+ },
2170
+ {
2171
+ "epoch": 1.44,
2172
+ "learning_rate": 1.081155433287483e-05,
2173
+ "loss": 0.3093,
2174
+ "step": 361
2175
+ },
2176
+ {
2177
+ "epoch": 1.45,
2178
+ "learning_rate": 1.078404401650619e-05,
2179
+ "loss": 0.3349,
2180
+ "step": 362
2181
+ },
2182
+ {
2183
+ "epoch": 1.45,
2184
+ "learning_rate": 1.0756533700137552e-05,
2185
+ "loss": 0.307,
2186
+ "step": 363
2187
+ },
2188
+ {
2189
+ "epoch": 1.46,
2190
+ "learning_rate": 1.0729023383768915e-05,
2191
+ "loss": 0.3127,
2192
+ "step": 364
2193
+ },
2194
+ {
2195
+ "epoch": 1.46,
2196
+ "learning_rate": 1.0701513067400275e-05,
2197
+ "loss": 0.2595,
2198
+ "step": 365
2199
+ },
2200
+ {
2201
+ "epoch": 1.46,
2202
+ "learning_rate": 1.0674002751031638e-05,
2203
+ "loss": 0.2923,
2204
+ "step": 366
2205
+ },
2206
+ {
2207
+ "epoch": 1.47,
2208
+ "learning_rate": 1.0646492434663e-05,
2209
+ "loss": 0.2882,
2210
+ "step": 367
2211
+ },
2212
+ {
2213
+ "epoch": 1.47,
2214
+ "learning_rate": 1.061898211829436e-05,
2215
+ "loss": 0.2729,
2216
+ "step": 368
2217
+ },
2218
+ {
2219
+ "epoch": 1.48,
2220
+ "learning_rate": 1.0591471801925723e-05,
2221
+ "loss": 0.3035,
2222
+ "step": 369
2223
+ },
2224
+ {
2225
+ "epoch": 1.48,
2226
+ "learning_rate": 1.0563961485557084e-05,
2227
+ "loss": 0.3469,
2228
+ "step": 370
2229
+ },
2230
+ {
2231
+ "epoch": 1.48,
2232
+ "learning_rate": 1.0536451169188447e-05,
2233
+ "loss": 0.329,
2234
+ "step": 371
2235
+ },
2236
+ {
2237
+ "epoch": 1.49,
2238
+ "learning_rate": 1.0508940852819808e-05,
2239
+ "loss": 0.3302,
2240
+ "step": 372
2241
+ },
2242
+ {
2243
+ "epoch": 1.49,
2244
+ "learning_rate": 1.048143053645117e-05,
2245
+ "loss": 0.3358,
2246
+ "step": 373
2247
+ },
2248
+ {
2249
+ "epoch": 1.5,
2250
+ "learning_rate": 1.0453920220082532e-05,
2251
+ "loss": 0.3315,
2252
+ "step": 374
2253
+ },
2254
+ {
2255
+ "epoch": 1.5,
2256
+ "learning_rate": 1.0426409903713894e-05,
2257
+ "loss": 0.2801,
2258
+ "step": 375
2259
+ },
2260
+ {
2261
+ "epoch": 1.5,
2262
+ "learning_rate": 1.0398899587345255e-05,
2263
+ "loss": 0.2818,
2264
+ "step": 376
2265
+ },
2266
+ {
2267
+ "epoch": 1.51,
2268
+ "learning_rate": 1.0371389270976618e-05,
2269
+ "loss": 0.2859,
2270
+ "step": 377
2271
+ },
2272
+ {
2273
+ "epoch": 1.51,
2274
+ "learning_rate": 1.034387895460798e-05,
2275
+ "loss": 0.355,
2276
+ "step": 378
2277
+ },
2278
+ {
2279
+ "epoch": 1.52,
2280
+ "learning_rate": 1.031636863823934e-05,
2281
+ "loss": 0.2643,
2282
+ "step": 379
2283
+ },
2284
+ {
2285
+ "epoch": 1.52,
2286
+ "learning_rate": 1.0288858321870703e-05,
2287
+ "loss": 0.257,
2288
+ "step": 380
2289
+ },
2290
+ {
2291
+ "epoch": 1.52,
2292
+ "learning_rate": 1.0261348005502063e-05,
2293
+ "loss": 0.354,
2294
+ "step": 381
2295
+ },
2296
+ {
2297
+ "epoch": 1.53,
2298
+ "learning_rate": 1.0233837689133425e-05,
2299
+ "loss": 0.3119,
2300
+ "step": 382
2301
+ },
2302
+ {
2303
+ "epoch": 1.53,
2304
+ "learning_rate": 1.0206327372764788e-05,
2305
+ "loss": 0.3228,
2306
+ "step": 383
2307
+ },
2308
+ {
2309
+ "epoch": 1.54,
2310
+ "learning_rate": 1.0178817056396148e-05,
2311
+ "loss": 0.3063,
2312
+ "step": 384
2313
+ },
2314
+ {
2315
+ "epoch": 1.54,
2316
+ "learning_rate": 1.015130674002751e-05,
2317
+ "loss": 0.3218,
2318
+ "step": 385
2319
+ },
2320
+ {
2321
+ "epoch": 1.54,
2322
+ "learning_rate": 1.0123796423658874e-05,
2323
+ "loss": 0.3326,
2324
+ "step": 386
2325
+ },
2326
+ {
2327
+ "epoch": 1.55,
2328
+ "learning_rate": 1.0096286107290235e-05,
2329
+ "loss": 0.3411,
2330
+ "step": 387
2331
+ },
2332
+ {
2333
+ "epoch": 1.55,
2334
+ "learning_rate": 1.0068775790921596e-05,
2335
+ "loss": 0.3243,
2336
+ "step": 388
2337
+ },
2338
+ {
2339
+ "epoch": 1.56,
2340
+ "learning_rate": 1.0041265474552959e-05,
2341
+ "loss": 0.349,
2342
+ "step": 389
2343
+ },
2344
+ {
2345
+ "epoch": 1.56,
2346
+ "learning_rate": 1.001375515818432e-05,
2347
+ "loss": 0.2975,
2348
+ "step": 390
2349
+ },
2350
+ {
2351
+ "epoch": 1.56,
2352
+ "learning_rate": 9.986244841815681e-06,
2353
+ "loss": 0.3078,
2354
+ "step": 391
2355
+ },
2356
+ {
2357
+ "epoch": 1.57,
2358
+ "learning_rate": 9.958734525447044e-06,
2359
+ "loss": 0.3176,
2360
+ "step": 392
2361
+ },
2362
+ {
2363
+ "epoch": 1.57,
2364
+ "learning_rate": 9.931224209078406e-06,
2365
+ "loss": 0.2235,
2366
+ "step": 393
2367
+ },
2368
+ {
2369
+ "epoch": 1.58,
2370
+ "learning_rate": 9.903713892709767e-06,
2371
+ "loss": 0.3255,
2372
+ "step": 394
2373
+ },
2374
+ {
2375
+ "epoch": 1.58,
2376
+ "learning_rate": 9.87620357634113e-06,
2377
+ "loss": 0.3365,
2378
+ "step": 395
2379
+ },
2380
+ {
2381
+ "epoch": 1.58,
2382
+ "learning_rate": 9.848693259972491e-06,
2383
+ "loss": 0.2892,
2384
+ "step": 396
2385
+ },
2386
+ {
2387
+ "epoch": 1.59,
2388
+ "learning_rate": 9.821182943603852e-06,
2389
+ "loss": 0.3018,
2390
+ "step": 397
2391
+ },
2392
+ {
2393
+ "epoch": 1.59,
2394
+ "learning_rate": 9.793672627235215e-06,
2395
+ "loss": 0.2908,
2396
+ "step": 398
2397
+ },
2398
+ {
2399
+ "epoch": 1.6,
2400
+ "learning_rate": 9.766162310866576e-06,
2401
+ "loss": 0.2969,
2402
+ "step": 399
2403
+ },
2404
+ {
2405
+ "epoch": 1.6,
2406
+ "learning_rate": 9.738651994497937e-06,
2407
+ "loss": 0.24,
2408
+ "step": 400
2409
+ },
2410
+ {
2411
+ "epoch": 1.6,
2412
+ "learning_rate": 9.711141678129299e-06,
2413
+ "loss": 0.2877,
2414
+ "step": 401
2415
+ },
2416
+ {
2417
+ "epoch": 1.61,
2418
+ "learning_rate": 9.683631361760661e-06,
2419
+ "loss": 0.2665,
2420
+ "step": 402
2421
+ },
2422
+ {
2423
+ "epoch": 1.61,
2424
+ "learning_rate": 9.656121045392023e-06,
2425
+ "loss": 0.3016,
2426
+ "step": 403
2427
+ },
2428
+ {
2429
+ "epoch": 1.62,
2430
+ "learning_rate": 9.628610729023384e-06,
2431
+ "loss": 0.3647,
2432
+ "step": 404
2433
+ },
2434
+ {
2435
+ "epoch": 1.62,
2436
+ "learning_rate": 9.601100412654745e-06,
2437
+ "loss": 0.2788,
2438
+ "step": 405
2439
+ },
2440
+ {
2441
+ "epoch": 1.62,
2442
+ "learning_rate": 9.573590096286108e-06,
2443
+ "loss": 0.2821,
2444
+ "step": 406
2445
+ },
2446
+ {
2447
+ "epoch": 1.63,
2448
+ "learning_rate": 9.54607977991747e-06,
2449
+ "loss": 0.2481,
2450
+ "step": 407
2451
+ },
2452
+ {
2453
+ "epoch": 1.63,
2454
+ "learning_rate": 9.51856946354883e-06,
2455
+ "loss": 0.3071,
2456
+ "step": 408
2457
+ },
2458
+ {
2459
+ "epoch": 1.64,
2460
+ "learning_rate": 9.491059147180193e-06,
2461
+ "loss": 0.2887,
2462
+ "step": 409
2463
+ },
2464
+ {
2465
+ "epoch": 1.64,
2466
+ "learning_rate": 9.463548830811555e-06,
2467
+ "loss": 0.3086,
2468
+ "step": 410
2469
+ },
2470
+ {
2471
+ "epoch": 1.64,
2472
+ "learning_rate": 9.436038514442917e-06,
2473
+ "loss": 0.3034,
2474
+ "step": 411
2475
+ },
2476
+ {
2477
+ "epoch": 1.65,
2478
+ "learning_rate": 9.408528198074279e-06,
2479
+ "loss": 0.3142,
2480
+ "step": 412
2481
+ },
2482
+ {
2483
+ "epoch": 1.65,
2484
+ "learning_rate": 9.381017881705642e-06,
2485
+ "loss": 0.2837,
2486
+ "step": 413
2487
+ },
2488
+ {
2489
+ "epoch": 1.66,
2490
+ "learning_rate": 9.353507565337003e-06,
2491
+ "loss": 0.2937,
2492
+ "step": 414
2493
+ },
2494
+ {
2495
+ "epoch": 1.66,
2496
+ "learning_rate": 9.325997248968364e-06,
2497
+ "loss": 0.3049,
2498
+ "step": 415
2499
+ },
2500
+ {
2501
+ "epoch": 1.66,
2502
+ "learning_rate": 9.298486932599725e-06,
2503
+ "loss": 0.2905,
2504
+ "step": 416
2505
+ },
2506
+ {
2507
+ "epoch": 1.67,
2508
+ "learning_rate": 9.270976616231088e-06,
2509
+ "loss": 0.3221,
2510
+ "step": 417
2511
+ },
2512
