Hugging Face's logo

nntoan209
/
bgem3-crosslingual-e1

Sentence Similarity
sentence-transformers
Safetensors
xlm-roberta
feature-extraction
Inference Endpoints
text-embeddings-inference
Model card Files Community
nntoan209 commited on
Commit
2d2fc18
•
1 Parent(s): 656dc89

Upload folder using huggingface_hub

Browse files
Files changed (19) hide show
  1. .gitattributes +1 -0
  2. 1_Pooling/config.json +10 -0
  3. README.md +57 -0
  4. colbert_linear.pt +3 -0
  5. config.json +28 -0
  6. config_sentence_transformers.json +9 -0
  7. latest +1 -0
  8. model.safetensors +3 -0
  9. modules.json +20 -0
  10. rng_state.pth +3 -0
  11. sentence_bert_config.json +4 -0
  12. sentencepiece.bpe.model +3 -0
  13. sparse_linear.pt +3 -0
  14. special_tokens_map.json +51 -0
  15. tokenizer.json +3 -0
  16. tokenizer_config.json +55 -0
  17. trainer_state.json +3010 -0
  18. training_args.bin +3 -0
  19. zero_to_fp32.py +592 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
1_Pooling/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "word_embedding_dimension": 1024,
3
+ "pooling_mode_cls_token": true,
4
+ "pooling_mode_mean_tokens": false,
5
+ "pooling_mode_max_tokens": false,
6
+ "pooling_mode_mean_sqrt_len_tokens": false,
7
+ "pooling_mode_weightedmean_tokens": false,
8
+ "pooling_mode_lasttoken": false,
9
+ "include_prompt": true
10
+ }
README.md ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: sentence-transformers
3
+ pipeline_tag: sentence-similarity
4
+ tags:
5
+ - sentence-transformers
6
+ - feature-extraction
7
+ - sentence-similarity
8
+
9
+ ---
10
+
11
+ # {MODEL_NAME}
12
+
13
+ This is a [sentence-transformers](https://www.SBERT.net) model: It maps sentences & paragraphs to a 1024 dimensional dense vector space and can be used for tasks like clustering or semantic search.
14
+
15
+ <!--- Describe your model here -->
16
+
17
+ ## Usage (Sentence-Transformers)
18
+
19
+ Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
20
+
21
+ ```
22
+ pip install -U sentence-transformers
23
+ ```
24
+
25
+ Then you can use the model like this:
26
+
27
+ ```python
28
+ from sentence_transformers import SentenceTransformer
29
+ sentences = ["This is an example sentence", "Each sentence is converted"]
30
+
31
+ model = SentenceTransformer('{MODEL_NAME}')
32
+ embeddings = model.encode(sentences)
33
+ print(embeddings)
34
+ ```
35
+
36
+
37
+
38
+ ## Evaluation Results
39
+
40
+ <!--- Describe how your model was evaluated -->
41
+
42
+ For an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: [https://seb.sbert.net](https://seb.sbert.net?model_name={MODEL_NAME})
43
+
44
+
45
+
46
+ ## Full Model Architecture
47
+ ```
48
+ SentenceTransformer(
49
+ (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: XLMRobertaModel
50
+ (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
51
+ (2): Normalize()
52
+ )
53
+ ```
54
+
55
+ ## Citing & Authors
56
+
57
+ <!--- Describe where people can find more information -->
colbert_linear.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:206fa2734e5d626bbb1cfe48b795e4495d2a238c66e095c6c2d5c47c6e46e8d7
3
+ size 2100674
config.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "saved_models/cross_lingual/bgem3_crosslingual_20240613/tmp-checkpoint-8546",
3
+ "architectures": [
4
+ "XLMRobertaModel"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": null,
9
+ "eos_token_id": 2,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout_prob": 0.1,
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4096,
15
+ "layer_norm_eps": 1e-05,
16
+ "max_position_embeddings": 8194,
17
+ "model_type": "xlm-roberta",
18
+ "num_attention_heads": 16,
19
+ "num_hidden_layers": 24,
20
+ "output_past": true,
21
+ "pad_token_id": 1,
22
+ "position_embedding_type": "absolute",
23
+ "torch_dtype": "float32",
24
+ "transformers_version": "4.38.2",
25
+ "type_vocab_size": 1,
26
+ "use_cache": true,
27
+ "vocab_size": 250002
28
+ }
config_sentence_transformers.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "__version__": {
3
+ "sentence_transformers": "2.5.1",
4
+ "transformers": "4.38.2",
5
+ "pytorch": "2.1.0+cu121"
6
+ },
7
+ "prompts": {},
8
+ "default_prompt_name": null
9
+ }
latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step8546
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0675f12f6c670bc0a8a880ec04dd4ab371f5c775da0cf0e897a709cbf4992e9a
3
+ size 2271064456
modules.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "idx": 0,
4
+ "name": "0",
5
+ "path": "",
6
+ "type": "sentence_transformers.models.Transformer"
7
+ },
8
+ {
9
+ "idx": 1,
10
+ "name": "1",
11
+ "path": "1_Pooling",
12
+ "type": "sentence_transformers.models.Pooling"
13
+ },
14
+ {
15
+ "idx": 2,
16
+ "name": "2",
17
+ "path": "2_Normalize",
18
+ "type": "sentence_transformers.models.Normalize"
19
+ }
20
+ ]
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e99e78c0db6807a7dcb6ba51f06893fe45acc3007468ac54e3d9e5c7a5accd0
3
+ size 14244
sentence_bert_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "max_seq_length": 8192,
3
+ "do_lower_case": false
4
+ }
sentencepiece.bpe.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cfc8146abe2a0488e9e2a0c56de7952f7c11ab059eca145a0a727afce0db2865
3
+ size 5069051
sparse_linear.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b753b82fa558b68bd5b49f9bee80eef9b60df0690648e4734970033a9c0d606
3
+ size 3516
special_tokens_map.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "cls_token": {
10
+ "content": "<s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "eos_token": {
17
+ "content": "</s>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "mask_token": {
24
+ "content": "<mask>",
25
+ "lstrip": true,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ },
30
+ "pad_token": {
31
+ "content": "<pad>",
32
+ "lstrip": false,
33
+ "normalized": false,
34
+ "rstrip": false,
35
+ "single_word": false
36
+ },
37
+ "sep_token": {
38
+ "content": "</s>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false
43
+ },
44
+ "unk_token": {
45
+ "content": "<unk>",
46
+ "lstrip": false,
47
+ "normalized": false,
48
+ "rstrip": false,
49
+ "single_word": false
50
+ }
51
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69564b696052886ed0ac63fa393e928384e0f8caada38c1f4864a9bfbf379c15
3
+ size 17098273
tokenizer_config.json ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "<s>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "1": {
12
+ "content": "<pad>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "2": {
20
+ "content": "</s>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "3": {
28
+ "content": "<unk>",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "250001": {
36
+ "content": "<mask>",
37
+ "lstrip": true,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "bos_token": "<s>",
45
+ "clean_up_tokenization_spaces": true,
46
+ "cls_token": "<s>",
47
+ "eos_token": "</s>",
48
+ "mask_token": "<mask>",
49
+ "model_max_length": 8192,
50
+ "pad_token": "<pad>",
51
+ "sep_token": "</s>",
52
+ "sp_model_kwargs": {},
53
+ "tokenizer_class": "XLMRobertaTokenizer",
54
+ "unk_token": "<unk>"
55
+ }
trainer_state.json ADDED
@@ -0,0 +1,3010 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 500,
6
+ "global_step": 8546,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0,
13
+ "grad_norm": 11.189342496234838,
14
+ "learning_rate": 4.2000000000000006e-07,
15
+ "loss": 0.6334,
16
+ "step": 20
17
+ },
18
+ {
19
+ "epoch": 0.0,
20
+ "grad_norm": 12.102946252189938,
21
+ "learning_rate": 6.515789473684211e-07,
22
+ "loss": 0.6334,