+ {
2513
+ "epoch": 1.67,
2514
+ "learning_rate": 9.24346629986245e-06,
2515
+ "loss": 0.3128,
2516
+ "step": 418
2517
+ },
2518
+ {
2519
+ "epoch": 1.68,
2520
+ "learning_rate": 9.21595598349381e-06,
2521
+ "loss": 0.318,
2522
+ "step": 419
2523
+ },
2524
+ {
2525
+ "epoch": 1.68,
2526
+ "learning_rate": 9.188445667125173e-06,
2527
+ "loss": 0.2852,
2528
+ "step": 420
2529
+ },
2530
+ {
2531
+ "epoch": 1.68,
2532
+ "learning_rate": 9.160935350756535e-06,
2533
+ "loss": 0.2334,
2534
+ "step": 421
2535
+ },
2536
+ {
2537
+ "epoch": 1.69,
2538
+ "learning_rate": 9.133425034387896e-06,
2539
+ "loss": 0.3457,
2540
+ "step": 422
2541
+ },
2542
+ {
2543
+ "epoch": 1.69,
2544
+ "learning_rate": 9.105914718019257e-06,
2545
+ "loss": 0.2861,
2546
+ "step": 423
2547
+ },
2548
+ {
2549
+ "epoch": 1.7,
2550
+ "learning_rate": 9.07840440165062e-06,
2551
+ "loss": 0.2817,
2552
+ "step": 424
2553
+ },
2554
+ {
2555
+ "epoch": 1.7,
2556
+ "learning_rate": 9.050894085281981e-06,
2557
+ "loss": 0.3062,
2558
+ "step": 425
2559
+ },
2560
+ {
2561
+ "epoch": 1.7,
2562
+ "learning_rate": 9.023383768913342e-06,
2563
+ "loss": 0.3121,
2564
+ "step": 426
2565
+ },
2566
+ {
2567
+ "epoch": 1.71,
2568
+ "learning_rate": 8.995873452544705e-06,
2569
+ "loss": 0.3047,
2570
+ "step": 427
2571
+ },
2572
+ {
2573
+ "epoch": 1.71,
2574
+ "learning_rate": 8.968363136176066e-06,
2575
+ "loss": 0.3511,
2576
+ "step": 428
2577
+ },
2578
+ {
2579
+ "epoch": 1.72,
2580
+ "learning_rate": 8.940852819807428e-06,
2581
+ "loss": 0.284,
2582
+ "step": 429
2583
+ },
2584
+ {
2585
+ "epoch": 1.72,
2586
+ "learning_rate": 8.91334250343879e-06,
2587
+ "loss": 0.2439,
2588
+ "step": 430
2589
+ },
2590
+ {
2591
+ "epoch": 1.72,
2592
+ "learning_rate": 8.885832187070152e-06,
2593
+ "loss": 0.2665,
2594
+ "step": 431
2595
+ },
2596
+ {
2597
+ "epoch": 1.73,
2598
+ "learning_rate": 8.858321870701515e-06,
2599
+ "loss": 0.3014,
2600
+ "step": 432
2601
+ },
2602
+ {
2603
+ "epoch": 1.73,
2604
+ "learning_rate": 8.830811554332876e-06,
2605
+ "loss": 0.2791,
2606
+ "step": 433
2607
+ },
2608
+ {
2609
+ "epoch": 1.74,
2610
+ "learning_rate": 8.803301237964237e-06,
2611
+ "loss": 0.3252,
2612
+ "step": 434
2613
+ },
2614
+ {
2615
+ "epoch": 1.74,
2616
+ "learning_rate": 8.7757909215956e-06,
2617
+ "loss": 0.2917,
2618
+ "step": 435
2619
+ },
2620
+ {
2621
+ "epoch": 1.74,
2622
+ "learning_rate": 8.748280605226961e-06,
2623
+ "loss": 0.2771,
2624
+ "step": 436
2625
+ },
2626
+ {
2627
+ "epoch": 1.75,
2628
+ "learning_rate": 8.720770288858322e-06,
2629
+ "loss": 0.2894,
2630
+ "step": 437
2631
+ },
2632
+ {
2633
+ "epoch": 1.75,
2634
+ "learning_rate": 8.693259972489685e-06,
2635
+ "loss": 0.315,
2636
+ "step": 438
2637
+ },
2638
+ {
2639
+ "epoch": 1.76,
2640
+ "learning_rate": 8.665749656121047e-06,
2641
+ "loss": 0.3168,
2642
+ "step": 439
2643
+ },
2644
+ {
2645
+ "epoch": 1.76,
2646
+ "learning_rate": 8.638239339752408e-06,
2647
+ "loss": 0.2843,
2648
+ "step": 440
2649
+ },
2650
+ {
2651
+ "epoch": 1.76,
2652
+ "learning_rate": 8.610729023383769e-06,
2653
+ "loss": 0.2928,
2654
+ "step": 441
2655
+ },
2656
+ {
2657
+ "epoch": 1.77,
2658
+ "learning_rate": 8.583218707015132e-06,
2659
+ "loss": 0.2945,
2660
+ "step": 442
2661
+ },
2662
+ {
2663
+ "epoch": 1.77,
2664
+ "learning_rate": 8.555708390646493e-06,
2665
+ "loss": 0.3129,
2666
+ "step": 443
2667
+ },
2668
+ {
2669
+ "epoch": 1.78,
2670
+ "learning_rate": 8.528198074277854e-06,
2671
+ "loss": 0.3279,
2672
+ "step": 444
2673
+ },
2674
+ {
2675
+ "epoch": 1.78,
2676
+ "learning_rate": 8.500687757909215e-06,
2677
+ "loss": 0.3032,
2678
+ "step": 445
2679
+ },
2680
+ {
2681
+ "epoch": 1.78,
2682
+ "learning_rate": 8.473177441540578e-06,
2683
+ "loss": 0.2923,
2684
+ "step": 446
2685
+ },
2686
+ {
2687
+ "epoch": 1.79,
2688
+ "learning_rate": 8.44566712517194e-06,
2689
+ "loss": 0.2889,
2690
+ "step": 447
2691
+ },
2692
+ {
2693
+ "epoch": 1.79,
2694
+ "learning_rate": 8.418156808803302e-06,
2695
+ "loss": 0.2625,
2696
+ "step": 448
2697
+ },
2698
+ {
2699
+ "epoch": 1.8,
2700
+ "learning_rate": 8.390646492434664e-06,
2701
+ "loss": 0.2831,
2702
+ "step": 449
2703
+ },
2704
+ {
2705
+ "epoch": 1.8,
2706
+ "learning_rate": 8.363136176066025e-06,
2707
+ "loss": 0.3488,
2708
+ "step": 450
2709
+ },
2710
+ {
2711
+ "epoch": 1.8,
2712
+ "learning_rate": 8.335625859697388e-06,
2713
+ "loss": 0.325,
2714
+ "step": 451
2715
+ },
2716
+ {
2717
+ "epoch": 1.81,
2718
+ "learning_rate": 8.308115543328749e-06,
2719
+ "loss": 0.2904,
2720
+ "step": 452
2721
+ },
2722
+ {
2723
+ "epoch": 1.81,
2724
+ "learning_rate": 8.280605226960112e-06,
2725
+ "loss": 0.315,
2726
+ "step": 453
2727
+ },
2728
+ {
2729
+ "epoch": 1.82,
2730
+ "learning_rate": 8.253094910591473e-06,
2731
+ "loss": 0.3033,
2732
+ "step": 454
2733
+ },
2734
+ {
2735
+ "epoch": 1.82,
2736
+ "learning_rate": 8.225584594222834e-06,
2737
+ "loss": 0.3404,
2738
+ "step": 455
2739
+ },
2740
+ {
2741
+ "epoch": 1.82,
2742
+ "learning_rate": 8.198074277854196e-06,
2743
+ "loss": 0.2838,
2744
+ "step": 456
2745
+ },
2746
+ {
2747
+ "epoch": 1.83,
2748
+ "learning_rate": 8.170563961485558e-06,
2749
+ "loss": 0.2823,
2750
+ "step": 457
2751
+ },
2752
+ {
2753
+ "epoch": 1.83,
2754
+ "learning_rate": 8.14305364511692e-06,
2755
+ "loss": 0.306,
2756
+ "step": 458
2757
+ },
2758
+ {
2759
+ "epoch": 1.84,
2760
+ "learning_rate": 8.115543328748281e-06,
2761
+ "loss": 0.3263,
2762
+ "step": 459
2763
+ },
2764
+ {
2765
+ "epoch": 1.84,
2766
+ "learning_rate": 8.088033012379644e-06,
2767
+ "loss": 0.3078,
2768
+ "step": 460
2769
+ },
2770
+ {
2771
+ "epoch": 1.84,
2772
+ "learning_rate": 8.060522696011005e-06,
2773
+ "loss": 0.317,
2774
+ "step": 461
2775
+ },
2776
+ {
2777
+ "epoch": 1.85,
2778
+ "learning_rate": 8.033012379642366e-06,
2779
+ "loss": 0.3439,
2780
+ "step": 462
2781
+ },
2782
+ {
2783
+ "epoch": 1.85,
2784
+ "learning_rate": 8.005502063273727e-06,
2785
+ "loss": 0.3117,
2786
+ "step": 463
2787
+ },
2788
+ {
2789
+ "epoch": 1.86,
2790
+ "learning_rate": 7.97799174690509e-06,
2791
+ "loss": 0.316,
2792
+ "step": 464
2793
+ },
2794
+ {
2795
+ "epoch": 1.86,
2796
+ "learning_rate": 7.950481430536452e-06,
2797
+ "loss": 0.328,
2798
+ "step": 465
2799
+ },
2800
+ {
2801
+ "epoch": 1.86,
2802
+ "learning_rate": 7.922971114167813e-06,
2803
+ "loss": 0.2783,
2804
+ "step": 466
2805
+ },
2806
+ {
2807
+ "epoch": 1.87,
2808
+ "learning_rate": 7.895460797799176e-06,
2809
+ "loss": 0.3354,
2810
+ "step": 467
2811
+ },
2812
+ {
2813
+ "epoch": 1.87,
2814
+ "learning_rate": 7.867950481430537e-06,
2815
+ "loss": 0.3286,
2816
+ "step": 468
2817
+ },
2818
+ {
2819
+ "epoch": 1.88,
2820
+ "learning_rate": 7.840440165061898e-06,
2821
+ "loss": 0.349,
2822
+ "step": 469
2823
+ },
2824
+ {
2825
+ "epoch": 1.88,
2826
+ "learning_rate": 7.812929848693261e-06,
2827
+ "loss": 0.3494,
2828
+ "step": 470
2829
+ },
2830
+ {
2831
+ "epoch": 1.88,
2832
+ "learning_rate": 7.785419532324622e-06,
2833
+ "loss": 0.2758,
2834
+ "step": 471
2835
+ },
2836
+ {
2837
+ "epoch": 1.89,
2838
+ "learning_rate": 7.757909215955985e-06,
2839
+ "loss": 0.3115,
2840
+ "step": 472
2841
+ },
2842
+ {
2843
+ "epoch": 1.89,
2844
+ "learning_rate": 7.730398899587346e-06,
2845
+ "loss": 0.2571,
2846
+ "step": 473
2847
+ },
2848
+ {
2849
+ "epoch": 1.9,
2850
+ "learning_rate": 7.702888583218707e-06,
2851
+ "loss": 0.3107,
2852
+ "step": 474
2853
+ },
2854
+ {
2855
+ "epoch": 1.9,
2856