23
+ "step": 40
24
+ },
25
+ {
26
+ "epoch": 0.01,
27
+ "grad_norm": 9.23868714422811,
28
+ "learning_rate": 8.831578947368421e-07,
29
+ "loss": 0.5919,
30
+ "step": 60
31
+ },
32
+ {
33
+ "epoch": 0.01,
34
+ "grad_norm": 10.163427827153294,
35
+ "learning_rate": 1.1147368421052632e-06,
36
+ "loss": 0.591,
37
+ "step": 80
38
+ },
39
+ {
40
+ "epoch": 0.01,
41
+ "grad_norm": 7.703301789458378,
42
+ "learning_rate": 1.3463157894736842e-06,
43
+ "loss": 0.5663,
44
+ "step": 100
45
+ },
46
+ {
47
+ "epoch": 0.01,
48
+ "grad_norm": 10.292384648986614,
49
+ "learning_rate": 1.5778947368421053e-06,
50
+ "loss": 0.5649,
51
+ "step": 120
52
+ },
53
+ {
54
+ "epoch": 0.02,
55
+ "grad_norm": 9.622475032798782,
56
+ "learning_rate": 1.8094736842105263e-06,
57
+ "loss": 0.5215,
58
+ "step": 140
59
+ },
60
+ {
61
+ "epoch": 0.02,
62
+ "grad_norm": 12.80974448742081,
63
+ "learning_rate": 2.0410526315789475e-06,
64
+ "loss": 0.5166,
65
+ "step": 160
66
+ },
67
+ {
68
+ "epoch": 0.02,
69
+ "grad_norm": 7.2494201428270495,
70
+ "learning_rate": 2.2726315789473687e-06,
71
+ "loss": 0.4708,
72
+ "step": 180
73
+ },
74
+ {
75
+ "epoch": 0.02,
76
+ "grad_norm": 7.525527996410342,
77
+ "learning_rate": 2.5042105263157895e-06,
78
+ "loss": 0.4809,
79
+ "step": 200
80
+ },
81
+ {
82
+ "epoch": 0.03,
83
+ "grad_norm": 9.553244905277726,
84
+ "learning_rate": 2.735789473684211e-06,
85
+ "loss": 0.4149,
86
+ "step": 220
87
+ },
88
+ {
89
+ "epoch": 0.03,
90
+ "grad_norm": 10.08934392514418,
91
+ "learning_rate": 2.967368421052632e-06,
92
+ "loss": 0.4709,
93
+ "step": 240
94
+ },
95
+ {
96
+ "epoch": 0.03,
97
+ "grad_norm": 8.77781615537256,
98
+ "learning_rate": 3.198947368421053e-06,
99
+ "loss": 0.4706,
100
+ "step": 260
101
+ },
102
+ {
103
+ "epoch": 0.03,
104
+ "grad_norm": 8.842231539252408,
105
+ "learning_rate": 3.430526315789474e-06,
106
+ "loss": 0.4033,
107
+ "step": 280
108
+ },
109
+ {
110
+ "epoch": 0.04,
111
+ "grad_norm": 10.007575408247385,
112
+ "learning_rate": 3.662105263157895e-06,
113
+ "loss": 0.4411,
114
+ "step": 300
115
+ },
116
+ {
117
+ "epoch": 0.04,
118
+ "grad_norm": 8.652379461692668,
119
+ "learning_rate": 3.893684210526316e-06,
120
+ "loss": 0.4011,
121
+ "step": 320
122
+ },
123
+ {
124
+ "epoch": 0.04,
125
+ "grad_norm": 9.03023367578369,
126
+ "learning_rate": 4.125263157894737e-06,
127
+ "loss": 0.4252,
128
+ "step": 340
129
+ },
130
+ {
131
+ "epoch": 0.04,
132
+ "grad_norm": 8.141636176102654,
133
+ "learning_rate": 4.356842105263159e-06,
134
+ "loss": 0.4364,
135
+ "step": 360
136
+ },
137
+ {
138
+ "epoch": 0.04,
139
+ "grad_norm": 8.361423248626743,
140
+ "learning_rate": 4.58842105263158e-06,
141
+ "loss": 0.4193,
142
+ "step": 380
143
+ },
144
+ {
145
+ "epoch": 0.05,
146
+ "grad_norm": 9.373028764113052,
147
+ "learning_rate": 4.8200000000000004e-06,
148
+ "loss": 0.4078,
149
+ "step": 400
150
+ },
151
+ {
152
+ "epoch": 0.05,
153
+ "grad_norm": 8.799844515033747,
154
+ "learning_rate": 5.051578947368422e-06,
155
+ "loss": 0.4291,
156
+ "step": 420
157
+ },
158
+ {
159
+ "epoch": 0.05,
160
+ "grad_norm": 8.642174028335415,
161
+ "learning_rate": 5.283157894736842e-06,
162
+ "loss": 0.382,
163
+ "step": 440
164
+ },
165
+ {
166
+ "epoch": 0.05,
167
+ "grad_norm": 9.451813409785364,
168
+ "learning_rate": 5.514736842105264e-06,
169
+ "loss": 0.4089,
170
+ "step": 460
171
+ },
172
+ {
173
+ "epoch": 0.06,
174
+ "grad_norm": 11.329024075313082,
175
+ "learning_rate": 5.7463157894736845e-06,
176
+ "loss": 0.4015,
177
+ "step": 480
178
+ },
179
+ {
180
+ "epoch": 0.06,
181
+ "grad_norm": 7.271736970582471,
182
+ "learning_rate": 5.977894736842106e-06,
183
+ "loss": 0.4028,
184
+ "step": 500
185
+ },
186
+ {
187
+ "epoch": 0.06,
188
+ "grad_norm": 10.856360235187129,
189
+ "learning_rate": 6.209473684210528e-06,
190
+ "loss": 0.3814,
191
+ "step": 520
192
+ },
193
+ {
194
+ "epoch": 0.06,
195
+ "grad_norm": 10.40735773829138,
196
+ "learning_rate": 6.441052631578948e-06,
197
+ "loss": 0.3868,
198
+ "step": 540
199
+ },
200
+ {
201
+ "epoch": 0.07,
202
+ "grad_norm": 7.176143531188922,
203
+ "learning_rate": 6.672631578947369e-06,
204
+ "loss": 0.4089,
205
+ "step": 560
206
+ },
207
+ {
208
+ "epoch": 0.07,
209
+ "grad_norm": 9.628282656024224,
210
+ "learning_rate": 6.90421052631579e-06,
211
+ "loss": 0.3771,
212
+ "step": 580
213
+ },
214
+ {
215
+ "epoch": 0.07,
216
+ "grad_norm": 11.199839686200068,
217
+ "learning_rate": 7.135789473684212e-06,
218
+ "loss": 0.383,
219
+ "step": 600
220
+ },
221
+ {
222
+ "epoch": 0.07,
223
+ "grad_norm": 8.042449860326105,
224
+ "learning_rate": 7.367368421052632e-06,
225
+ "loss": 0.3875,
226
+ "step": 620
227
+ },
228
+ {
229
+ "epoch": 0.07,
230
+ "grad_norm": 6.256777636127039,
231
+ "learning_rate": 7.598947368421053e-06,
232
+ "loss": 0.399,
233
+ "step": 640
234
+ },
235
+ {
236
+ "epoch": 0.08,
237
+ "grad_norm": 12.140400652487944,
238
+ "learning_rate": 7.830526315789473e-06,
239
+ "loss": 0.399,
240
+ "step": 660
241
+ },
242
+ {
243
+ "epoch": 0.08,
244
+ "grad_norm": 9.31740323216306,
245
+ "learning_rate": 8.062105263157896e-06,
246
+ "loss": 0.383,
247
+ "step": 680
248
+ },
249
+ {
250
+ "epoch": 0.08,
251
+ "grad_norm": 11.518507949844242,
252
+ "learning_rate": 8.293684210526317e-06,
253
+ "loss": 0.3673,
254
+ "step": 700
255
+ },
256
+ {
257
+ "epoch": 0.08,
258
+ "grad_norm": 12.23485079993539,
259
+ "learning_rate": 8.525263157894737e-06,
260
+ "loss": 0.4105,
261
+ "step": 720
262
+ },
263
+ {
264
+ "epoch": 0.09,
265
+ "grad_norm": 7.506176757455166,
266
+ "learning_rate": 8.756842105263158e-06,
267
+ "loss": 0.425,
268
+ "step": 740
269
+ },
270
+ {
271
+ "epoch": 0.09,
272
+ "grad_norm": 28.3327677352401,
273
+ "learning_rate": 8.988421052631579e-06,
274
+ "loss": 0.3874,
275
+ "step": 760
276
+ },
277
+ {
278
+ "epoch": 0.09,
279
+ "grad_norm": 10.006030362522525,
280
+ "learning_rate": 9.220000000000002e-06,
281
+ "loss": 0.4055,
282
+ "step": 780
283
+ },
284
+ {
285
+ "epoch": 0.09,
286
+ "grad_norm": 10.010446813711976,
287
+ "learning_rate": 9.451578947368422e-06,
288
+ "loss": 0.3599,
289
+ "step": 800
290
+ },
291
+ {
292
+ "epoch": 0.1,
293
+ "grad_norm": 7.043688000572858,
294
+ "learning_rate": 9.683157894736843e-06,
295
+ "loss": 0.3624,
296
+ "step": 820
297
+ },
298
+ {
299
+ "epoch": 0.1,
300
+ "grad_norm": 9.607021831412599,
301
+ "learning_rate": 9.914736842105264e-06,
302
+ "loss": 0.3952,
303
+ "step": 840
304
+ },
305
+ {
306
+ "epoch": 0.1,
307
+ "grad_norm": 9.147395412600936,
308
+ "learning_rate": 1.0146315789473686e-05,
309
+ "loss": 0.3819,
310
+ "step": 860
311
+ },
312
+ {
313
+ "epoch": 0.1,
314
+ "grad_norm": 7.993375419567414,
315
+ "learning_rate": 1.0377894736842107e-05,
316
+ "loss": 0.3678,
317
+ "step": 880
318
+ },
319
+ {
320
+ "epoch": 0.11,
321
+ "grad_norm": 7.52609874523302,
322
+ "learning_rate": 1.0609473684210528e-05,
323
+ "loss": 0.375,
324
+ "step": 900
325
+ },
326
+ {
327
+ "epoch": 0.11,
328
+ "grad_norm": 9.52738539095041,
329
+ "learning_rate": 1.084105263157895e-05,
330
+ "loss": 0.3681,
331
+ "step": 920
332
+ },
333
+ {
334
+ "epoch": 0.11,
335
+ "grad_norm": 11.370588516129013,
336
+ "learning_rate": 1.107263157894737e-05,
337
+ "loss": 0.3911,
338
+ "step": 940
339
+ },
340
+ {
341
+ "epoch": 0.11,
342
+ "grad_norm": 9.391054368953709,
343
+ "learning_rate": 1.130421052631579e-05,
344
+ "loss": 0.329,
345
+ "step": 960
346
+ },
347
+ {
348
+ "epoch": 0.11,
349
+ "grad_norm": 7.997549873904863,
350
+ "learning_rate": 1.1535789473684211e-05,
351
+ "loss": 0.359,