+ "learning_rate": 7.67537826685007e-06,
2857
+ "loss": 0.2853,
2858
+ "step": 475
2859
+ },
2860
+ {
2861
+ "epoch": 1.9,
2862
+ "learning_rate": 7.647867950481432e-06,
2863
+ "loss": 0.3056,
2864
+ "step": 476
2865
+ },
2866
+ {
2867
+ "epoch": 1.91,
2868
+ "learning_rate": 7.620357634112793e-06,
2869
+ "loss": 0.3009,
2870
+ "step": 477
2871
+ },
2872
+ {
2873
+ "epoch": 1.91,
2874
+ "learning_rate": 7.592847317744154e-06,
2875
+ "loss": 0.298,
2876
+ "step": 478
2877
+ },
2878
+ {
2879
+ "epoch": 1.92,
2880
+ "learning_rate": 7.565337001375517e-06,
2881
+ "loss": 0.3062,
2882
+ "step": 479
2883
+ },
2884
+ {
2885
+ "epoch": 1.92,
2886
+ "learning_rate": 7.537826685006878e-06,
2887
+ "loss": 0.3205,
2888
+ "step": 480
2889
+ },
2890
+ {
2891
+ "epoch": 1.92,
2892
+ "learning_rate": 7.510316368638239e-06,
2893
+ "loss": 0.2877,
2894
+ "step": 481
2895
+ },
2896
+ {
2897
+ "epoch": 1.93,
2898
+ "learning_rate": 7.482806052269602e-06,
2899
+ "loss": 0.2873,
2900
+ "step": 482
2901
+ },
2902
+ {
2903
+ "epoch": 1.93,
2904
+ "learning_rate": 7.4552957359009634e-06,
2905
+ "loss": 0.3208,
2906
+ "step": 483
2907
+ },
2908
+ {
2909
+ "epoch": 1.94,
2910
+ "learning_rate": 7.4277854195323255e-06,
2911
+ "loss": 0.2843,
2912
+ "step": 484
2913
+ },
2914
+ {
2915
+ "epoch": 1.94,
2916
+ "learning_rate": 7.400275103163687e-06,
2917
+ "loss": 0.3534,
2918
+ "step": 485
2919
+ },
2920
+ {
2921
+ "epoch": 1.94,
2922
+ "learning_rate": 7.372764786795049e-06,
2923
+ "loss": 0.2732,
2924
+ "step": 486
2925
+ },
2926
+ {
2927
+ "epoch": 1.95,
2928
+ "learning_rate": 7.345254470426411e-06,
2929
+ "loss": 0.2605,
2930
+ "step": 487
2931
+ },
2932
+ {
2933
+ "epoch": 1.95,
2934
+ "learning_rate": 7.317744154057772e-06,
2935
+ "loss": 0.3282,
2936
+ "step": 488
2937
+ },
2938
+ {
2939
+ "epoch": 1.96,
2940
+ "learning_rate": 7.290233837689133e-06,
2941
+ "loss": 0.3356,
2942
+ "step": 489
2943
+ },
2944
+ {
2945
+ "epoch": 1.96,
2946
+ "learning_rate": 7.262723521320496e-06,
2947
+ "loss": 0.2869,
2948
+ "step": 490
2949
+ },
2950
+ {
2951
+ "epoch": 1.96,
2952
+ "learning_rate": 7.235213204951857e-06,
2953
+ "loss": 0.3214,
2954
+ "step": 491
2955
+ },
2956
+ {
2957
+ "epoch": 1.97,
2958
+ "learning_rate": 7.207702888583219e-06,
2959
+ "loss": 0.2519,
2960
+ "step": 492
2961
+ },
2962
+ {
2963
+ "epoch": 1.97,
2964
+ "learning_rate": 7.1801925722145814e-06,
2965
+ "loss": 0.3467,
2966
+ "step": 493
2967
+ },
2968
+ {
2969
+ "epoch": 1.98,
2970
+ "learning_rate": 7.152682255845943e-06,
2971
+ "loss": 0.3123,
2972
+ "step": 494
2973
+ },
2974
+ {
2975
+ "epoch": 1.98,
2976
+ "learning_rate": 7.125171939477305e-06,
2977
+ "loss": 0.3143,
2978
+ "step": 495
2979
+ },
2980
+ {
2981
+ "epoch": 1.98,
2982
+ "learning_rate": 7.097661623108666e-06,
2983
+ "loss": 0.2994,
2984
+ "step": 496
2985
+ },
2986
+ {
2987
+ "epoch": 1.99,
2988
+ "learning_rate": 7.070151306740029e-06,
2989
+ "loss": 0.3241,
2990
+ "step": 497
2991
+ },
2992
+ {
2993
+ "epoch": 1.99,
2994
+ "learning_rate": 7.04264099037139e-06,
2995
+ "loss": 0.3239,
2996
+ "step": 498
2997
+ },
2998
+ {
2999
+ "epoch": 2.0,
3000
+ "learning_rate": 7.015130674002751e-06,
3001
+ "loss": 0.2942,
3002
+ "step": 499
3003
+ },
3004
+ {
3005
+ "epoch": 2.0,
3006
+ "learning_rate": 6.9876203576341124e-06,
3007
+ "loss": 0.3155,
3008
+ "step": 500
3009
+ },
3010
+ {
3011
+ "epoch": 2.0,
3012
+ "learning_rate": 6.960110041265475e-06,
3013
+ "loss": 0.1219,
3014
+ "step": 501
3015
+ },
3016
+ {
3017
+ "epoch": 2.01,
3018
+ "learning_rate": 6.9325997248968365e-06,
3019
+ "loss": 0.0989,
3020
+ "step": 502
3021
+ },
3022
+ {
3023
+ "epoch": 2.01,
3024
+ "learning_rate": 6.905089408528199e-06,
3025
+ "loss": 0.1012,
3026
+ "step": 503
3027
+ },
3028
+ {
3029
+ "epoch": 2.02,
3030
+ "learning_rate": 6.877579092159561e-06,
3031
+ "loss": 0.0988,
3032
+ "step": 504
3033
+ },
3034
+ {
3035
+ "epoch": 2.02,
3036
+ "learning_rate": 6.850068775790923e-06,
3037
+ "loss": 0.0915,
3038
+ "step": 505
3039
+ },
3040
+ {
3041
+ "epoch": 2.02,
3042
+ "learning_rate": 6.822558459422284e-06,
3043
+ "loss": 0.097,
3044
+ "step": 506
3045
+ },
3046
+ {
3047
+ "epoch": 2.03,
3048
+ "learning_rate": 6.795048143053645e-06,
3049
+ "loss": 0.0733,
3050
+ "step": 507
3051
+ },
3052
+ {
3053
+ "epoch": 2.03,
3054
+ "learning_rate": 6.767537826685008e-06,
3055
+ "loss": 0.0898,
3056
+ "step": 508
3057
+ },
3058
+ {
3059
+ "epoch": 2.04,
3060
+ "learning_rate": 6.740027510316369e-06,
3061
+ "loss": 0.098,
3062
+ "step": 509
3063
+ },
3064
+ {
3065
+ "epoch": 2.04,
3066
+ "learning_rate": 6.7125171939477305e-06,
3067
+ "loss": 0.1005,
3068
+ "step": 510
3069
+ },
3070
+ {
3071
+ "epoch": 2.04,
3072
+ "learning_rate": 6.6850068775790925e-06,
3073
+ "loss": 0.114,
3074
+ "step": 511
3075
+ },
3076
+ {
3077
+ "epoch": 2.05,
3078
+ "learning_rate": 6.6574965612104546e-06,
3079
+ "loss": 0.0894,
3080
+ "step": 512
3081
+ },
3082
+ {
3083
+ "epoch": 2.05,
3084
+ "learning_rate": 6.629986244841816e-06,
3085
+ "loss": 0.0914,
3086
+ "step": 513
3087
+ },
3088
+ {
3089
+ "epoch": 2.06,
3090
+ "learning_rate": 6.602475928473178e-06,
3091
+ "loss": 0.0679,
3092
+ "step": 514
3093
+ },
3094
+ {
3095
+ "epoch": 2.06,
3096
+ "learning_rate": 6.57496561210454e-06,
3097
+ "loss": 0.0758,
3098
+ "step": 515
3099
+ },
3100
+ {
3101
+ "epoch": 2.06,
3102
+ "learning_rate": 6.547455295735902e-06,
3103
+ "loss": 0.0863,
3104
+ "step": 516
3105
+ },
3106
+ {
3107
+ "epoch": 2.07,
3108
+ "learning_rate": 6.519944979367263e-06,
3109
+ "loss": 0.0891,
3110
+ "step": 517
3111
+ },
3112
+ {
3113
+ "epoch": 2.07,
3114
+ "learning_rate": 6.492434662998624e-06,
3115
+ "loss": 0.0713,
3116
+ "step": 518
3117
+ },
3118
+ {
3119
+ "epoch": 2.08,
3120
+ "learning_rate": 6.464924346629987e-06,
3121
+ "loss": 0.0867,
3122
+ "step": 519
3123
+ },
3124
+ {
3125
+ "epoch": 2.08,
3126
+ "learning_rate": 6.4374140302613485e-06,
3127
+ "loss": 0.0871,
3128
+ "step": 520
3129
+ },
3130
+ {
3131
+ "epoch": 2.08,
3132
+ "learning_rate": 6.40990371389271e-06,
3133
+ "loss": 0.0778,
3134
+ "step": 521
3135
+ },
3136
+ {
3137
+ "epoch": 2.09,
3138
+ "learning_rate": 6.382393397524072e-06,
3139
+ "loss": 0.0896,
3140
+ "step": 522
3141
+ },
3142
+ {
3143
+ "epoch": 2.09,
3144
+ "learning_rate": 6.354883081155434e-06,
3145
+ "loss": 0.0913,
3146
+ "step": 523
3147
+ },
3148
+ {
3149
+ "epoch": 2.1,
3150
+ "learning_rate": 6.327372764786796e-06,
3151
+ "loss": 0.0807,
3152
+ "step": 524
3153
+ },
3154
+ {
3155
+ "epoch": 2.1,
3156
+ "learning_rate": 6.299862448418157e-06,
3157
+ "loss": 0.0811,
3158
+ "step": 525
3159
+ },
3160
+ {
3161
+ "epoch": 2.1,
3162
+ "learning_rate": 6.27235213204952e-06,
3163
+ "loss": 0.0919,
3164
+ "step": 526
3165
+ },
3166
+ {
3167
+ "epoch": 2.11,
3168
+ "learning_rate": 6.244841815680881e-06,
3169
+ "loss": 0.0605,
3170
+ "step": 527
3171
+ },
3172
+ {
3173
+ "epoch": 2.11,
3174
+ "learning_rate": 6.217331499312242e-06,
3175
+ "loss": 0.0743,
3176
+ "step": 528
3177
+ },
3178
+ {
3179
+ "epoch": 2.12,
3180
+ "learning_rate": 6.189821182943604e-06,
3181
+ "loss": 0.0813,
3182
+ "step": 529
3183
+ },
3184
+ {
3185
+ "epoch": 2.12,
3186
+ "learning_rate": 6.1623108665749665e-06,
3187
+ "loss": 0.0786,
3188
+ "step": 530
3189
+ },
3190
+ {
3191
+ "epoch": 2.12,
3192
+ "learning_rate": 6.134800550206328e-06,
3193
+ "loss": 0.0809,
3194
+ "step": 531
3195
+ },
3196
+ {
3197
+ "epoch": 2.13,
3198
+ "learning_rate": 6.10729023383769e-06,
3199
+ "loss": 0.0692,
3200
+ "step": 532
3201
+ },
3202
+ {
3203
+ "epoch": 2.13,
3204
+ "learning_rate": 6.079779917469052e-06,
3205
+ "loss": 0.0696,
3206
+ "step": 533
3207
+ },
3208
+ {
3209
+ "epoch": 2.14,
3210
+ "learning_rate": 6.052269601100413e-06,
3211
+ "loss": 0.0947,
3212
+ "step": 534
3213
+ },
3214
+ {
3215
+ "epoch": 2.14,
3216
+ "learning_rate": 6.024759284731775e-06,
3217
+ "loss": 0.0876,
3218
+ "step": 535
3219
+ },
3220
+ {
3221
+ "epoch": 2.14,
3222
+ "learning_rate": 5.997248968363136e-06,
3223
+ "loss": 0.0816,
3224
+ "step": 536
3225
+ },
3226
+ {
3227
+ "epoch": 2.15,
3228
+ "learning_rate": 5.969738651994499e-06,
3229
+ "loss": 0.0722,
3230
+ "step": 537
3231
+ },
3232
+ {
3233
+ "epoch": 2.15,
3234
+ "learning_rate": 5.94222833562586e-06,
3235
+ "loss": 0.0823,
3236
+ "step": 538
3237
+ },
3238
+ {
3239
+ "epoch": 2.16,
3240
+ "learning_rate": 5.914718019257222e-06,
3241
+ "loss": 0.0916,
3242
+ "step": 539
3243
+ },
3244
+ {
3245
+ "epoch": 2.16,
3246
+ "learning_rate": 5.887207702888584e-06,
3247
+ "loss": 0.0696,
3248
+ "step": 540
3249
+ },
3250
+ {
3251
+ "epoch": 2.16,
3252
+ "learning_rate": 5.859697386519946e-06,
3253
+ "loss": 0.107,
3254
+ "step": 541
3255
+ },
3256
+ {
3257
+ "epoch": 2.17,
3258
+ "learning_rate": 5.832187070151307e-06,
3259
+ "loss": 0.0832,
3260
+ "step": 542
3261
+ },
3262
+ {
3263
+ "epoch": 2.17,
3264
+ "learning_rate": 5.804676753782669e-06,
3265
+ "loss": 0.0955,
3266
+ "step": 543
3267
+ },
3268
+ {
3269
+ "epoch": 2.18,
3270
+ "learning_rate": 5.777166437414031e-06,
3271
+ "loss": 0.0934,
3272
+ "step": 544
3273
+ },
3274
+ {
3275
+ "epoch": 2.18,
3276
+ "learning_rate": 5.749656121045393e-06,
3277
+ "loss": 0.0974,
3278
+ "step": 545
3279
+ },
3280
+ {
3281
+ "epoch": 2.18,
3282
+ "learning_rate": 5.722145804676754e-06,
3283
+ "loss": 0.076,
3284
+ "step": 546
3285
+ },
3286
+ {
3287
+ "epoch": 2.19,
3288
+ "learning_rate": 5.6946354883081155e-06,
3289
+ "loss": 0.0603,
3290
+ "step": 547
3291
+ },
3292
+ {
3293
+ "epoch": 2.19,
3294
+ "learning_rate": 5.667125171939478e-06,
3295
+ "loss": 0.1073,
3296
+ "step": 548
3297
+ },
3298
+ {
3299
+ "epoch": 2.2,
3300
+ "learning_rate": 5.63961485557084e-06,
3301
+ "loss": 0.0638,
3302
+ "step": 549
3303
+ },
3304
+ {
3305
+ "epoch": 2.2,
3306
+ "learning_rate": 5.612104539202201e-06,
3307
+ "loss": 0.0747,
3308
+ "step": 550
3309
+ },
3310
+ {
3311
+ "epoch": 2.2,
3312
+ "learning_rate": 5.584594222833563e-06,
3313
+ "loss": 0.0803,
3314
+ "step": 551
3315
+ },
3316
+ {
3317
+ "epoch": 2.21,
3318
+ "learning_rate": 5.557083906464925e-06,
3319
+ "loss": 0.0687,
3320
+ "step": 552
3321
+ },
3322
+ {
3323
+ "epoch": 2.21,
3324
+ "learning_rate": 5.529573590096287e-06,
3325
+ "loss": 0.0822,
3326
+ "step": 553
3327
+ },
3328
+ {
3329
+ "epoch": 2.22,
3330
+ "learning_rate": 5.502063273727648e-06,
3331
+ "loss": 0.087,
3332
+ "step": 554
3333
+ },
3334
+ {
3335
+ "epoch": 2.22,
3336
+ "learning_rate": 5.47455295735901e-06,
3337
+ "loss": 0.0738,
3338
+ "step": 555
3339
+ },
3340
+ {
3341
+ "epoch": 2.22,
3342
+ "learning_rate": 5.447042640990372e-06,
3343
+ "loss": 0.072,
3344
+ "step": 556
3345
+ },
3346
+ {
3347
+ "epoch": 2.23,
3348
+ "learning_rate": 5.4195323246217335e-06,
3349
+ "loss": 0.08,
3350
+ "step": 557
3351
+ },
3352
+ {
3353
+ "epoch": 2.23,
3354
+ "learning_rate": 5.392022008253095e-06,
3355
+ "loss": 0.1007,
3356
+ "step": 558
3357
+ },
3358
+ {
3359
+ "epoch": 2.24,
3360
+ "learning_rate": 5.364511691884458e-06,
3361
+ "loss": 0.0886,
3362
+ "step": 559
3363
+ },
3364
+ {
3365
+ "epoch": 2.24,
3366
+ "learning_rate": 5.337001375515819e-06,
3367
+ "loss": 0.0782,
3368
+ "step": 560
3369
+ },
3370
+ {
3371
+ "epoch": 2.24,
3372
+ "learning_rate": 5.30949105914718e-06,
3373
+ "loss": 0.0879,
3374
+ "step": 561
3375
+ },
3376
+ {
3377
+ "epoch": 2.25,
3378
+ "learning_rate": 5.281980742778542e-06,
3379
+ "loss": 0.0709,
3380
+ "step": 562
3381
+ },
3382
+ {
3383
+ "epoch": 2.25,
3384
+ "learning_rate": 5.254470426409904e-06,
3385
+ "loss": 0.0539,
3386
+ "step": 563
3387
+ },
3388
+ {
3389
+ "epoch": 2.26,
3390
+ "learning_rate": 5.226960110041266e-06,
3391
+ "loss": 0.0815,
3392
+ "step": 564
3393
+ },
3394
+ {
3395
+ "epoch": 2.26,
3396
+ "learning_rate": 5.199449793672627e-06,
3397
+ "loss": 0.0745,
3398
+ "step": 565
3399
+ },
3400
+ {
3401
+ "epoch": 2.26,
3402
+ "learning_rate": 5.17193947730399e-06,
3403
+ "loss": 0.0769,
3404
+ "step": 566
3405
+ },
3406
+ {
3407
+ "epoch": 2.27,
3408
+ "learning_rate": 5.1444291609353515e-06,
3409
+ "loss": 0.0723,
3410
+ "step": 567
3411
+ },
3412
+ {
3413
+ "epoch": 2.27,
3414
+ "learning_rate": 5.116918844566713e-06,
3415
+ "loss": 0.0713,
3416
+ "step": 568
3417
+ },
3418
+ {
3419
+ "epoch": 2.28,
3420
+ "learning_rate": 5.089408528198074e-06,
3421
+ "loss": 0.085,
3422
+ "step": 569
3423
+ },
3424
+ {
3425
+ "epoch": 2.28,
3426
+ "learning_rate": 5.061898211829437e-06,
3427
+ "loss": 0.0857,
3428
+ "step": 570
3429
+ },
3430
+ {
3431
+ "epoch": 2.28,
3432
+ "learning_rate": 5.034387895460798e-06,
3433
+ "loss": 0.0728,
3434
+ "step": 571
3435
+ },
3436
+ {
3437
+ "epoch": 2.29,
3438
+ "learning_rate": 5.00687757909216e-06,
3439
+ "loss": 0.0733,
3440
+ "step": 572
3441
+ },
3442
+ {
3443
+ "epoch": 2.29,
3444
+ "learning_rate": 4.979367262723522e-06,
3445
+ "loss": 0.0727,
3446
+ "step": 573
3447
+ },
3448
+ {
3449
+ "epoch": 2.3,
3450
+ "learning_rate": 4.951856946354883e-06,
3451
+ "loss": 0.0765,
3452
+ "step": 574
3453
+ },
3454
+ {
3455
+ "epoch": 2.3,
3456
+ "learning_rate": 4.9243466299862454e-06,
3457
+ "loss": 0.0732,
3458
+ "step": 575
3459
+ },
3460
+ {
3461
+ "epoch": 2.3,
3462
+ "learning_rate": 4.8968363136176075e-06,
3463
+ "loss": 0.0889,
3464
+ "step": 576
3465
+ },
3466
+ {
3467
+ "epoch": 2.31,
3468
+ "learning_rate": 4.869325997248969e-06,
3469
+ "loss": 0.0803,
3470
+ "step": 577
3471
+ },
3472
+ {
3473
+ "epoch": 2.31,
3474
+ "learning_rate": 4.841815680880331e-06,
3475
+ "loss": 0.0759,
3476
+ "step": 578
3477
+ },
3478
+ {
3479
+ "epoch": 2.32,
3480
+ "learning_rate": 4.814305364511692e-06,
3481
+ "loss": 0.0818,
3482
+ "step": 579
3483
+ },
3484
+ {
3485
+ "epoch": 2.32,
3486
+ "learning_rate": 4.786795048143054e-06,
3487
+ "loss": 0.0996,
3488
+ "step": 580
3489
+ },
3490
+ {
3491
+ "epoch": 2.32,
3492
+ "learning_rate": 4.759284731774415e-06,
3493
+ "loss": 0.0707,
3494
+ "step": 581
3495
+ },
3496
+ {
3497
+ "epoch": 2.33,
3498
+ "learning_rate": 4.731774415405777e-06,
3499
+ "loss": 0.0908,
3500
+ "step": 582
3501
+ },
3502
+ {
3503
+ "epoch": 2.33,
3504
+ "learning_rate": 4.704264099037139e-06,
3505
+ "loss": 0.0724,
3506
+ "step": 583
3507
+ },
3508
+ {
3509
+ "epoch": 2.34,
3510
+ "learning_rate": 4.676753782668501e-06,
3511
+ "loss": 0.072,
3512
+ "step": 584
3513
+ },
3514
+ {
3515
+ "epoch": 2.34,
3516
+ "learning_rate": 4.649243466299863e-06,
3517
+ "loss": 0.071,
3518
+ "step": 585
3519
+ },
3520
+ {
3521
+ "epoch": 2.34,
3522
+ "learning_rate": 4.621733149931225e-06,
3523
+ "loss": 0.0762,
3524
+ "step": 586
3525
+ },
3526
+ {
3527
+ "epoch": 2.35,
3528
+ "learning_rate": 4.594222833562587e-06,
3529
+ "loss": 0.0753,
3530
+ "step": 587
3531
+ },
3532
+ {
3533
+ "epoch": 2.35,
3534
+ "learning_rate": 4.566712517193948e-06,
3535
+ "loss": 0.0938,
3536
+ "step": 588
3537
+ },
3538
+ {
3539
+ "epoch": 2.36,
3540
+ "learning_rate": 4.53920220082531e-06,
3541
+ "loss": 0.0718,