352
+ "step": 980
353
+ },
354
+ {
355
+ "epoch": 0.12,
356
+ "grad_norm": 10.755576173466588,
357
+ "learning_rate": 1.1767368421052634e-05,
358
+ "loss": 0.393,
359
+ "step": 1000
360
+ },
361
+ {
362
+ "epoch": 0.12,
363
+ "grad_norm": 6.458117606549104,
364
+ "learning_rate": 1.1998947368421054e-05,
365
+ "loss": 0.363,
366
+ "step": 1020
367
+ },
368
+ {
369
+ "epoch": 0.12,
370
+ "grad_norm": 9.937055469862932,
371
+ "learning_rate": 1.2230526315789475e-05,
372
+ "loss": 0.3766,
373
+ "step": 1040
374
+ },
375
+ {
376
+ "epoch": 0.12,
377
+ "grad_norm": 7.485816768913991,
378
+ "learning_rate": 1.2462105263157898e-05,
379
+ "loss": 0.3845,
380
+ "step": 1060
381
+ },
382
+ {
383
+ "epoch": 0.13,
384
+ "grad_norm": 8.01333127284645,
385
+ "learning_rate": 1.2693684210526319e-05,
386
+ "loss": 0.3617,
387
+ "step": 1080
388
+ },
389
+ {
390
+ "epoch": 0.13,
391
+ "grad_norm": 8.911089094498925,
392
+ "learning_rate": 1.292526315789474e-05,
393
+ "loss": 0.3599,
394
+ "step": 1100
395
+ },
396
+ {
397
+ "epoch": 0.13,
398
+ "grad_norm": 8.32079434857686,
399
+ "learning_rate": 1.3156842105263158e-05,
400
+ "loss": 0.3722,
401
+ "step": 1120
402
+ },
403
+ {
404
+ "epoch": 0.13,
405
+ "grad_norm": 9.806231592184165,
406
+ "learning_rate": 1.3388421052631581e-05,
407
+ "loss": 0.3755,
408
+ "step": 1140
409
+ },
410
+ {
411
+ "epoch": 0.14,
412
+ "grad_norm": 8.798076540805445,
413
+ "learning_rate": 1.3620000000000002e-05,
414
+ "loss": 0.3616,
415
+ "step": 1160
416
+ },
417
+ {
418
+ "epoch": 0.14,
419
+ "grad_norm": 8.343613755617508,
420
+ "learning_rate": 1.3851578947368423e-05,
421
+ "loss": 0.3631,
422
+ "step": 1180
423
+ },
424
+ {
425
+ "epoch": 0.14,
426
+ "grad_norm": 8.35814690526204,
427
+ "learning_rate": 1.4083157894736843e-05,
428
+ "loss": 0.3638,
429
+ "step": 1200
430
+ },
431
+ {
432
+ "epoch": 0.14,
433
+ "grad_norm": 7.706542458286787,
434
+ "learning_rate": 1.4314736842105266e-05,
435
+ "loss": 0.375,
436
+ "step": 1220
437
+ },
438
+ {
439
+ "epoch": 0.15,
440
+ "grad_norm": 10.89212756963322,
441
+ "learning_rate": 1.4546315789473687e-05,
442
+ "loss": 0.4049,
443
+ "step": 1240
444
+ },
445
+ {
446
+ "epoch": 0.15,
447
+ "grad_norm": 11.46198790559678,
448
+ "learning_rate": 1.4777894736842107e-05,
449
+ "loss": 0.332,
450
+ "step": 1260
451
+ },
452
+ {
453
+ "epoch": 0.15,
454
+ "grad_norm": 8.365156167551179,
455
+ "learning_rate": 1.500947368421053e-05,
456
+ "loss": 0.3586,
457
+ "step": 1280
458
+ },
459
+ {
460
+ "epoch": 0.15,
461
+ "grad_norm": 7.841036992292662,
462
+ "learning_rate": 1.5241052631578949e-05,
463
+ "loss": 0.368,
464
+ "step": 1300
465
+ },
466
+ {
467
+ "epoch": 0.15,
468
+ "grad_norm": 9.84275168548854,
469
+ "learning_rate": 1.547263157894737e-05,
470
+ "loss": 0.3767,
471
+ "step": 1320
472
+ },
473
+ {
474
+ "epoch": 0.16,
475
+ "grad_norm": 10.506254240871392,
476
+ "learning_rate": 1.5704210526315792e-05,
477
+ "loss": 0.378,
478
+ "step": 1340
479
+ },
480
+ {
481
+ "epoch": 0.16,
482
+ "grad_norm": 7.844581104033147,
483
+ "learning_rate": 1.5935789473684213e-05,
484
+ "loss": 0.3453,
485
+ "step": 1360
486
+ },
487
+ {
488
+ "epoch": 0.16,
489
+ "grad_norm": 8.193758138016792,
490
+ "learning_rate": 1.6167368421052634e-05,
491
+ "loss": 0.3752,
492
+ "step": 1380
493
+ },
494
+ {
495
+ "epoch": 0.16,
496
+ "grad_norm": 7.475620318034102,
497
+ "learning_rate": 1.6398947368421055e-05,
498
+ "loss": 0.3642,
499
+ "step": 1400
500
+ },
501
+ {
502
+ "epoch": 0.17,
503
+ "grad_norm": 7.4964850772334355,
504
+ "learning_rate": 1.6630526315789475e-05,
505
+ "loss": 0.3377,
506
+ "step": 1420
507
+ },
508
+ {
509
+ "epoch": 0.17,
510
+ "grad_norm": 7.903753436810423,
511
+ "learning_rate": 1.6862105263157896e-05,
512
+ "loss": 0.3586,
513
+ "step": 1440
514
+ },
515
+ {
516
+ "epoch": 0.17,
517
+ "grad_norm": 10.49594728277505,
518
+ "learning_rate": 1.7093684210526317e-05,
519
+ "loss": 0.3468,
520
+ "step": 1460
521
+ },
522
+ {
523
+ "epoch": 0.17,
524
+ "grad_norm": 8.027713457118994,
525
+ "learning_rate": 1.7325263157894738e-05,
526
+ "loss": 0.3491,
527
+ "step": 1480
528
+ },
529
+ {
530
+ "epoch": 0.18,
531
+ "grad_norm": 7.2748681125287415,
532
+ "learning_rate": 1.755684210526316e-05,
533
+ "loss": 0.3581,
534
+ "step": 1500
535
+ },
536
+ {
537
+ "epoch": 0.18,
538
+ "grad_norm": 6.077326940397276,
539
+ "learning_rate": 1.7788421052631583e-05,
540
+ "loss": 0.3356,
541
+ "step": 1520
542
+ },
543
+ {
544
+ "epoch": 0.18,
545
+ "grad_norm": 6.724980645701413,
546
+ "learning_rate": 1.802e-05,
547
+ "loss": 0.3656,
548
+ "step": 1540
549
+ },
550
+ {
551
+ "epoch": 0.18,
552
+ "grad_norm": 8.592315221136559,
553
+ "learning_rate": 1.825157894736842e-05,
554
+ "loss": 0.3467,
555
+ "step": 1560
556
+ },
557
+ {
558
+ "epoch": 0.18,
559
+ "grad_norm": 9.036253074638084,
560
+ "learning_rate": 1.8483157894736845e-05,
561
+ "loss": 0.3414,
562
+ "step": 1580
563
+ },
564
+ {
565
+ "epoch": 0.19,
566
+ "grad_norm": 11.349080832480995,
567
+ "learning_rate": 1.8714736842105266e-05,
568
+ "loss": 0.3655,
569
+ "step": 1600
570
+ },
571
+ {
572
+ "epoch": 0.19,
573
+ "grad_norm": 9.361475025905081,
574
+ "learning_rate": 1.8946315789473687e-05,
575
+ "loss": 0.3564,
576
+ "step": 1620
577
+ },
578
+ {
579
+ "epoch": 0.19,
580
+ "grad_norm": 6.53643855564322,
581
+ "learning_rate": 1.9177894736842108e-05,
582
+ "loss": 0.3714,
583
+ "step": 1640
584
+ },
585
+ {
586
+ "epoch": 0.19,
587
+ "grad_norm": 6.65699298275954,
588
+ "learning_rate": 1.940947368421053e-05,
589
+ "loss": 0.3546,
590
+ "step": 1660
591
+ },
592
+ {
593
+ "epoch": 0.2,
594
+ "grad_norm": 9.856127510930383,
595
+ "learning_rate": 1.964105263157895e-05,
596
+ "loss": 0.3308,
597
+ "step": 1680
598
+ },
599
+ {
600
+ "epoch": 0.2,
601
+ "grad_norm": 8.532146476622113,
602
+ "learning_rate": 1.987263157894737e-05,
603
+ "loss": 0.366,
604
+ "step": 1700
605
+ },
606
+ {
607
+ "epoch": 0.2,
608
+ "grad_norm": 8.327666937449752,
609
+ "learning_rate": 1.999999532098205e-05,
610
+ "loss": 0.3247,
611
+ "step": 1720
612
+ },
613
+ {
614
+ "epoch": 0.2,
615
+ "grad_norm": 10.876035443701728,
616
+ "learning_rate": 1.99999578888647e-05,
617
+ "loss": 0.3543,
618
+ "step": 1740
619
+ },
620
+ {
621
+ "epoch": 0.21,
622
+ "grad_norm": 8.263711805253918,
623
+ "learning_rate": 1.999988302477013e-05,
624
+ "loss": 0.3664,
625
+ "step": 1760
626
+ },
627
+ {
628
+ "epoch": 0.21,
629
+ "grad_norm": 7.575419338189409,
630
+ "learning_rate": 1.9999770728978607e-05,
631
+ "loss": 0.345,
632
+ "step": 1780
633
+ },
634
+ {
635
+ "epoch": 0.21,
636
+ "grad_norm": 7.814279338388738,
637
+ "learning_rate": 1.9999621001910508e-05,
638
+ "loss": 0.3573,
639
+ "step": 1800
640
+ },
641
+ {
642
+ "epoch": 0.21,
643
+ "grad_norm": 13.040107479471628,
644
+ "learning_rate": 1.999943384412636e-05,
645
+ "loss": 0.3344,
646
+ "step": 1820
647
+ },
648
+ {
649
+ "epoch": 0.22,
650
+ "grad_norm": 7.535441012066771,
651
+ "learning_rate": 1.99992092563268e-05,
652
+ "loss": 0.3552,
653
+ "step": 1840
654
+ },
655
+ {
656
+ "epoch": 0.22,
657
+ "grad_norm": 8.059491683038381,
658
+ "learning_rate": 1.9998947239352593e-05,
659
+ "loss": 0.3442,
660
+ "step": 1860
661
+ },
662
+ {
663
+ "epoch": 0.22,
664
+ "grad_norm": 6.700609139607897,
665
+ "learning_rate": 1.999864779418462e-05,
666
+ "loss": 0.3584,
667
+ "step": 1880
668
+ },
669
+ {
670
+ "epoch": 0.22,
671
+ "grad_norm": 6.831136707088909,
672
+ "learning_rate": 1.999831092194389e-05,
673
+ "loss": 0.3341,
674
+ "step": 1900
675
+ },
676
+ {