3542
+ "step": 589
3543
+ },
3544
+ {
3545
+ "epoch": 2.36,
3546
+ "learning_rate": 4.511691884456671e-06,
3547
+ "loss": 0.0851,
3548
+ "step": 590
3549
+ },
3550
+ {
3551
+ "epoch": 2.36,
3552
+ "learning_rate": 4.484181568088033e-06,
3553
+ "loss": 0.0891,
3554
+ "step": 591
3555
+ },
3556
+ {
3557
+ "epoch": 2.37,
3558
+ "learning_rate": 4.456671251719395e-06,
3559
+ "loss": 0.0827,
3560
+ "step": 592
3561
+ },
3562
+ {
3563
+ "epoch": 2.37,
3564
+ "learning_rate": 4.429160935350757e-06,
3565
+ "loss": 0.0808,
3566
+ "step": 593
3567
+ },
3568
+ {
3569
+ "epoch": 2.38,
3570
+ "learning_rate": 4.4016506189821186e-06,
3571
+ "loss": 0.0902,
3572
+ "step": 594
3573
+ },
3574
+ {
3575
+ "epoch": 2.38,
3576
+ "learning_rate": 4.374140302613481e-06,
3577
+ "loss": 0.0676,
3578
+ "step": 595
3579
+ },
3580
+ {
3581
+ "epoch": 2.38,
3582
+ "learning_rate": 4.346629986244843e-06,
3583
+ "loss": 0.0716,
3584
+ "step": 596
3585
+ },
3586
+ {
3587
+ "epoch": 2.39,
3588
+ "learning_rate": 4.319119669876204e-06,
3589
+ "loss": 0.0769,
3590
+ "step": 597
3591
+ },
3592
+ {
3593
+ "epoch": 2.39,
3594
+ "learning_rate": 4.291609353507566e-06,
3595
+ "loss": 0.0815,
3596
+ "step": 598
3597
+ },
3598
+ {
3599
+ "epoch": 2.4,
3600
+ "learning_rate": 4.264099037138927e-06,
3601
+ "loss": 0.0698,
3602
+ "step": 599
3603
+ },
3604
+ {
3605
+ "epoch": 2.4,
3606
+ "learning_rate": 4.236588720770289e-06,
3607
+ "loss": 0.0777,
3608
+ "step": 600
3609
+ },
3610
+ {
3611
+ "epoch": 2.4,
3612
+ "learning_rate": 4.209078404401651e-06,
3613
+ "loss": 0.0933,
3614
+ "step": 601
3615
+ },
3616
+ {
3617
+ "epoch": 2.41,
3618
+ "learning_rate": 4.1815680880330125e-06,
3619
+ "loss": 0.0835,
3620
+ "step": 602
3621
+ },
3622
+ {
3623
+ "epoch": 2.41,
3624
+ "learning_rate": 4.1540577716643745e-06,
3625
+ "loss": 0.0704,
3626
+ "step": 603
3627
+ },
3628
+ {
3629
+ "epoch": 2.42,
3630
+ "learning_rate": 4.1265474552957366e-06,
3631
+ "loss": 0.0784,
3632
+ "step": 604
3633
+ },
3634
+ {
3635
+ "epoch": 2.42,
3636
+ "learning_rate": 4.099037138927098e-06,
3637
+ "loss": 0.0867,
3638
+ "step": 605
3639
+ },
3640
+ {
3641
+ "epoch": 2.42,
3642
+ "learning_rate": 4.07152682255846e-06,
3643
+ "loss": 0.0947,
3644
+ "step": 606
3645
+ },
3646
+ {
3647
+ "epoch": 2.43,
3648
+ "learning_rate": 4.044016506189822e-06,
3649
+ "loss": 0.0907,
3650
+ "step": 607
3651
+ },
3652
+ {
3653
+ "epoch": 2.43,
3654
+ "learning_rate": 4.016506189821183e-06,
3655
+ "loss": 0.0735,
3656
+ "step": 608
3657
+ },
3658
+ {
3659
+ "epoch": 2.44,
3660
+ "learning_rate": 3.988995873452545e-06,
3661
+ "loss": 0.0667,
3662
+ "step": 609
3663
+ },
3664
+ {
3665
+ "epoch": 2.44,
3666
+ "learning_rate": 3.961485557083906e-06,
3667
+ "loss": 0.0788,
3668
+ "step": 610
3669
+ },
3670
+ {
3671
+ "epoch": 2.44,
3672
+ "learning_rate": 3.933975240715268e-06,
3673
+ "loss": 0.0684,
3674
+ "step": 611
3675
+ },
3676
+ {
3677
+ "epoch": 2.45,
3678
+ "learning_rate": 3.9064649243466305e-06,
3679
+ "loss": 0.078,
3680
+ "step": 612
3681
+ },
3682
+ {
3683
+ "epoch": 2.45,
3684
+ "learning_rate": 3.8789546079779925e-06,
3685
+ "loss": 0.103,
3686
+ "step": 613
3687
+ },
3688
+ {
3689
+ "epoch": 2.46,
3690
+ "learning_rate": 3.851444291609354e-06,
3691
+ "loss": 0.0702,
3692
+ "step": 614
3693
+ },
3694
+ {
3695
+ "epoch": 2.46,
3696
+ "learning_rate": 3.823933975240716e-06,
3697
+ "loss": 0.0681,
3698
+ "step": 615
3699
+ },
3700
+ {
3701
+ "epoch": 2.46,
3702
+ "learning_rate": 3.796423658872077e-06,
3703
+ "loss": 0.0629,
3704
+ "step": 616
3705
+ },
3706
+ {
3707
+ "epoch": 2.47,
3708
+ "learning_rate": 3.768913342503439e-06,
3709
+ "loss": 0.0641,
3710
+ "step": 617
3711
+ },
3712
+ {
3713
+ "epoch": 2.47,
3714
+ "learning_rate": 3.741403026134801e-06,
3715
+ "loss": 0.069,
3716
+ "step": 618
3717
+ },
3718
+ {
3719
+ "epoch": 2.48,
3720
+ "learning_rate": 3.7138927097661627e-06,
3721
+ "loss": 0.0894,
3722
+ "step": 619
3723
+ },
3724
+ {
3725
+ "epoch": 2.48,
3726
+ "learning_rate": 3.6863823933975244e-06,
3727
+ "loss": 0.0581,
3728
+ "step": 620
3729
+ },
3730
+ {
3731
+ "epoch": 2.48,
3732
+ "learning_rate": 3.658872077028886e-06,
3733
+ "loss": 0.076,
3734
+ "step": 621
3735
+ },
3736
+ {
3737
+ "epoch": 2.49,
3738
+ "learning_rate": 3.631361760660248e-06,
3739
+ "loss": 0.079,
3740
+ "step": 622
3741
+ },
3742
+ {
3743
+ "epoch": 2.49,
3744
+ "learning_rate": 3.6038514442916097e-06,
3745
+ "loss": 0.0614,
3746
+ "step": 623
3747
+ },
3748
+ {
3749
+ "epoch": 2.5,
3750
+ "learning_rate": 3.5763411279229713e-06,
3751
+ "loss": 0.0906,
3752
+ "step": 624
3753
+ },
3754
+ {
3755
+ "epoch": 2.5,
3756
+ "learning_rate": 3.548830811554333e-06,
3757
+ "loss": 0.0733,
3758
+ "step": 625
3759
+ },
3760
+ {
3761
+ "epoch": 2.5,
3762
+ "learning_rate": 3.521320495185695e-06,
3763
+ "loss": 0.0835,
3764
+ "step": 626
3765
+ },
3766
+ {
3767
+ "epoch": 2.51,
3768
+ "learning_rate": 3.4938101788170562e-06,
3769
+ "loss": 0.0871,
3770
+ "step": 627
3771
+ },
3772
+ {
3773
+ "epoch": 2.51,
3774
+ "learning_rate": 3.4662998624484183e-06,
3775
+ "loss": 0.0605,
3776
+ "step": 628
3777
+ },
3778
+ {
3779
+ "epoch": 2.52,
3780
+ "learning_rate": 3.4387895460797803e-06,
3781
+ "loss": 0.0884,
3782
+ "step": 629
3783
+ },
3784
+ {
3785
+ "epoch": 2.52,
3786
+ "learning_rate": 3.411279229711142e-06,
3787
+ "loss": 0.0839,
3788
+ "step": 630
3789
+ },
3790
+ {
3791
+ "epoch": 2.52,
3792
+ "learning_rate": 3.383768913342504e-06,
3793
+ "loss": 0.0678,
3794
+ "step": 631
3795
+ },
3796
+ {
3797
+ "epoch": 2.53,
3798
+ "learning_rate": 3.3562585969738652e-06,
3799
+ "loss": 0.086,
3800
+ "step": 632
3801
+ },
3802
+ {
3803
+ "epoch": 2.53,
3804
+ "learning_rate": 3.3287482806052273e-06,
3805
+ "loss": 0.0747,
3806
+ "step": 633
3807
+ },
3808
+ {
3809
+ "epoch": 2.54,
3810
+ "learning_rate": 3.301237964236589e-06,
3811
+ "loss": 0.081,
3812
+ "step": 634
3813
+ },
3814
+ {
3815
+ "epoch": 2.54,
3816
+ "learning_rate": 3.273727647867951e-06,
3817
+ "loss": 0.0888,
3818
+ "step": 635
3819
+ },
3820
+ {
3821
+ "epoch": 2.54,
3822
+ "learning_rate": 3.246217331499312e-06,
3823
+ "loss": 0.0762,
3824
+ "step": 636
3825
+ },
3826
+ {
3827
+ "epoch": 2.55,
3828
+ "learning_rate": 3.2187070151306742e-06,
3829
+ "loss": 0.07,
3830
+ "step": 637
3831
+ },
3832
+ {
3833
+ "epoch": 2.55,
3834
+ "learning_rate": 3.191196698762036e-06,
3835
+ "loss": 0.0863,
3836
+ "step": 638
3837
+ },
3838
+ {
3839
+ "epoch": 2.56,
3840
+ "learning_rate": 3.163686382393398e-06,
3841
+ "loss": 0.087,
3842
+ "step": 639
3843
+ },
3844
+ {
3845
+ "epoch": 2.56,
3846
+ "learning_rate": 3.13617606602476e-06,
3847
+ "loss": 0.0826,
3848
+ "step": 640
3849
+ },
3850
+ {
3851
+ "epoch": 2.56,
3852
+ "learning_rate": 3.108665749656121e-06,
3853
+ "loss": 0.084,
3854
+ "step": 641
3855
+ },
3856
+ {
3857
+ "epoch": 2.57,
3858
+ "learning_rate": 3.0811554332874832e-06,
3859
+ "loss": 0.0635,
3860
+ "step": 642
3861
+ },
3862
+ {
3863
+ "epoch": 2.57,
3864
+ "learning_rate": 3.053645116918845e-06,
3865
+ "loss": 0.0718,
3866
+ "step": 643
3867
+ },
3868
+ {
3869
+ "epoch": 2.58,
3870
+ "learning_rate": 3.0261348005502065e-06,
3871
+ "loss": 0.0676,
3872
+ "step": 644
3873
+ },
3874
+ {
3875
+ "epoch": 2.58,
3876
+ "learning_rate": 2.998624484181568e-06,
3877
+ "loss": 0.0934,
3878
+ "step": 645
3879
+ },
3880
+ {
3881
+ "epoch": 2.58,
3882
+ "learning_rate": 2.97111416781293e-06,
3883
+ "loss": 0.0605,
3884
+ "step": 646