677
+ "epoch": 0.22,
678
+ "grad_norm": 10.30262237054047,
679
+ "learning_rate": 1.9997936623891498e-05,
680
+ "loss": 0.3704,
681
+ "step": 1920
682
+ },
683
+ {
684
+ "epoch": 0.23,
685
+ "grad_norm": 6.703691792033561,
686
+ "learning_rate": 1.9997524901428677e-05,
687
+ "loss": 0.3769,
688
+ "step": 1940
689
+ },
690
+ {
691
+ "epoch": 0.23,
692
+ "grad_norm": 7.251026212477321,
693
+ "learning_rate": 1.9997075756096734e-05,
694
+ "loss": 0.3234,
695
+ "step": 1960
696
+ },
697
+ {
698
+ "epoch": 0.23,
699
+ "grad_norm": 6.14471190830255,
700
+ "learning_rate": 1.9996589189577094e-05,
701
+ "loss": 0.3443,
702
+ "step": 1980
703
+ },
704
+ {
705
+ "epoch": 0.23,
706
+ "grad_norm": 7.254885638733633,
707
+ "learning_rate": 1.999606520369125e-05,
708
+ "loss": 0.3178,
709
+ "step": 2000
710
+ },
711
+ {
712
+ "epoch": 0.24,
713
+ "grad_norm": 7.330977870771951,
714
+ "learning_rate": 1.9995503800400805e-05,
715
+ "loss": 0.3309,
716
+ "step": 2020
717
+ },
718
+ {
719
+ "epoch": 0.24,
720
+ "grad_norm": 7.291895603037368,
721
+ "learning_rate": 1.9994904981807407e-05,
722
+ "loss": 0.3372,
723
+ "step": 2040
724
+ },
725
+ {
726
+ "epoch": 0.24,
727
+ "grad_norm": 12.772894555548028,
728
+ "learning_rate": 1.9994268750152793e-05,
729
+ "loss": 0.3747,
730
+ "step": 2060
731
+ },
732
+ {
733
+ "epoch": 0.24,
734
+ "grad_norm": 9.497788021161657,
735
+ "learning_rate": 1.999359510781875e-05,
736
+ "loss": 0.3529,
737
+ "step": 2080
738
+ },
739
+ {
740
+ "epoch": 0.25,
741
+ "grad_norm": 6.581748417513793,
742
+ "learning_rate": 1.9992884057327118e-05,
743
+ "loss": 0.3395,
744
+ "step": 2100
745
+ },
746
+ {
747
+ "epoch": 0.25,
748
+ "grad_norm": 9.523030924078585,
749
+ "learning_rate": 1.9992135601339776e-05,
750
+ "loss": 0.3469,
751
+ "step": 2120
752
+ },
753
+ {
754
+ "epoch": 0.25,
755
+ "grad_norm": 17.640537636796516,
756
+ "learning_rate": 1.999134974265864e-05,
757
+ "loss": 0.3365,
758
+ "step": 2140
759
+ },
760
+ {
761
+ "epoch": 0.25,
762
+ "grad_norm": 5.4385511763432985,
763
+ "learning_rate": 1.999052648422563e-05,
764
+ "loss": 0.3845,
765
+ "step": 2160
766
+ },
767
+ {
768
+ "epoch": 0.26,
769
+ "grad_norm": 7.585921156892153,
770
+ "learning_rate": 1.9989665829122704e-05,
771
+ "loss": 0.3464,
772
+ "step": 2180
773
+ },
774
+ {
775
+ "epoch": 0.26,
776
+ "grad_norm": 12.163408304702841,
777
+ "learning_rate": 1.9988767780571778e-05,
778
+ "loss": 0.3127,
779
+ "step": 2200
780
+ },
781
+ {
782
+ "epoch": 0.26,
783
+ "grad_norm": 8.246994395835456,
784
+ "learning_rate": 1.9987832341934793e-05,
785
+ "loss": 0.3424,
786
+ "step": 2220
787
+ },
788
+ {
789
+ "epoch": 0.26,
790
+ "grad_norm": 9.628684787640838,
791
+ "learning_rate": 1.9986859516713636e-05,
792
+ "loss": 0.3472,
793
+ "step": 2240
794
+ },
795
+ {
796
+ "epoch": 0.26,
797
+ "grad_norm": 10.912326639203211,
798
+ "learning_rate": 1.9985849308550167e-05,
799
+ "loss": 0.3326,
800
+ "step": 2260
801
+ },
802
+ {
803
+ "epoch": 0.27,
804
+ "grad_norm": 6.419410251137725,
805
+ "learning_rate": 1.9984801721226188e-05,
806
+ "loss": 0.3297,
807
+ "step": 2280
808
+ },
809
+ {
810
+ "epoch": 0.27,
811
+ "grad_norm": 7.798212276898864,
812
+ "learning_rate": 1.998371675866344e-05,
813
+ "loss": 0.3389,
814
+ "step": 2300
815
+ },
816
+ {
817
+ "epoch": 0.27,
818
+ "grad_norm": 7.811483576458649,
819
+ "learning_rate": 1.9982594424923566e-05,
820
+ "loss": 0.332,
821
+ "step": 2320
822
+ },
823
+ {
824
+ "epoch": 0.27,
825
+ "grad_norm": 6.967736687994363,
826
+ "learning_rate": 1.9981434724208126e-05,
827
+ "loss": 0.3099,
828
+ "step": 2340
829
+ },
830
+ {
831
+ "epoch": 0.28,
832
+ "grad_norm": 9.635115819134949,
833
+ "learning_rate": 1.9980237660858565e-05,
834
+ "loss": 0.3285,
835
+ "step": 2360
836
+ },
837
+ {
838
+ "epoch": 0.28,
839
+ "grad_norm": 6.720069081541748,
840
+ "learning_rate": 1.997900323935619e-05,
841
+ "loss": 0.343,
842
+ "step": 2380
843
+ },
844
+ {
845
+ "epoch": 0.28,
846
+ "grad_norm": 10.300889386045979,
847
+ "learning_rate": 1.9977731464322162e-05,
848
+ "loss": 0.3667,
849
+ "step": 2400
850
+ },
851
+ {
852
+ "epoch": 0.28,
853
+ "grad_norm": 7.888854177455323,
854
+ "learning_rate": 1.9976422340517495e-05,
855
+ "loss": 0.3397,
856
+ "step": 2420
857
+ },
858
+ {
859
+ "epoch": 0.29,
860
+ "grad_norm": 9.643051064776678,
861
+ "learning_rate": 1.9975075872843e-05,
862
+ "loss": 0.3371,
863
+ "step": 2440
864
+ },
865
+ {
866
+ "epoch": 0.29,
867
+ "grad_norm": 6.166262467342929,
868
+ "learning_rate": 1.9973692066339298e-05,
869
+ "loss": 0.3424,
870
+ "step": 2460
871
+ },
872
+ {
873
+ "epoch": 0.29,
874
+ "grad_norm": 7.520601841401043,
875
+ "learning_rate": 1.9972270926186794e-05,
876
+ "loss": 0.3083,
877
+ "step": 2480
878
+ },
879
+ {
880
+ "epoch": 0.29,
881
+ "grad_norm": 6.095472200047568,
882
+ "learning_rate": 1.9970812457705642e-05,
883
+ "loss": 0.3512,
884
+ "step": 2500
885
+ },
886
+ {
887
+ "epoch": 0.29,
888
+ "grad_norm": 7.259492119293066,
889
+ "learning_rate": 1.9969316666355757e-05,
890
+ "loss": 0.3309,
891
+ "step": 2520
892
+ },
893
+ {
894
+ "epoch": 0.3,
895
+ "grad_norm": 8.355978245335697,
896
+ "learning_rate": 1.996778355773676e-05,
897
+ "loss": 0.3296,
898
+ "step": 2540
899
+ },
900
+ {
901
+ "epoch": 0.3,
902
+ "grad_norm": 9.612502269967832,
903
+ "learning_rate": 1.996621313758798e-05,
904
+ "loss": 0.3391,
905
+ "step": 2560
906
+ },
907
+ {
908
+ "epoch": 0.3,
909
+ "grad_norm": 8.148285817739712,
910
+ "learning_rate": 1.9964605411788424e-05,
911
+ "loss": 0.3416,
912
+ "step": 2580
913
+ },
914
+ {
915
+ "epoch": 0.3,
916
+ "grad_norm": 7.645206131186397,
917
+ "learning_rate": 1.9962960386356748e-05,
918
+ "loss": 0.352,
919
+ "step": 2600
920
+ },
921
+ {
922
+ "epoch": 0.31,
923
+ "grad_norm": 5.406075006199623,
924
+ "learning_rate": 1.9961278067451258e-05,
925
+ "loss": 0.3194,
926
+ "step": 2620
927
+ },
928
+ {
929
+ "epoch": 0.31,
930
+ "grad_norm": 7.23674984010716,
931
+ "learning_rate": 1.995955846136986e-05,
932
+ "loss": 0.3216,
933
+ "step": 2640
934
+ },
935
+ {
936
+ "epoch": 0.31,
937
+ "grad_norm": 8.044071872202302,
938
+ "learning_rate": 1.9957801574550044e-05,
939
+ "loss": 0.321,
940
+ "step": 2660
941
+ },
942
+ {
943
+ "epoch": 0.31,
944
+ "grad_norm": 7.2624583573058485,
945
+ "learning_rate": 1.9956007413568874e-05,
946
+ "loss": 0.3352,
947
+ "step": 2680
948
+ },
949
+ {
950
+ "epoch": 0.32,
951
+ "grad_norm": 9.277277549105909,
952
+ "learning_rate": 1.9954175985142955e-05,
953
+ "loss": 0.3391,
954
+ "step": 2700
955
+ },
956
+ {
957
+ "epoch": 0.32,
958
+ "grad_norm": 10.975275864316648,
959
+ "learning_rate": 1.9952307296128395e-05,
960
+ "loss": 0.3298,
961
+ "step": 2720
962
+ },
963
+ {
964
+ "epoch": 0.32,
965
+ "grad_norm": 11.416959735607907,
966
+ "learning_rate": 1.9950401353520795e-05,
967
+ "loss": 0.3382,
968
+ "step": 2740
969
+ },
970
+ {
971
+ "epoch": 0.32,
972
+ "grad_norm": 7.384993942351236,
973
+ "learning_rate": 1.994845816445522e-05,
974
+ "loss": 0.3422,
975
+ "step": 2760
976
+ },
977
+ {
978
+ "epoch": 0.33,
979
+ "grad_norm": 6.775839902047113,
980
+ "learning_rate": 1.9946477736206167e-05,
981
+ "loss": 0.3694,
982
+ "step": 2780
983
+ },
984
+ {
985
+ "epoch": 0.33,
986
+ "grad_norm": 8.246414330131564,
987
+ "learning_rate": 1.9944460076187546e-05,
988
+ "loss": 0.3458,
989
+ "step": 2800
990
+ },
991
+ {
992
+ "epoch": 0.33,
993
+ "grad_norm": 7.152660285783071,
994
+ "learning_rate": 1.9942405191952646e-05,
995
+ "loss": 0.3426,
996
+ "step": 2820
997
+ },
998
+ {
999