3885
+ },
3886
+ {
3887
+ "epoch": 2.59,
3888
+ "learning_rate": 2.943603851444292e-06,
3889
+ "loss": 0.0711,
3890
+ "step": 647
3891
+ },
3892
+ {
3893
+ "epoch": 2.59,
3894
+ "learning_rate": 2.9160935350756535e-06,
3895
+ "loss": 0.0872,
3896
+ "step": 648
3897
+ },
3898
+ {
3899
+ "epoch": 2.6,
3900
+ "learning_rate": 2.8885832187070155e-06,
3901
+ "loss": 0.0681,
3902
+ "step": 649
3903
+ },
3904
+ {
3905
+ "epoch": 2.6,
3906
+ "learning_rate": 2.861072902338377e-06,
3907
+ "loss": 0.0856,
3908
+ "step": 650
3909
+ },
3910
+ {
3911
+ "epoch": 2.6,
3912
+ "learning_rate": 2.833562585969739e-06,
3913
+ "loss": 0.0611,
3914
+ "step": 651
3915
+ },
3916
+ {
3917
+ "epoch": 2.61,
3918
+ "learning_rate": 2.8060522696011004e-06,
3919
+ "loss": 0.0888,
3920
+ "step": 652
3921
+ },
3922
+ {
3923
+ "epoch": 2.61,
3924
+ "learning_rate": 2.7785419532324625e-06,
3925
+ "loss": 0.0741,
3926
+ "step": 653
3927
+ },
3928
+ {
3929
+ "epoch": 2.62,
3930
+ "learning_rate": 2.751031636863824e-06,
3931
+ "loss": 0.0763,
3932
+ "step": 654
3933
+ },
3934
+ {
3935
+ "epoch": 2.62,
3936
+ "learning_rate": 2.723521320495186e-06,
3937
+ "loss": 0.0909,
3938
+ "step": 655
3939
+ },
3940
+ {
3941
+ "epoch": 2.62,
3942
+ "learning_rate": 2.6960110041265474e-06,
3943
+ "loss": 0.0681,
3944
+ "step": 656
3945
+ },
3946
+ {
3947
+ "epoch": 2.63,
3948
+ "learning_rate": 2.6685006877579094e-06,
3949
+ "loss": 0.079,
3950
+ "step": 657
3951
+ },
3952
+ {
3953
+ "epoch": 2.63,
3954
+ "learning_rate": 2.640990371389271e-06,
3955
+ "loss": 0.078,
3956
+ "step": 658
3957
+ },
3958
+ {
3959
+ "epoch": 2.64,
3960
+ "learning_rate": 2.613480055020633e-06,
3961
+ "loss": 0.0895,
3962
+ "step": 659
3963
+ },
3964
+ {
3965
+ "epoch": 2.64,
3966
+ "learning_rate": 2.585969738651995e-06,
3967
+ "loss": 0.0763,
3968
+ "step": 660
3969
+ },
3970
+ {
3971
+ "epoch": 2.64,
3972
+ "learning_rate": 2.5584594222833564e-06,
3973
+ "loss": 0.0683,
3974
+ "step": 661
3975
+ },
3976
+ {
3977
+ "epoch": 2.65,
3978
+ "learning_rate": 2.5309491059147184e-06,
3979
+ "loss": 0.0771,
3980
+ "step": 662
3981
+ },
3982
+ {
3983
+ "epoch": 2.65,
3984
+ "learning_rate": 2.50343878954608e-06,
3985
+ "loss": 0.0685,
3986
+ "step": 663
3987
+ },
3988
+ {
3989
+ "epoch": 2.66,
3990
+ "learning_rate": 2.4759284731774417e-06,
3991
+ "loss": 0.0835,
3992
+ "step": 664
3993
+ },
3994
+ {
3995
+ "epoch": 2.66,
3996
+ "learning_rate": 2.4484181568088037e-06,
3997
+ "loss": 0.098,
3998
+ "step": 665
3999
+ },
4000
+ {
4001
+ "epoch": 2.66,
4002
+ "learning_rate": 2.4209078404401654e-06,
4003
+ "loss": 0.0889,
4004
+ "step": 666
4005
+ },
4006
+ {
4007
+ "epoch": 2.67,
4008
+ "learning_rate": 2.393397524071527e-06,
4009
+ "loss": 0.0776,
4010
+ "step": 667
4011
+ },
4012
+ {
4013
+ "epoch": 2.67,
4014
+ "learning_rate": 2.3658872077028886e-06,
4015
+ "loss": 0.0817,
4016
+ "step": 668
4017
+ },
4018
+ {
4019
+ "epoch": 2.68,
4020
+ "learning_rate": 2.3383768913342507e-06,
4021
+ "loss": 0.0595,
4022
+ "step": 669
4023
+ },
4024
+ {
4025
+ "epoch": 2.68,
4026
+ "learning_rate": 2.3108665749656123e-06,
4027
+ "loss": 0.075,
4028
+ "step": 670
4029
+ },
4030
+ {
4031
+ "epoch": 2.68,
4032
+ "learning_rate": 2.283356258596974e-06,
4033
+ "loss": 0.0643,
4034
+ "step": 671
4035
+ },
4036
+ {
4037
+ "epoch": 2.69,
4038
+ "learning_rate": 2.2558459422283356e-06,
4039
+ "loss": 0.078,
4040
+ "step": 672
4041
+ },
4042
+ {
4043
+ "epoch": 2.69,
4044
+ "learning_rate": 2.2283356258596976e-06,
4045
+ "loss": 0.0559,
4046
+ "step": 673
4047
+ },
4048
+ {
4049
+ "epoch": 2.7,
4050
+ "learning_rate": 2.2008253094910593e-06,
4051
+ "loss": 0.0719,
4052
+ "step": 674
4053
+ },
4054
+ {
4055
+ "epoch": 2.7,
4056
+ "learning_rate": 2.1733149931224213e-06,
4057
+ "loss": 0.0862,
4058
+ "step": 675
4059
+ },
4060
+ {
4061
+ "epoch": 2.7,
4062
+ "learning_rate": 2.145804676753783e-06,
4063
+ "loss": 0.0607,
4064
+ "step": 676
4065
+ },
4066
+ {
4067
+ "epoch": 2.71,
4068
+ "learning_rate": 2.1182943603851446e-06,
4069
+ "loss": 0.0696,
4070
+ "step": 677
4071
+ },
4072
+ {
4073
+ "epoch": 2.71,
4074
+ "learning_rate": 2.0907840440165062e-06,
4075
+ "loss": 0.0795,
4076
+ "step": 678
4077
+ },
4078
+ {
4079
+ "epoch": 2.72,
4080
+ "learning_rate": 2.0632737276478683e-06,
4081
+ "loss": 0.0901,
4082
+ "step": 679
4083
+ },
4084
+ {
4085
+ "epoch": 2.72,
4086
+ "learning_rate": 2.03576341127923e-06,
4087
+ "loss": 0.0813,
4088
+ "step": 680
4089
+ },
4090
+ {
4091
+ "epoch": 2.72,
4092
+ "learning_rate": 2.0082530949105915e-06,
4093
+ "loss": 0.0811,
4094
+ "step": 681
4095
+ },
4096
+ {
4097
+ "epoch": 2.73,
4098
+ "learning_rate": 1.980742778541953e-06,
4099
+ "loss": 0.0665,
4100
+ "step": 682
4101
+ },
4102
+ {
4103
+ "epoch": 2.73,
4104
+ "learning_rate": 1.9532324621733152e-06,
4105
+ "loss": 0.0742,
4106
+ "step": 683
4107
+ },
4108
+ {
4109
+ "epoch": 2.74,
4110
+ "learning_rate": 1.925722145804677e-06,
4111
+ "loss": 0.0786,
4112
+ "step": 684
4113
+ },
4114
+ {
4115
+ "epoch": 2.74,
4116
+ "learning_rate": 1.8982118294360385e-06,
4117
+ "loss": 0.0844,
4118
+ "step": 685
4119
+ },
4120
+ {
4121
+ "epoch": 2.74,
4122
+ "learning_rate": 1.8707015130674006e-06,
4123
+ "loss": 0.0747,
4124
+ "step": 686
4125
+ },
4126
+ {
4127
+ "epoch": 2.75,
4128
+ "learning_rate": 1.8431911966987622e-06,
4129
+ "loss": 0.0818,
4130
+ "step": 687
4131
+ },
4132
+ {
4133
+ "epoch": 2.75,
4134
+ "learning_rate": 1.815680880330124e-06,
4135
+ "loss": 0.0938,
4136
+ "step": 688
4137
+ },
4138
+ {
4139
+ "epoch": 2.76,
4140
+ "learning_rate": 1.7881705639614857e-06,
4141
+ "loss": 0.0863,
4142
+ "step": 689
4143
+ },
4144
+ {
4145
+ "epoch": 2.76,
4146
+ "learning_rate": 1.7606602475928475e-06,
4147
+ "loss": 0.0875,
4148
+ "step": 690
4149
+ },
4150
+ {
4151
+ "epoch": 2.76,
4152
+ "learning_rate": 1.7331499312242091e-06,
4153
+ "loss": 0.0824,
4154
+ "step": 691
4155
+ },
4156
+ {
4157
+ "epoch": 2.77,
4158
+ "learning_rate": 1.705639614855571e-06,
4159
+ "loss": 0.0751,
4160
+ "step": 692
4161
+ },
4162
+ {
4163
+ "epoch": 2.77,
4164
+ "learning_rate": 1.6781292984869326e-06,
4165
+ "loss": 0.08,
4166
+ "step": 693
4167
+ },
4168
+ {
4169
+ "epoch": 2.78,
4170
+ "learning_rate": 1.6506189821182945e-06,
4171
+ "loss": 0.0569,
4172
+ "step": 694
4173
+ },
4174
+ {
4175
+ "epoch": 2.78,
4176
+ "learning_rate": 1.623108665749656e-06,
4177
+ "loss": 0.0696,
4178
+ "step": 695
4179
+ },
4180
+ {
4181
+ "epoch": 2.78,
4182
+ "learning_rate": 1.595598349381018e-06,
4183
+ "loss": 0.0717,
4184
+ "step": 696
4185
+ },
4186
+ {
4187
+ "epoch": 2.79,
4188
+ "learning_rate": 1.56808803301238e-06,
4189
+ "loss": 0.0548,
4190
+ "step": 697
4191
+ },
4192
+ {
4193
+ "epoch": 2.79,
4194
+ "learning_rate": 1.5405777166437416e-06,
4195
+ "loss": 0.084,
4196
+ "step": 698
4197
+ },
4198
+ {
4199
+ "epoch": 2.8,
4200
+ "learning_rate": 1.5130674002751033e-06,
4201
+ "loss": 0.085,
4202
+ "step": 699
4203
+ },
4204
+ {
4205
+ "epoch": 2.8,
4206
+ "learning_rate": 1.485557083906465e-06,
4207
+ "loss": 0.0848,
4208
+ "step": 700
4209
+ },
4210
+ {
4211
+ "epoch": 2.8,
4212
+ "learning_rate": 1.4580467675378267e-06,
4213
+ "loss": 0.0793,
4214
+ "step": 701
4215
+ },
4216
+ {
4217
+ "epoch": 2.81,
4218
+ "learning_rate": 1.4305364511691886e-06,
4219
+ "loss": 0.0653,
4220
+ "step": 702
4221
+ },
4222
+ {
4223
+ "epoch": 2.81,
4224
+ "learning_rate": 1.4030261348005502e-06,
4225
+ "loss": 0.0813,
4226
+ "step": 703