+ "epoch": 0.33,
1000
+ "grad_norm": 6.9408668390055315,
1001
+ "learning_rate": 1.9940313091194096e-05,
1002
+ "loss": 0.3097,
1003
+ "step": 2840
1004
+ },
1005
+ {
1006
+ "epoch": 0.33,
1007
+ "grad_norm": 5.797513942002627,
1008
+ "learning_rate": 1.9938183781743873e-05,
1009
+ "loss": 0.3102,
1010
+ "step": 2860
1011
+ },
1012
+ {
1013
+ "epoch": 0.34,
1014
+ "grad_norm": 7.361935180345544,
1015
+ "learning_rate": 1.9936017271573223e-05,
1016
+ "loss": 0.3102,
1017
+ "step": 2880
1018
+ },
1019
+ {
1020
+ "epoch": 0.34,
1021
+ "grad_norm": 7.4796257480282,
1022
+ "learning_rate": 1.9933813568792678e-05,
1023
+ "loss": 0.3256,
1024
+ "step": 2900
1025
+ },
1026
+ {
1027
+ "epoch": 0.34,
1028
+ "grad_norm": 5.500192812227557,
1029
+ "learning_rate": 1.9931572681651984e-05,
1030
+ "loss": 0.3107,
1031
+ "step": 2920
1032
+ },
1033
+ {
1034
+ "epoch": 0.34,
1035
+ "grad_norm": 13.84776864376109,
1036
+ "learning_rate": 1.9929294618540103e-05,
1037
+ "loss": 0.3062,
1038
+ "step": 2940
1039
+ },
1040
+ {
1041
+ "epoch": 0.35,
1042
+ "grad_norm": 7.937410128828084,
1043
+ "learning_rate": 1.992697938798516e-05,
1044
+ "loss": 0.3293,
1045
+ "step": 2960
1046
+ },
1047
+ {
1048
+ "epoch": 0.35,
1049
+ "grad_norm": 5.447364190742796,
1050
+ "learning_rate": 1.9924626998654433e-05,
1051
+ "loss": 0.3318,
1052
+ "step": 2980
1053
+ },
1054
+ {
1055
+ "epoch": 0.35,
1056
+ "grad_norm": 7.825239369201605,
1057
+ "learning_rate": 1.9922237459354287e-05,
1058
+ "loss": 0.3238,
1059
+ "step": 3000
1060
+ },
1061
+ {
1062
+ "epoch": 0.35,
1063
+ "grad_norm": 6.466926354506395,
1064
+ "learning_rate": 1.9919810779030177e-05,
1065
+ "loss": 0.3136,
1066
+ "step": 3020
1067
+ },
1068
+ {
1069
+ "epoch": 0.36,
1070
+ "grad_norm": 10.115420483963225,
1071
+ "learning_rate": 1.9917346966766596e-05,
1072
+ "loss": 0.3329,
1073
+ "step": 3040
1074
+ },
1075
+ {
1076
+ "epoch": 0.36,
1077
+ "grad_norm": 6.08100630613984,
1078
+ "learning_rate": 1.991484603178704e-05,
1079
+ "loss": 0.2999,
1080
+ "step": 3060
1081
+ },
1082
+ {
1083
+ "epoch": 0.36,
1084
+ "grad_norm": 6.153474908414458,
1085
+ "learning_rate": 1.991230798345398e-05,
1086
+ "loss": 0.3344,
1087
+ "step": 3080
1088
+ },
1089
+ {
1090
+ "epoch": 0.36,
1091
+ "grad_norm": 6.956842076140817,
1092
+ "learning_rate": 1.990973283126882e-05,
1093
+ "loss": 0.3356,
1094
+ "step": 3100
1095
+ },
1096
+ {
1097
+ "epoch": 0.37,
1098
+ "grad_norm": 7.977473013654998,
1099
+ "learning_rate": 1.990712058487187e-05,
1100
+ "loss": 0.2984,
1101
+ "step": 3120
1102
+ },
1103
+ {
1104
+ "epoch": 0.37,
1105
+ "grad_norm": 7.472067890046691,
1106
+ "learning_rate": 1.9904471254042304e-05,
1107
+ "loss": 0.308,
1108
+ "step": 3140
1109
+ },
1110
+ {
1111
+ "epoch": 0.37,
1112
+ "grad_norm": 6.98090482864238,
1113
+ "learning_rate": 1.9901784848698125e-05,
1114
+ "loss": 0.3036,
1115
+ "step": 3160
1116
+ },
1117
+ {
1118
+ "epoch": 0.37,
1119
+ "grad_norm": 7.229553593404789,
1120
+ "learning_rate": 1.989906137889613e-05,
1121
+ "loss": 0.3027,
1122
+ "step": 3180
1123
+ },
1124
+ {
1125
+ "epoch": 0.37,
1126
+ "grad_norm": 5.9700766482137295,
1127
+ "learning_rate": 1.9896300854831862e-05,
1128
+ "loss": 0.3329,
1129
+ "step": 3200
1130
+ },
1131
+ {
1132
+ "epoch": 0.38,
1133
+ "grad_norm": 7.864131148182889,
1134
+ "learning_rate": 1.989350328683959e-05,
1135
+ "loss": 0.308,
1136
+ "step": 3220
1137
+ },
1138
+ {
1139
+ "epoch": 0.38,
1140
+ "grad_norm": 20.04577205534678,
1141
+ "learning_rate": 1.989066868539225e-05,
1142
+ "loss": 0.3217,
1143
+ "step": 3240
1144
+ },
1145
+ {
1146
+ "epoch": 0.38,
1147
+ "grad_norm": 8.46751929768339,
1148
+ "learning_rate": 1.9887797061101424e-05,
1149
+ "loss": 0.3036,
1150
+ "step": 3260
1151
+ },
1152
+ {
1153
+ "epoch": 0.38,
1154
+ "grad_norm": 7.667111535917501,
1155
+ "learning_rate": 1.9884888424717296e-05,
1156
+ "loss": 0.3084,
1157
+ "step": 3280
1158
+ },
1159
+ {
1160
+ "epoch": 0.39,
1161
+ "grad_norm": 6.615069450077278,
1162
+ "learning_rate": 1.9881942787128594e-05,
1163
+ "loss": 0.3177,
1164
+ "step": 3300
1165
+ },
1166
+ {
1167
+ "epoch": 0.39,
1168
+ "grad_norm": 9.324876177990959,
1169
+ "learning_rate": 1.9878960159362572e-05,
1170
+ "loss": 0.3496,
1171
+ "step": 3320
1172
+ },
1173
+ {
1174
+ "epoch": 0.39,
1175
+ "grad_norm": 7.1979554134343635,
1176
+ "learning_rate": 1.987594055258496e-05,
1177
+ "loss": 0.3167,
1178
+ "step": 3340
1179
+ },
1180
+ {
1181
+ "epoch": 0.39,
1182
+ "grad_norm": 29.076306154493622,
1183
+ "learning_rate": 1.987288397809993e-05,
1184
+ "loss": 0.34,
1185
+ "step": 3360
1186
+ },
1187
+ {
1188
+ "epoch": 0.4,
1189
+ "grad_norm": 7.861933206527291,
1190
+ "learning_rate": 1.986979044735002e-05,
1191
+ "loss": 0.3424,
1192
+ "step": 3380
1193
+ },
1194
+ {
1195
+ "epoch": 0.4,
1196
+ "grad_norm": 5.7878723156599445,
1197
+ "learning_rate": 1.9866659971916146e-05,
1198
+ "loss": 0.3157,
1199
+ "step": 3400
1200
+ },
1201
+ {
1202
+ "epoch": 0.4,
1203
+ "grad_norm": 8.470531768859948,
1204
+ "learning_rate": 1.9863492563517513e-05,
1205
+ "loss": 0.3201,
1206
+ "step": 3420
1207
+ },
1208
+ {
1209
+ "epoch": 0.4,
1210
+ "grad_norm": 8.104846079245535,
1211
+ "learning_rate": 1.98602882340116e-05,
1212
+ "loss": 0.322,
1213
+ "step": 3440
1214
+ },
1215
+ {
1216
+ "epoch": 0.4,
1217
+ "grad_norm": 6.180289951670451,
1218
+ "learning_rate": 1.9857046995394084e-05,
1219
+ "loss": 0.2991,
1220
+ "step": 3460
1221
+ },
1222
+ {
1223
+ "epoch": 0.41,
1224
+ "grad_norm": 4.962160167060266,
1225
+ "learning_rate": 1.985376885979884e-05,
1226
+ "loss": 0.3097,
1227
+ "step": 3480
1228
+ },
1229
+ {
1230
+ "epoch": 0.41,
1231
+ "grad_norm": 6.813966925636774,
1232
+ "learning_rate": 1.9850453839497844e-05,
1233
+ "loss": 0.3294,
1234
+ "step": 3500
1235
+ },
1236
+ {
1237
+ "epoch": 0.41,
1238
+ "grad_norm": 5.902763508299491,
1239
+ "learning_rate": 1.9847101946901172e-05,
1240
+ "loss": 0.2967,
1241
+ "step": 3520
1242
+ },
1243
+ {
1244
+ "epoch": 0.41,
1245
+ "grad_norm": 7.359798475906595,
1246
+ "learning_rate": 1.984371319455693e-05,
1247
+ "loss": 0.3268,
1248
+ "step": 3540
1249
+ },
1250
+ {
1251
+ "epoch": 0.42,
1252
+ "grad_norm": 7.377881796357637,
1253
+ "learning_rate": 1.9840287595151205e-05,
1254
+ "loss": 0.3251,
1255
+ "step": 3560
1256
+ },
1257
+ {
1258
+ "epoch": 0.42,
1259
+ "grad_norm": 7.261601866242624,
1260
+ "learning_rate": 1.9836825161508032e-05,
1261
+ "loss": 0.284,
1262
+ "step": 3580
1263
+ },
1264
+ {
1265
+ "epoch": 0.42,
1266
+ "grad_norm": 6.123609423918951,
1267
+ "learning_rate": 1.983332590658934e-05,
1268
+ "loss": 0.3137,
1269
+ "step": 3600
1270
+ },
1271
+ {
1272
+ "epoch": 0.42,
1273
+ "grad_norm": 7.205290164862431,
1274
+ "learning_rate": 1.9829789843494885e-05,
1275
+ "loss": 0.3,
1276
+ "step": 3620
1277
+ },
1278
+ {
1279
+ "epoch": 0.43,
1280
+ "grad_norm": 6.0471563544848035,
1281
+ "learning_rate": 1.9826216985462245e-05,
1282
+ "loss": 0.2991,
1283
+ "step": 3640
1284
+ },
1285
+ {
1286
+ "epoch": 0.43,
1287
+ "grad_norm": 7.436589545934819,
1288
+ "learning_rate": 1.9822607345866715e-05,
1289
+ "loss": 0.3299,
1290
+ "step": 3660
1291
+ },
1292
+ {
1293
+ "epoch": 0.43,
1294
+ "grad_norm": 8.240853065109379,
1295
+ "learning_rate": 1.9818960938221308e-05,
1296
+ "loss": 0.2886,
1297
+ "step": 3680
1298
+ },
1299
+ {
1300
+ "epoch": 0.43,
1301
+ "grad_norm": 5.451034188795461,
1302
+ "learning_rate": 1.9815277776176663e-05,
1303
+ "loss": 0.3138,
1304
+ "step": 3700
1305
+ },
1306
+ {
1307
+ "epoch": 0.44,
1308
+ "grad_norm": 7.874196329434049,