4227
+ },
4228
+ {
4229
+ "epoch": 2.82,
4230
+ "learning_rate": 1.375515818431912e-06,
4231
+ "loss": 0.0653,
4232
+ "step": 704
4233
+ },
4234
+ {
4235
+ "epoch": 2.82,
4236
+ "learning_rate": 1.3480055020632737e-06,
4237
+ "loss": 0.0773,
4238
+ "step": 705
4239
+ },
4240
+ {
4241
+ "epoch": 2.82,
4242
+ "learning_rate": 1.3204951856946355e-06,
4243
+ "loss": 0.093,
4244
+ "step": 706
4245
+ },
4246
+ {
4247
+ "epoch": 2.83,
4248
+ "learning_rate": 1.2929848693259976e-06,
4249
+ "loss": 0.0606,
4250
+ "step": 707
4251
+ },
4252
+ {
4253
+ "epoch": 2.83,
4254
+ "learning_rate": 1.2654745529573592e-06,
4255
+ "loss": 0.0667,
4256
+ "step": 708
4257
+ },
4258
+ {
4259
+ "epoch": 2.84,
4260
+ "learning_rate": 1.2379642365887208e-06,
4261
+ "loss": 0.0685,
4262
+ "step": 709
4263
+ },
4264
+ {
4265
+ "epoch": 2.84,
4266
+ "learning_rate": 1.2104539202200827e-06,
4267
+ "loss": 0.0701,
4268
+ "step": 710
4269
+ },
4270
+ {
4271
+ "epoch": 2.84,
4272
+ "learning_rate": 1.1829436038514443e-06,
4273
+ "loss": 0.0699,
4274
+ "step": 711
4275
+ },
4276
+ {
4277
+ "epoch": 2.85,
4278
+ "learning_rate": 1.1554332874828062e-06,
4279
+ "loss": 0.0796,
4280
+ "step": 712
4281
+ },
4282
+ {
4283
+ "epoch": 2.85,
4284
+ "learning_rate": 1.1279229711141678e-06,
4285
+ "loss": 0.0907,
4286
+ "step": 713
4287
+ },
4288
+ {
4289
+ "epoch": 2.86,
4290
+ "learning_rate": 1.1004126547455296e-06,
4291
+ "loss": 0.0626,
4292
+ "step": 714
4293
+ },
4294
+ {
4295
+ "epoch": 2.86,
4296
+ "learning_rate": 1.0729023383768915e-06,
4297
+ "loss": 0.0551,
4298
+ "step": 715
4299
+ },
4300
+ {
4301
+ "epoch": 2.86,
4302
+ "learning_rate": 1.0453920220082531e-06,
4303
+ "loss": 0.069,
4304
+ "step": 716
4305
+ },
4306
+ {
4307
+ "epoch": 2.87,
4308
+ "learning_rate": 1.017881705639615e-06,
4309
+ "loss": 0.0748,
4310
+ "step": 717
4311
+ },
4312
+ {
4313
+ "epoch": 2.87,
4314
+ "learning_rate": 9.903713892709766e-07,
4315
+ "loss": 0.0696,
4316
+ "step": 718
4317
+ },
4318
+ {
4319
+ "epoch": 2.88,
4320
+ "learning_rate": 9.628610729023384e-07,
4321
+ "loss": 0.064,
4322
+ "step": 719
4323
+ },
4324
+ {
4325
+ "epoch": 2.88,
4326
+ "learning_rate": 9.353507565337003e-07,
4327
+ "loss": 0.0614,
4328
+ "step": 720
4329
+ },
4330
+ {
4331
+ "epoch": 2.88,
4332
+ "learning_rate": 9.07840440165062e-07,
4333
+ "loss": 0.0667,
4334
+ "step": 721
4335
+ },
4336
+ {
4337
+ "epoch": 2.89,
4338
+ "learning_rate": 8.803301237964238e-07,
4339
+ "loss": 0.0737,
4340
+ "step": 722
4341
+ },
4342
+ {
4343
+ "epoch": 2.89,
4344
+ "learning_rate": 8.528198074277855e-07,
4345
+ "loss": 0.0585,
4346
+ "step": 723
4347
+ },
4348
+ {
4349
+ "epoch": 2.9,
4350
+ "learning_rate": 8.253094910591472e-07,
4351
+ "loss": 0.0814,
4352
+ "step": 724
4353
+ },
4354
+ {
4355
+ "epoch": 2.9,
4356
+ "learning_rate": 7.97799174690509e-07,
4357
+ "loss": 0.0751,
4358
+ "step": 725
4359
+ },
4360
+ {
4361
+ "epoch": 2.9,
4362
+ "learning_rate": 7.702888583218708e-07,
4363
+ "loss": 0.0693,
4364
+ "step": 726
4365
+ },
4366
+ {
4367
+ "epoch": 2.91,
4368
+ "learning_rate": 7.427785419532325e-07,
4369
+ "loss": 0.0771,
4370
+ "step": 727
4371
+ },
4372
+ {
4373
+ "epoch": 2.91,
4374
+ "learning_rate": 7.152682255845943e-07,
4375
+ "loss": 0.0678,
4376
+ "step": 728
4377
+ },
4378
+ {
4379
+ "epoch": 2.92,
4380
+ "learning_rate": 6.87757909215956e-07,
4381
+ "loss": 0.0857,
4382
+ "step": 729
4383
+ },
4384
+ {
4385
+ "epoch": 2.92,
4386
+ "learning_rate": 6.602475928473178e-07,
4387
+ "loss": 0.0857,
4388
+ "step": 730
4389
+ },
4390
+ {
4391
+ "epoch": 2.92,
4392
+ "learning_rate": 6.327372764786796e-07,
4393
+ "loss": 0.0864,
4394
+ "step": 731
4395
+ },
4396
+ {
4397
+ "epoch": 2.93,
4398
+ "learning_rate": 6.052269601100413e-07,
4399
+ "loss": 0.0851,
4400
+ "step": 732
4401
+ },
4402
+ {
4403
+ "epoch": 2.93,
4404
+ "learning_rate": 5.777166437414031e-07,
4405
+ "loss": 0.0889,
4406
+ "step": 733
4407
+ },
4408
+ {
4409
+ "epoch": 2.94,
4410
+ "learning_rate": 5.502063273727648e-07,
4411
+ "loss": 0.0771,
4412
+ "step": 734
4413
+ },
4414
+ {
4415
+ "epoch": 2.94,
4416
+ "learning_rate": 5.226960110041266e-07,
4417
+ "loss": 0.0822,
4418
+ "step": 735
4419
+ },
4420
+ {
4421
+ "epoch": 2.94,
4422
+ "learning_rate": 4.951856946354883e-07,
4423
+ "loss": 0.0642,
4424
+ "step": 736
4425
+ },
4426
+ {
4427
+ "epoch": 2.95,
4428
+ "learning_rate": 4.6767537826685014e-07,
4429
+ "loss": 0.0765,
4430
+ "step": 737
4431
+ },
4432
+ {
4433
+ "epoch": 2.95,
4434
+ "learning_rate": 4.401650618982119e-07,
4435
+ "loss": 0.0821,
4436
+ "step": 738
4437
+ },
4438
+ {
4439
+ "epoch": 2.96,
4440
+ "learning_rate": 4.126547455295736e-07,
4441
+ "loss": 0.0821,
4442
+ "step": 739
4443
+ },
4444
+ {
4445
+ "epoch": 2.96,
4446
+ "learning_rate": 3.851444291609354e-07,
4447
+ "loss": 0.0783,
4448
+ "step": 740
4449
+ },
4450
+ {
4451
+ "epoch": 2.96,
4452
+ "learning_rate": 3.5763411279229714e-07,
4453
+ "loss": 0.0722,
4454
+ "step": 741
4455
+ },
4456
+ {
4457
+ "epoch": 2.97,
4458
+ "learning_rate": 3.301237964236589e-07,
4459
+ "loss": 0.0935,
4460
+ "step": 742
4461
+ },
4462
+ {
4463
+ "epoch": 2.97,
4464
+ "learning_rate": 3.0261348005502067e-07,
4465
+ "loss": 0.0929,
4466
+ "step": 743
4467
+ },
4468
+ {
4469
+ "epoch": 2.98,
4470
+ "learning_rate": 2.751031636863824e-07,
4471
+ "loss": 0.0869,
4472
+ "step": 744
4473
+ },
4474
+ {
4475
+ "epoch": 2.98,
4476
+ "learning_rate": 2.4759284731774415e-07,
4477
+ "loss": 0.1,
4478
+ "step": 745
4479
+ },
4480
+ {
4481
+ "epoch": 2.98,
4482
+ "learning_rate": 2.2008253094910594e-07,
4483
+ "loss": 0.0941,
4484
+ "step": 746
4485
+ },
4486
+ {
4487
+ "epoch": 2.99,
4488
+ "learning_rate": 1.925722145804677e-07,
4489
+ "loss": 0.0596,
4490
+ "step": 747
4491
+ },
4492
+ {
4493
+ "epoch": 2.99,
4494
+ "learning_rate": 1.6506189821182944e-07,
4495
+ "loss": 0.0703,
4496
+ "step": 748
4497
+ },
4498
+ {
4499
+ "epoch": 3.0,
4500
+ "learning_rate": 1.375515818431912e-07,
4501
+ "loss": 0.0806,
4502
+ "step": 749
4503
+ },
4504
+ {
4505
+ "epoch": 3.0,
4506
+ "learning_rate": 1.1004126547455297e-07,
4507
+ "loss": 0.0651,
4508
+ "step": 750
4509
+ },
4510
+ {
4511
+ "epoch": 3.0,
4512
+ "step": 750,
4513
+ "total_flos": 21264552591360.0,
4514
+ "train_loss": 0.3743656844943762,
4515
+ "train_runtime": 11966.0892,
4516
+ "train_samples_per_second": 8.023,
4517
+ "train_steps_per_second": 0.063
4518
+ }
4519
+ ],
4520
+ "max_steps": 750,
4521
+ "num_train_epochs": 3,
4522
+ "total_flos": 21264552591360.0,
4523
+ "trial_name": null,
4524
+ "trial_params": null
4525
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c7ead6a6dbc616da8543def0721208b1345118c05e3572db5baaef2a9bbdc95
3
+ size 5051
zero_to_fp32.py ADDED
@@ -0,0 +1,483 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+ '''Copyright The Microsoft DeepSpeed Team'''
3
+
4
+ # This script extracts fp32 consolidated weights from a zero 2 and 3 DeepSpeed checkpoints. It gets
5
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
6
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
7
+ # application.
8
+ #
9
+ # example: python zero_to_fp32.py . pytorch_model.bin
10
+
11
+ import argparse
12
+ import torch
13
+ import glob
14
+ import math
15
+ import os
16
+ import re
17
+ from collections import OrderedDict
18
+
19
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
20
+ # DeepSpeed data structures it has to be available in the current python environment.