1309
+ "learning_rate": 1.9811557873521027e-05,
1310
+ "loss": 0.2946,
1311
+ "step": 3720
1312
+ },
1313
+ {
1314
+ "epoch": 0.44,
1315
+ "grad_norm": 6.547528436867746,
1316
+ "learning_rate": 1.9807801244180174e-05,
1317
+ "loss": 0.2978,
1318
+ "step": 3740
1319
+ },
1320
+ {
1321
+ "epoch": 0.44,
1322
+ "grad_norm": 5.816214082411462,
1323
+ "learning_rate": 1.980400790221739e-05,
1324
+ "loss": 0.3067,
1325
+ "step": 3760
1326
+ },
1327
+ {
1328
+ "epoch": 0.44,
1329
+ "grad_norm": 6.245697982302123,
1330
+ "learning_rate": 1.9800177861833372e-05,
1331
+ "loss": 0.3191,
1332
+ "step": 3780
1333
+ },
1334
+ {
1335
+ "epoch": 0.44,
1336
+ "grad_norm": 7.932045331416635,
1337
+ "learning_rate": 1.9796311137366212e-05,
1338
+ "loss": 0.3439,
1339
+ "step": 3800
1340
+ },
1341
+ {
1342
+ "epoch": 0.45,
1343
+ "grad_norm": 6.088006858952151,
1344
+ "learning_rate": 1.9792407743291343e-05,
1345
+ "loss": 0.33,
1346
+ "step": 3820
1347
+ },
1348
+ {
1349
+ "epoch": 0.45,
1350
+ "grad_norm": 5.770262468473867,
1351
+ "learning_rate": 1.9788467694221453e-05,
1352
+ "loss": 0.2897,
1353
+ "step": 3840
1354
+ },
1355
+ {
1356
+ "epoch": 0.45,
1357
+ "grad_norm": 7.978864884873927,
1358
+ "learning_rate": 1.9784491004906468e-05,
1359
+ "loss": 0.3118,
1360
+ "step": 3860
1361
+ },
1362
+ {
1363
+ "epoch": 0.45,
1364
+ "grad_norm": 6.631368778757732,
1365
+ "learning_rate": 1.9780477690233476e-05,
1366
+ "loss": 0.3165,
1367
+ "step": 3880
1368
+ },
1369
+ {
1370
+ "epoch": 0.46,
1371
+ "grad_norm": 6.541290304773085,
1372
+ "learning_rate": 1.9776427765226664e-05,
1373
+ "loss": 0.3379,
1374
+ "step": 3900
1375
+ },
1376
+ {
1377
+ "epoch": 0.46,
1378
+ "grad_norm": 6.42886929503141,
1379
+ "learning_rate": 1.9772341245047285e-05,
1380
+ "loss": 0.3136,
1381
+ "step": 3920
1382
+ },
1383
+ {
1384
+ "epoch": 0.46,
1385
+ "grad_norm": 9.004651033459043,
1386
+ "learning_rate": 1.9768218144993593e-05,
1387
+ "loss": 0.311,
1388
+ "step": 3940
1389
+ },
1390
+ {
1391
+ "epoch": 0.46,
1392
+ "grad_norm": 6.241942743461084,
1393
+ "learning_rate": 1.976405848050077e-05,
1394
+ "loss": 0.3224,
1395
+ "step": 3960
1396
+ },
1397
+ {
1398
+ "epoch": 0.47,
1399
+ "grad_norm": 7.030346892956473,
1400
+ "learning_rate": 1.975986226714089e-05,
1401
+ "loss": 0.3513,
1402
+ "step": 3980
1403
+ },
1404
+ {
1405
+ "epoch": 0.47,
1406
+ "grad_norm": 6.984184655106913,
1407
+ "learning_rate": 1.9755629520622835e-05,
1408
+ "loss": 0.3031,
1409
+ "step": 4000
1410
+ },
1411
+ {
1412
+ "epoch": 0.47,
1413
+ "grad_norm": 3.6594620639728253,
1414
+ "learning_rate": 1.9751360256792274e-05,
1415
+ "loss": 0.2819,
1416
+ "step": 4020
1417
+ },
1418
+ {
1419
+ "epoch": 0.47,
1420
+ "grad_norm": 3.9453364305431418,
1421
+ "learning_rate": 1.9747054491631567e-05,
1422
+ "loss": 0.2806,
1423
+ "step": 4040
1424
+ },
1425
+ {
1426
+ "epoch": 0.48,
1427
+ "grad_norm": 3.8004852838340475,
1428
+ "learning_rate": 1.9742712241259726e-05,
1429
+ "loss": 0.2712,
1430
+ "step": 4060
1431
+ },
1432
+ {
1433
+ "epoch": 0.48,
1434
+ "grad_norm": 4.671864793440983,
1435
+ "learning_rate": 1.9738333521932348e-05,
1436
+ "loss": 0.264,
1437
+ "step": 4080
1438
+ },
1439
+ {
1440
+ "epoch": 0.48,
1441
+ "grad_norm": 3.9158215253486452,
1442
+ "learning_rate": 1.9733918350041546e-05,
1443
+ "loss": 0.2466,
1444
+ "step": 4100
1445
+ },
1446
+ {
1447
+ "epoch": 0.48,
1448
+ "grad_norm": 3.232730170815344,
1449
+ "learning_rate": 1.972946674211591e-05,
1450
+ "loss": 0.28,
1451
+ "step": 4120
1452
+ },
1453
+ {
1454
+ "epoch": 0.48,
1455
+ "grad_norm": 5.26102706338969,
1456
+ "learning_rate": 1.972497871482043e-05,
1457
+ "loss": 0.2793,
1458
+ "step": 4140
1459
+ },
1460
+ {
1461
+ "epoch": 0.49,
1462
+ "grad_norm": 6.036896745386087,
1463
+ "learning_rate": 1.972045428495642e-05,
1464
+ "loss": 0.2551,
1465
+ "step": 4160
1466
+ },
1467
+ {
1468
+ "epoch": 0.49,
1469
+ "grad_norm": 3.4899809349872823,
1470
+ "learning_rate": 1.971589346946149e-05,
1471
+ "loss": 0.2746,
1472
+ "step": 4180
1473
+ },
1474
+ {
1475
+ "epoch": 0.49,
1476
+ "grad_norm": 4.003065603442425,
1477
+ "learning_rate": 1.9711296285409443e-05,
1478
+ "loss": 0.2639,
1479
+ "step": 4200
1480
+ },
1481
+ {
1482
+ "epoch": 0.49,
1483
+ "grad_norm": 2.8477299266017626,
1484
+ "learning_rate": 1.9706662750010255e-05,
1485
+ "loss": 0.2689,
1486
+ "step": 4220
1487
+ },
1488
+ {
1489
+ "epoch": 0.5,
1490
+ "grad_norm": 3.5491058459283837,
1491
+ "learning_rate": 1.970199288060996e-05,
1492
+ "loss": 0.2642,
1493
+ "step": 4240
1494
+ },
1495
+ {
1496
+ "epoch": 0.5,
1497
+ "grad_norm": 3.658049963633355,
1498
+ "learning_rate": 1.969728669469063e-05,
1499
+ "loss": 0.2699,
1500
+ "step": 4260
1501
+ },
1502
+ {
1503
+ "epoch": 0.5,
1504
+ "grad_norm": 3.7093939532444127,
1505
+ "learning_rate": 1.9692544209870287e-05,
1506
+ "loss": 0.2697,
1507
+ "step": 4280
1508
+ },
1509
+ {
1510
+ "epoch": 0.5,
1511
+ "grad_norm": 3.0342042036143444,
1512
+ "learning_rate": 1.968776544390283e-05,
1513
+ "loss": 0.2654,
1514
+ "step": 4300
1515
+ },
1516
+ {
1517
+ "epoch": 0.51,
1518
+ "grad_norm": 3.8174199112021174,
1519
+ "learning_rate": 1.9682950414678004e-05,
1520
+ "loss": 0.2525,
1521
+ "step": 4320
1522
+ },
1523
+ {
1524
+ "epoch": 0.51,
1525
+ "grad_norm": 5.35961771156863,
1526
+ "learning_rate": 1.9678099140221284e-05,
1527
+ "loss": 0.2464,
1528
+ "step": 4340
1529
+ },
1530
+ {
1531
+ "epoch": 0.51,
1532
+ "grad_norm": 5.063501894675061,
1533
+ "learning_rate": 1.9673211638693844e-05,
1534
+ "loss": 0.2788,
1535
+ "step": 4360
1536
+ },
1537
+ {
1538
+ "epoch": 0.51,
1539
+ "grad_norm": 3.8581072620927723,
1540
+ "learning_rate": 1.9668287928392478e-05,
1541
+ "loss": 0.2743,
1542
+ "step": 4380
1543
+ },
1544
+ {
1545
+ "epoch": 0.51,
1546
+ "grad_norm": 3.386524586463125,
1547
+ "learning_rate": 1.9663328027749523e-05,
1548
+ "loss": 0.2547,
1549
+ "step": 4400
1550
+ },
1551
+ {
1552
+ "epoch": 0.52,
1553
+ "grad_norm": 4.049063661443696,
1554
+ "learning_rate": 1.9658331955332807e-05,
1555
+ "loss": 0.2787,
1556
+ "step": 4420
1557
+ },
1558
+ {
1559
+ "epoch": 0.52,
1560
+ "grad_norm": 5.12761026294579,
1561
+ "learning_rate": 1.9653299729845564e-05,
1562
+ "loss": 0.2577,
1563
+ "step": 4440
1564
+ },
1565
+ {
1566
+ "epoch": 0.52,
1567
+ "grad_norm": 3.808933278135121,
1568
+ "learning_rate": 1.9648231370126375e-05,
1569
+ "loss": 0.2654,
1570
+ "step": 4460
1571
+ },
1572
+ {
1573
+ "epoch": 0.52,
1574
+ "grad_norm": 4.402659150192465,
1575
+ "learning_rate": 1.9643126895149087e-05,
1576
+ "loss": 0.2603,
1577
+ "step": 4480
1578
+ },
1579
+ {
1580
+ "epoch": 0.53,
1581
+ "grad_norm": 8.492295307503369,
1582
+ "learning_rate": 1.963798632402275e-05,
1583
+ "loss": 0.2691,
1584
+ "step": 4500
1585
+ },
1586
+ {
1587
+ "epoch": 0.53,
1588
+ "grad_norm": 3.628014593555578,
1589
+ "learning_rate": 1.9632809675991544e-05,
1590
+ "loss": 0.2711,
1591
+ "step": 4520
1592
+ },
1593
+ {
1594
+ "epoch": 0.53,
1595
+ "grad_norm": 5.207256154885955,
1596
+ "learning_rate": 1.9627596970434708e-05,
1597
+ "loss": 0.2604,
1598
+ "step": 4540
1599
+ },
1600
+ {
1601
+ "epoch": 0.53,
1602
+ "grad_norm": 3.892539047603965,
1603
+ "learning_rate": 1.962234822686646e-05,
1604
+ "loss": 0.2693,
1605
+ "step": 4560
1606
+ },
1607
+ {
1608
+ "epoch": 0.54,
1609
+ "grad_norm": 4.005334872796194,
1610
+ "learning_rate": 1.961706346493594e-05,
1611
+ "loss": 0.2774,
1612
+ "step": 4580
1613
+ },
1614
+ {
1615
+ "epoch": 0.54,
1616
+ "grad_norm": 4.456881895914548,
1617
+ "learning_rate": 1.9611742704427106e-05,
1618
+ "loss": 0.2531,
1619
+ "step": 4600
1620
+ },
1621
+ {
1622
+ "epoch": 0.54,
1623
+ "grad_norm": 3.925512807597538,
1624
+ "learning_rate": 1.9606385965258697e-05,
1625
+ "loss": 0.2507,
1626
+ "step": 4620
1627
+ },
1628
+ {
1629
+ "epoch": 0.54,
1630
+ "grad_norm": 4.401211069070843,
1631
+ "learning_rate": 1.9600993267484147e-05,
1632
+ "loss": 0.262,
1633
+ "step": 4640
1634
+ },
1635
+ {
1636
+ "epoch": 0.55,
1637
+ "grad_norm": 5.3437799820979075,
1638
+ "learning_rate": 1.9595564631291478e-05,
1639
+ "loss": 0.2516,
1640
+ "step": 4660
1641
+ },
1642
+ {
1643
+ "epoch": 0.55,
1644
+ "grad_norm": 4.553539308066039,
1645
+ "learning_rate": 1.9590100077003278e-05,
1646
+ "loss": 0.2603,
1647
+ "step": 4680
1648
+ },
1649
+ {
1650
+ "epoch": 0.55,
1651
+ "grad_norm": 3.169670353545694,
1652
+ "learning_rate": 1.958459962507658e-05,
1653
+ "loss": 0.2875,
1654
+ "step": 4700
1655
+ },
1656
+ {
1657
+ "epoch": 0.55,
1658
+ "grad_norm": 3.572973282405317,
1659
+ "learning_rate": 1.9579063296102812e-05,
1660
+ "loss": 0.2671,
1661
+ "step": 4720
1662
+ },
1663
+ {
1664
+ "epoch": 0.55,
1665
+ "grad_norm": 3.59782258517998,
1666
+ "learning_rate": 1.957349111080771e-05,
1667
+ "loss": 0.2837,
1668
+ "step": 4740
1669
+ },
1670
+ {
1671
+ "epoch": 0.56,
1672
+ "grad_norm": 5.199297439358037,
1673
+ "learning_rate": 1.956788309005123e-05,
1674
+ "loss": 0.2753,
1675
+ "step": 4760
1676
+ },
1677
+ {
1678
+ "epoch": 0.56,
1679
+ "grad_norm": 4.993257936146954,
1680
+ "learning_rate": 1.95622392548275e-05,
1681
+ "loss": 0.2714,
1682
+ "step": 4780
1683
+ },
1684
+ {
1685
+ "epoch": 0.56,
1686
+ "grad_norm": 3.7288744650364642,
1687
+ "learning_rate": 1.955655962626471e-05,
1688
+ "loss": 0.2609,
1689
+ "step": 4800
1690
+ },
1691
+ {
1692
+ "epoch": 0.56,
1693
+ "grad_norm": 5.377784540016259,
1694
+ "learning_rate": 1.9550844225625043e-05,
1695
+ "loss": 0.2605,
1696
+ "step": 4820
1697
+ },
1698
+ {
1699
+ "epoch": 0.57,
1700
+ "grad_norm": 3.1237811954279353,
1701
+ "learning_rate": 1.954509307430461e-05,
1702
+ "loss": 0.2569,
1703
+ "step": 4840
1704
+ },
1705
+ {
1706
+ "epoch": 0.57,
1707
+ "grad_norm": 4.101705029145583,
1708
+ "learning_rate": 1.953930619383335e-05,
1709
+ "loss": 0.2719,
1710
+ "step": 4860
1711
+ },
1712
+ {
1713
+ "epoch": 0.57,
1714
+ "grad_norm": 2.734092828632754,
1715
+ "learning_rate": 1.953348360587496e-05,
1716
+ "loss": 0.2503,
1717
+ "step": 4880
1718
+ },
1719
+ {
1720
+ "epoch": 0.57,
1721
+ "grad_norm": 3.4516150854343604,
1722
+ "learning_rate": 1.9527625332226804e-05,
1723
+ "loss": 0.2588,
1724
+ "step": 4900
1725
+ },
1726
+ {
1727
+ "epoch": 0.58,
1728
+ "grad_norm": 3.0561153265178085,
1729
+ "learning_rate": 1.952173139481985e-05,
1730
+ "loss": 0.2779,
1731
+ "step": 4920
1732
+ },
1733
+ {
1734
+ "epoch": 0.58,
1735
+ "grad_norm": 3.5096519082422986,
1736
+ "learning_rate": 1.9515801815718566e-05,
1737
+ "loss": 0.2567,
1738
+ "step": 4940
1739
+ },
1740
+ {
1741
+ "epoch": 0.58,
1742
+ "grad_norm": 3.1690240083159993,
1743
+ "learning_rate": 1.9509836617120856e-05,
1744
+ "loss": 0.2466,
1745
+ "step": 4960
1746
+ },
1747
+ {
1748
+ "epoch": 0.58,
1749
+ "grad_norm": 3.2970371296868484,
1750
+ "learning_rate": 1.950383582135796e-05,
1751
+ "loss": 0.2762,
1752
+ "step": 4980
1753
+ },
1754
+ {
1755
+ "epoch": 0.59,
1756
+ "grad_norm": 4.326207366177782,
1757
+ "learning_rate": 1.9497799450894387e-05,
1758
+ "loss": 0.2681,
1759
+ "step": 5000
1760
+ },
1761
+ {
1762
+ "epoch": 0.59,
1763
+ "grad_norm": 3.977255410491104,
1764
+ "learning_rate": 1.949172752832782e-05,
1765
+ "loss": 0.235,
1766
+ "step": 5020
1767
+ },
1768
+ {
1769
+ "epoch": 0.59,
1770
+ "grad_norm": 3.5634772315616665,
1771
+ "learning_rate": 1.9485620076389036e-05,
1772
+ "loss": 0.254,
1773
+ "step": 5040
1774
+ },
1775
+ {
1776
+ "epoch": 0.59,
1777
+ "grad_norm": 2.9667663371741377,
1778
+ "learning_rate": 1.9479477117941815e-05,
1779
+ "loss": 0.2582,
1780
+ "step": 5060
1781
+ },
1782
+ {
1783
+ "epoch": 0.59,
1784
+ "grad_norm": 3.628580495546253,
1785
+ "learning_rate": 1.9473298675982863e-05,
1786
+ "loss": 0.2526,
1787
+ "step": 5080
1788
+ },
1789
+ {
1790
+ "epoch": 0.6,
1791
+ "grad_norm": 7.393689026504584,
1792
+ "learning_rate": 1.946708477364172e-05,
1793
+ "loss": 0.2587,
1794
+ "step": 5100
1795
+ },
1796
+ {
1797
+ "epoch": 0.6,
1798
+ "grad_norm": 3.1975448011133407,
1799
+ "learning_rate": 1.9460835434180677e-05,
1800
+ "loss": 0.2679,
1801
+ "step": 5120
1802
+ },
1803
+ {
1804
+ "epoch": 0.6,
1805
+ "grad_norm": 3.231683615655966,
1806
+ "learning_rate": 1.9454550680994684e-05,
1807
+ "loss": 0.2589,
1808
+ "step": 5140
1809
+ },
1810
+ {
1811
+ "epoch": 0.6,
1812
+ "grad_norm": 2.956736460495959,
1813
+ "learning_rate": 1.9448230537611266e-05,
1814
+ "loss": 0.2594,
1815
+ "step": 5160
1816
+ },
1817
+ {
1818
+ "epoch": 0.61,
1819
+ "grad_norm": 3.326543297808526,
1820
+ "learning_rate": 1.9441875027690435e-05,
1821
+ "loss": 0.2556,
1822
+ "step": 5180
1823
+ },
1824
+ {
1825
+ "epoch": 0.61,
1826
+ "grad_norm": 3.635188025395725,
1827
+ "learning_rate": 1.94354841750246e-05,
1828
+ "loss": 0.2391,
1829
+ "step": 5200
1830
+ },
1831
+ {
1832
+ "epoch": 0.61,
1833
+ "grad_norm": 4.934642109517512,
1834
+ "learning_rate": 1.942905800353848e-05,
1835
+ "loss": 0.2576,
1836
+ "step": 5220
1837
+ },
1838
+ {
1839
+ "epoch": 0.61,
1840
+ "grad_norm": 3.8685357103793767,
1841
+ "learning_rate": 1.9422596537289016e-05,
1842
+ "loss": 0.2735,
1843
+ "step": 5240
1844
+ },
1845
+ {
1846
+ "epoch": 0.62,
1847
+ "grad_norm": 4.387694986472,
1848
+ "learning_rate": 1.941609980046527e-05,
1849
+ "loss": 0.2584,
1850
+ "step": 5260
1851
+ },
1852
+ {
1853
+ "epoch": 0.62,
1854
+ "grad_norm": 2.8046325848231977,
1855
+ "learning_rate": 1.9409567817388347e-05,
1856
+ "loss": 0.2527,
1857
+ "step": 5280
1858
+ },
1859
+ {
1860
+ "epoch": 0.62,
1861
+ "grad_norm": 3.4012361074638466,
1862
+ "learning_rate": 1.94030006125113e-05,
1863
+ "loss": 0.2566,
1864
+ "step": 5300
1865
+ },
1866
+ {
1867
+ "epoch": 0.62,
1868
+ "grad_norm": 5.450089536158873,
1869
+ "learning_rate": 1.9396398210419038e-05,
1870
+ "loss": 0.2667,
1871
+ "step": 5320
1872
+ },
1873
+ {
1874
+ "epoch": 0.62,
1875
+ "grad_norm": 3.7967573233998677,
1876
+ "learning_rate": 1.938976063582824e-05,
1877
+ "loss": 0.2536,
1878
+ "step": 5340
1879
+ },
1880
+ {
1881
+ "epoch": 0.63,
1882
+ "grad_norm": 3.0487888683110866,
1883
+ "learning_rate": 1.938308791358723e-05,
1884
+ "loss": 0.2673,
1885
+ "step": 5360
1886
+ },
1887
+ {
1888
+ "epoch": 0.63,
1889
+ "grad_norm": 3.2736066169116285,
1890
+ "learning_rate": 1.9376380068675956e-05,
1891
+ "loss": 0.2592,
1892
+ "step": 5380
1893
+ },
1894
+ {
1895
+ "epoch": 0.63,
1896
+ "grad_norm": 9.727952335441667,
1897
+ "learning_rate": 1.93696371262058e-05,
1898
+ "loss": 0.273,
1899
+ "step": 5400
1900
+ },
1901
+ {
1902
+ "epoch": 0.63,
1903
+ "grad_norm": 3.3824664284674686,
1904
+ "learning_rate": 1.936285911141958e-05,
1905
+ "loss": 0.2787,
1906
+ "step": 5420
1907
+ },
1908
+ {
1909
+ "epoch": 0.64,
1910
+ "grad_norm": 6.879080376487243,
1911
+ "learning_rate": 1.935604604969138e-05,
1912
+ "loss": 0.2698,
1913
+ "step": 5440
1914
+ },
1915
+ {
1916
+ "epoch": 0.64,
1917
+ "grad_norm": 4.579168728014292,
1918
+ "learning_rate": 1.934919796652649e-05,
1919
+ "loss": 0.2589,
1920
+ "step": 5460
1921
+ },
1922
+ {
1923
+ "epoch": 0.64,
1924
+ "grad_norm": 4.6776324362587784,
1925
+ "learning_rate": 1.9342314887561323e-05,