21
+ from deepspeed.utils import logger
22
+ from deepspeed.checkpoint.constants import (DS_VERSION,
23
+ OPTIMIZER_STATE_DICT,
24
+ SINGLE_PARTITION_OF_FP32_GROUPS,
25
+ FP32_FLAT_GROUPS,
26
+ ZERO_STAGE,
27
+ PARTITION_COUNT,
28
+ PARAM_SHAPES,
29
+ BUFFER_NAMES)
30
+
31
+ debug = 0
32
+
33
+ # load to cpu
34
+ device = torch.device('cpu')
35
+
36
+
37
+ def atoi(text):
38
+ return int(text) if text.isdigit() else text
39
+
40
+
41
+ def natural_keys(text):
42
+ '''
43
+ alist.sort(key=natural_keys) sorts in human order
44
+ http://nedbatchelder.com/blog/200712/human_sorting.html
45
+ (See Toothy's implementation in the comments)
46
+ '''
47
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
48
+
49
+
50
+ def get_model_state_file(checkpoint_dir, zero_stage):
51
+ if not os.path.isdir(checkpoint_dir):
52
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
53
+
54
+ # there should be only one file
55
+ if zero_stage == 2:
56
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
57
+ elif zero_stage == 3:
58
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
59
+
60
+ if not os.path.exists(file):
61
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
62
+
63
+ return file
64
+
65
+
66
+ def get_optim_files(checkpoint_dir):
67
+ # XXX: need to test that this simple glob rule works for multi-node setup too
68
+ optim_files = sorted(glob.glob(os.path.join(checkpoint_dir,
69
+ "*_optim_states.pt")),
70
+ key=natural_keys)
71
+
72
+ if len(optim_files) == 0:
73
+ raise FileNotFoundError(
74
+ f"can't find '*_optim_states.pt' files in directory '{checkpoint_dir}'")
75
+
76
+ return optim_files
77
+
78
+
79
+ def parse_model_state(file):
80
+ state_dict = torch.load(file, map_location=device)
81
+
82
+ if BUFFER_NAMES not in state_dict:
83
+ raise ValueError(f"{file} is not a model state checkpoint")
84
+ buffer_names = state_dict[BUFFER_NAMES]
85
+ if debug:
86
+ print("Found buffers:", buffer_names)
87
+
88
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
89
+ buffers = {
90
+ k: v.float()
91
+ for k,
92
+ v in state_dict["module"].items() if k in buffer_names
93
+ }
94
+ param_shapes = state_dict[PARAM_SHAPES]
95
+
96
+ ds_version = state_dict.get(DS_VERSION, None)
97
+
98
+ return buffers, param_shapes, ds_version
99
+
100
+
101
+ def parse_optim_states(files, ds_checkpoint_dir):
102
+
103
+ total_files = len(files)
104
+ state_dicts = []
105
+ for f in files:
106
+ state_dicts.append(torch.load(f, map_location=device))
107
+
108
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
109
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
110
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
111
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
112
+
113
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
114
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
115
+ # use the max of the partition_count to get the dp world_size.
116
+
117
+ if type(world_size) is list:
118
+ world_size = max(world_size)
119
+
120
+ if world_size != total_files:
121
+ raise ValueError(
122
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
123
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
124
+ )
125
+
126
+ # the groups are named differently in each stage
127
+ if zero_stage == 2:
128
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
129
+ elif zero_stage == 3:
130
+ fp32_groups_key = FP32_FLAT_GROUPS
131
+ else:
132
+ raise ValueError(f"unknown zero stage {zero_stage}")
133
+
134
+ if zero_stage == 2:
135
+ fp32_flat_groups = [
136
+ state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key]
137
+ for i in range(len(state_dicts))
138
+ ]
139
+ elif zero_stage == 3:
140
+ # if there is more than one param group, there will be multiple flattened tensors - one
141
+ # flattened tensor per group - for simplicity merge them into a single tensor
142
+ #
143
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
144
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
145
+
146
+ fp32_flat_groups = [
147
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key],
148
+ 0) for i in range(len(state_dicts))
149
+ ]
150
+
151
+ return zero_stage, world_size, fp32_flat_groups
152
+
153
+
154
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
155
+ """
156
+ Returns fp32 state_dict reconstructed from ds checkpoint
157
+
158
+ Args:
159
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
160
+
161
+ """
162
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
163
+
164
+ optim_files = get_optim_files(ds_checkpoint_dir)
165
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
166
+ print(
167
+ f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
168
+
169
+ model_file = get_model_state_file(ds_checkpoint_dir, zero_stage)
170
+ buffers, param_shapes, ds_version = parse_model_state(model_file)
171
+ print(f'Parsing checkpoint created by deepspeed=={ds_version}')
172
+
173
+ if zero_stage == 2:
174
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size,
175
+ param_shapes,
176
+ fp32_flat_groups,
177
+ buffers)
178
+ elif zero_stage == 3:
179
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size,
180
+ param_shapes,
181
+ fp32_flat_groups,
182
+ buffers)
183
+
184
+
185
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size,
186
+ param_shapes,
187
+ fp32_flat_groups,
188
+ buffers):
189
+
190
+ # Reconstruction protocol:
191
+ #
192
+ # XXX: document this
193
+
194
+ if debug:
195
+ for i in range(world_size):
196
+ for j in range(len(fp32_flat_groups[0])):
197
+ print(
198
+ f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
199
+
200
+ # XXX: memory usage doubles here (zero2)
201
+ num_param_groups = len(fp32_flat_groups[0])
202
+ merged_single_partition_of_fp32_groups = []
203
+ for i in range(num_param_groups):
204
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
205
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
206
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
207
+ avail_numel = sum([
208
+ full_single_fp32_vector.numel()
209
+ for full_single_fp32_vector in merged_single_partition_of_fp32_groups
210
+ ])
211
+
212
+ if debug:
213
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
214
+ wanted_numel = sum(
215
+ [sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
216
+ # not asserting if there is a mismatch due to possible padding
217
+ print(f"Have {avail_numel} numels to process.")
218
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
219
+
220
+ state_dict = OrderedDict()
221
+
222
+ # buffers
223
+ state_dict.update(buffers)
224
+ if debug:
225
+ print(f"added {len(buffers)} buffers")
226
+
227
+ # params
228
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
229
+ # out-of-core computing solution
230
+ total_numel = 0
231
+ total_params = 0
232
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
233
+ offset = 0
234
+ avail_numel = full_single_fp32_vector.numel()
235
+ for name, shape in shapes.items():
236
+
237
+ unpartitioned_numel = shape.numel()
238
+ total_numel += unpartitioned_numel
239
+ total_params += 1
240
+
241
+ if debug:
242
+ print(
243
+ f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} "
244
+ )
245
+ state_dict[name] = full_single_fp32_vector.narrow(
246
+ 0,
247
+ offset,
248
+ unpartitioned_numel).view(shape)
249
+ offset += unpartitioned_numel
250
+
251
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
252
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
253
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
254
+ # live optimizer object, so we are checking that the numbers are within the right range
255
+ align_to = 2 * world_size
256
+
257
+ def zero2_align(x):
258
+ return align_to * math.ceil(x / align_to)
259
+
260
+ if debug:
261
+ print(f"original offset={offset}, avail_numel={avail_numel}")
262
+
263
+ offset = zero2_align(offset)
264
+ avail_numel = zero2_align(avail_numel)
265
+
266
+ if debug:
267
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
268
+
269
+ # Sanity check
270
+ if offset != avail_numel:
271
+ raise ValueError(
272
+ f"consumed {offset} numels out of {avail_numel} - something is wrong")
273
+
274
+ print(
275
+ f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements"
276
+ )
277
+
278
+ return state_dict
279
+
280
+
281
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
282
+ remainder = unpartitioned_numel % world_size
283
+ padding_numel = (world_size - remainder) if remainder else 0
284
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
285
+ return partitioned_numel, padding_numel
286
+
287
+
288
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size,
289
+ param_shapes,
290
+ fp32_flat_groups,
291
+ buffers):
292
+
293
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
294
+ # param, re-consolidating each param, while dealing with padding if any
295
+
296
+ avail_numel = fp32_flat_groups[0].numel() * world_size
297
+ # merge list of dicts, preserving order
298
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
299
+
300
+ if debug:
301
+ for i in range(world_size):
302
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
303
+
304
+ wanted_params = len(param_shapes)
305
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
306
+ # not asserting if there is a mismatch due to possible padding
307
+ print(f"Have {avail_numel} numels to process.")
308
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
309
+
310
+ state_dict = OrderedDict()
311
+
312
+ # buffers
313
+ state_dict.update(buffers)
314
+ if debug:
315
+ print(f"added {len(buffers)} buffers")
316
+
317
+ # params
318
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
319
+ # out-of-core computing solution
320
+ offset = 0
321
+ total_numel = 0
322
+ total_params = 0
323
+ for name, shape in param_shapes.items():
324
+
325
+ unpartitioned_numel = shape.numel()
326
+ total_numel += unpartitioned_numel
327
+ total_params += 1
328
+
329
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
330
+
331
+ if debug:
332
+ print(
333
+ f"{total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
334
+ )
335
+
336
+ # XXX: memory usage doubles here
337
+ state_dict[name] = torch.cat(
338
+ tuple(fp32_flat_groups[i].narrow(0,
339
+ offset,
340
+ partitioned_numel)
341
+ for i in range(world_size)),
342
+ 0).narrow(0,
343
+ 0,
344
+ unpartitioned_numel).view(shape)
345
+ offset += partitioned_numel
346
+
347
+ offset *= world_size
348
+
349
+ # Sanity check
350
+ if offset != avail_numel:
351
+ raise ValueError(
352
+ f"consumed {offset} numels out of {avail_numel} - something is wrong")
353
+
354
+ print(
355
+ f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements"
356
+ )
357
+
358
+ return state_dict
359
+
360
+
361
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
362
+ """
363
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
364
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
365
+ via a model hub.
366
+
367
+ Args:
368
+ - ``checkpoint_dir``: path to the desired checkpoint folder
369
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
370
+
371
+ Returns:
372
+ - pytorch ``state_dict``
373
+
374
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
375
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
376
+ the checkpoint.
377
+
378
+ A typical usage might be ::
379
+
380
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
381
+ # do the training and checkpoint saving
382
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
383
+ model = model.cpu() # move to cpu
384
+ model.load_state_dict(state_dict)
385
+ # submit to model hub or save the model to share with others
386
+
387
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
388
+ application. i.e. you will need to re-initialize the deepspeed engine, since
389
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
390
+
391
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
392
+
393
+ """
394
+ if tag is None:
395
+ latest_path = os.path.join(checkpoint_dir, 'latest')
396
+ if os.path.isfile(latest_path):
397
+ with open(latest_path, 'r') as fd:
398
+ tag = fd.read().strip()
399
+ else:
400
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
401
+
402
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
403
+
404
+ if not os.path.isdir(ds_checkpoint_dir):
405
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
406
+
407
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
408
+
409
+
410
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
411
+ """
412
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
413
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
414
+
415
+ Args:
416
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
417
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
418
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
419
+ """
420
+
421
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
422
+ print(f"Saving fp32 state dict to {output_file}")
423
+ torch.save(state_dict, output_file)
424
+
425
+
426
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
427
+ """
428
+ 1. Put the provided model to cpu
429
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
430
+ 3. Load it into the provided model
431
+
432
+ Args:
433
+ - ``model``: the model object to update
434
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
435
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
436
+
437
+ Returns:
438
+ - ``model`: modified model
439
+
440
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
441
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
442
+ conveniently placed for you in the checkpoint folder.
443
+
444
+ A typical usage might be ::
445
+
446
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
447
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
448
+ # submit to model hub or save the model to share with others
449
+
450
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
451
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
452
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
453
+
454
+ """
455
+ logger.info(f"Extracting fp32 weights")
456
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
457
+
458
+ logger.info(f"Overwriting model with fp32 weights")
459
+ model = model.cpu()
460
+ model.load_state_dict(state_dict, strict=False)
461
+
462
+ return model
463
+
464
+
465
+ if __name__ == "__main__":
466
+
467
+ parser = argparse.ArgumentParser()
468
+ parser.add_argument(
469
+ "checkpoint_dir",
470
+ type=str,
471
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
472
+ parser.add_argument(
473
+ "output_file",
474
+ type=str,
475
+ help=
476
+ "path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)"
477
+ )
478
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
479
+ args = parser.parse_args()
480
+
481
+ debug = args.debug
482
+
483
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file)