1926
+ "loss": 0.2544,
1927
+ "step": 5480
1928
+ },
1929
+ {
1930
+ "epoch": 0.64,
1931
+ "grad_norm": 3.693997622998789,
1932
+ "learning_rate": 1.9335396838563282e-05,
1933
+ "loss": 0.2446,
1934
+ "step": 5500
1935
+ },
1936
+ {
1937
+ "epoch": 0.65,
1938
+ "grad_norm": 5.06873813162095,
1939
+ "learning_rate": 1.932844384543069e-05,
1940
+ "loss": 0.2618,
1941
+ "step": 5520
1942
+ },
1943
+ {
1944
+ "epoch": 0.65,
1945
+ "grad_norm": 6.816411566527297,
1946
+ "learning_rate": 1.9321455934192693e-05,
1947
+ "loss": 0.2614,
1948
+ "step": 5540
1949
+ },
1950
+ {
1951
+ "epoch": 0.65,
1952
+ "grad_norm": 3.5475232040366853,
1953
+ "learning_rate": 1.9314433131009146e-05,
1954
+ "loss": 0.2942,
1955
+ "step": 5560
1956
+ },
1957
+ {
1958
+ "epoch": 0.65,
1959
+ "grad_norm": 5.287550198652632,
1960
+ "learning_rate": 1.9307375462170527e-05,
1961
+ "loss": 0.2482,
1962
+ "step": 5580
1963
+ },
1964
+ {
1965
+ "epoch": 0.66,
1966
+ "grad_norm": 11.389854038766874,
1967
+ "learning_rate": 1.9300282954097847e-05,
1968
+ "loss": 0.245,
1969
+ "step": 5600
1970
+ },
1971
+ {
1972
+ "epoch": 0.66,
1973
+ "grad_norm": 4.178435806663215,
1974
+ "learning_rate": 1.9293155633342534e-05,
1975
+ "loss": 0.2534,
1976
+ "step": 5620
1977
+ },
1978
+ {
1979
+ "epoch": 0.66,
1980
+ "grad_norm": 3.1710922703595035,
1981
+ "learning_rate": 1.928599352658633e-05,
1982
+ "loss": 0.2622,
1983
+ "step": 5640
1984
+ },
1985
+ {
1986
+ "epoch": 0.66,
1987
+ "grad_norm": 2.6946627953356996,
1988
+ "learning_rate": 1.927879666064122e-05,
1989
+ "loss": 0.2626,
1990
+ "step": 5660
1991
+ },
1992
+ {
1993
+ "epoch": 0.66,
1994
+ "grad_norm": 4.8581812333670955,
1995
+ "learning_rate": 1.92715650624493e-05,
1996
+ "loss": 0.2346,
1997
+ "step": 5680
1998
+ },
1999
+ {
2000
+ "epoch": 0.67,
2001
+ "grad_norm": 3.092927765677092,
2002
+ "learning_rate": 1.926429875908269e-05,
2003
+ "loss": 0.2585,
2004
+ "step": 5700
2005
+ },
2006
+ {
2007
+ "epoch": 0.67,
2008
+ "grad_norm": 4.029661826813332,
2009
+ "learning_rate": 1.9256997777743437e-05,
2010
+ "loss": 0.2467,
2011
+ "step": 5720
2012
+ },
2013
+ {
2014
+ "epoch": 0.67,
2015
+ "grad_norm": 4.642691104141709,
2016
+ "learning_rate": 1.924966214576341e-05,
2017
+ "loss": 0.2681,
2018
+ "step": 5740
2019
+ },
2020
+ {
2021
+ "epoch": 0.67,
2022
+ "grad_norm": 7.315047122076067,
2023
+ "learning_rate": 1.9242291890604186e-05,
2024
+ "loss": 0.2508,
2025
+ "step": 5760
2026
+ },
2027
+ {
2028
+ "epoch": 0.68,
2029
+ "grad_norm": 3.540549397482083,
2030
+ "learning_rate": 1.923488703985696e-05,
2031
+ "loss": 0.2592,
2032
+ "step": 5780
2033
+ },
2034
+ {
2035
+ "epoch": 0.68,
2036
+ "grad_norm": 3.475018859373003,
2037
+ "learning_rate": 1.9227447621242458e-05,
2038
+ "loss": 0.2422,
2039
+ "step": 5800
2040
+ },
2041
+ {
2042
+ "epoch": 0.68,
2043
+ "grad_norm": 3.8676644079356355,
2044
+ "learning_rate": 1.921997366261078e-05,
2045
+ "loss": 0.2603,
2046
+ "step": 5820
2047
+ },
2048
+ {
2049
+ "epoch": 0.68,
2050
+ "grad_norm": 3.5885966628483,
2051
+ "learning_rate": 1.9212465191941346e-05,
2052
+ "loss": 0.2542,
2053
+ "step": 5840
2054
+ },
2055
+ {
2056
+ "epoch": 0.69,
2057
+ "grad_norm": 3.6495890137770886,
2058
+ "learning_rate": 1.920492223734279e-05,
2059
+ "loss": 0.2374,
2060
+ "step": 5860
2061
+ },
2062
+ {
2063
+ "epoch": 0.69,
2064
+ "grad_norm": 4.323488251305847,
2065
+ "learning_rate": 1.9197344827052813e-05,
2066
+ "loss": 0.2687,
2067
+ "step": 5880
2068
+ },
2069
+ {
2070
+ "epoch": 0.69,
2071
+ "grad_norm": 5.051497760408599,
2072
+ "learning_rate": 1.9189732989438123e-05,
2073
+ "loss": 0.2678,
2074
+ "step": 5900
2075
+ },
2076
+ {
2077
+ "epoch": 0.69,
2078
+ "grad_norm": 3.5559342765441495,
2079
+ "learning_rate": 1.9182086752994295e-05,
2080
+ "loss": 0.2726,
2081
+ "step": 5920
2082
+ },
2083
+ {
2084
+ "epoch": 0.7,
2085
+ "grad_norm": 3.029184640641301,
2086
+ "learning_rate": 1.9174406146345694e-05,
2087
+ "loss": 0.246,
2088
+ "step": 5940
2089
+ },
2090
+ {
2091
+ "epoch": 0.7,
2092
+ "grad_norm": 3.4792605785250994,
2093
+ "learning_rate": 1.9166691198245343e-05,
2094
+ "loss": 0.2843,
2095
+ "step": 5960
2096
+ },
2097
+ {
2098
+ "epoch": 0.7,
2099
+ "grad_norm": 3.0247490770922,
2100
+ "learning_rate": 1.915894193757483e-05,
2101
+ "loss": 0.2528,
2102
+ "step": 5980
2103
+ },
2104
+ {
2105
+ "epoch": 0.7,
2106
+ "grad_norm": 3.7727567007247855,
2107
+ "learning_rate": 1.915115839334419e-05,
2108
+ "loss": 0.2442,
2109
+ "step": 6000
2110
+ },
2111
+ {
2112
+ "epoch": 0.7,
2113
+ "grad_norm": 2.31881357193246,
2114
+ "learning_rate": 1.9143340594691806e-05,
2115
+ "loss": 0.2561,
2116
+ "step": 6020
2117
+ },
2118
+ {
2119
+ "epoch": 0.71,
2120
+ "grad_norm": 3.138315746753083,
2121
+ "learning_rate": 1.913548857088429e-05,
2122
+ "loss": 0.2636,
2123
+ "step": 6040
2124
+ },
2125
+ {
2126
+ "epoch": 0.71,
2127
+ "grad_norm": 8.010145906221986,
2128
+ "learning_rate": 1.9127602351316393e-05,
2129
+ "loss": 0.2876,
2130
+ "step": 6060
2131
+ },
2132
+ {
2133
+ "epoch": 0.71,
2134
+ "grad_norm": 3.867596475572881,
2135
+ "learning_rate": 1.9119681965510858e-05,
2136
+ "loss": 0.2659,
2137
+ "step": 6080
2138
+ },
2139
+ {
2140
+ "epoch": 0.71,
2141
+ "grad_norm": 3.3587166074477777,
2142
+ "learning_rate": 1.9111727443118357e-05,
2143
+ "loss": 0.2634,
2144
+ "step": 6100
2145
+ },
2146
+ {
2147
+ "epoch": 0.72,
2148
+ "grad_norm": 2.89895288044486,
2149
+ "learning_rate": 1.9103738813917336e-05,
2150
+ "loss": 0.2533,
2151
+ "step": 6120
2152
+ },
2153
+ {
2154
+ "epoch": 0.72,
2155
+ "grad_norm": 4.037752097559083,
2156
+ "learning_rate": 1.909571610781393e-05,
2157
+ "loss": 0.2485,
2158
+ "step": 6140
2159
+ },
2160
+ {
2161
+ "epoch": 0.72,
2162
+ "grad_norm": 3.763608653829743,
2163
+ "learning_rate": 1.908765935484185e-05,
2164
+ "loss": 0.259,
2165
+ "step": 6160
2166
+ },
2167
+ {
2168
+ "epoch": 0.72,
2169
+ "grad_norm": 3.161229351032073,
2170
+ "learning_rate": 1.9079568585162248e-05,
2171
+ "loss": 0.2508,
2172
+ "step": 6180
2173
+ },
2174
+ {
2175
+ "epoch": 0.73,
2176
+ "grad_norm": 2.725004696623228,
2177
+ "learning_rate": 1.9071443829063644e-05,
2178
+ "loss": 0.2524,
2179
+ "step": 6200
2180
+ },
2181
+ {
2182
+ "epoch": 0.73,
2183
+ "grad_norm": 3.7116897623627985,
2184
+ "learning_rate": 1.906328511696177e-05,
2185
+ "loss": 0.2598,
2186
+ "step": 6220
2187
+ },
2188
+ {
2189
+ "epoch": 0.73,
2190
+ "grad_norm": 2.9004110110815544,
2191
+ "learning_rate": 1.9055092479399474e-05,
2192
+ "loss": 0.2553,
2193
+ "step": 6240
2194
+ },
2195
+ {
2196
+ "epoch": 0.73,
2197
+ "grad_norm": 4.311030925472568,
2198
+ "learning_rate": 1.9046865947046627e-05,
2199
+ "loss": 0.2678,
2200
+ "step": 6260
2201
+ },
2202
+ {
2203
+ "epoch": 0.73,
2204
+ "grad_norm": 3.5377476420175977,
2205
+ "learning_rate": 1.903860555069997e-05,
2206
+ "loss": 0.2379,
2207
+ "step": 6280
2208
+ },
2209
+ {
2210
+ "epoch": 0.74,
2211
+ "grad_norm": 3.58934533910836,
2212
+ "learning_rate": 1.9030311321283022e-05,
2213
+ "loss": 0.2608,
2214
+ "step": 6300
2215
+ },
2216
+ {
2217
+ "epoch": 0.74,
2218
+ "grad_norm": 3.4612535457241624,
2219
+ "learning_rate": 1.9021983289845956e-05,
2220
+ "loss": 0.246,
2221
+ "step": 6320
2222
+ },
2223
+ {
2224
+ "epoch": 0.74,
2225
+ "grad_norm": 3.9763337015158258,
2226
+ "learning_rate": 1.9013621487565498e-05,
2227
+ "loss": 0.238,
2228
+ "step": 6340
2229
+ },
2230
+ {
2231
+ "epoch": 0.74,
2232
+ "grad_norm": 3.05488683337997,
2233
+ "learning_rate": 1.9005225945744782e-05,
2234
+ "loss": 0.2472,
2235
+ "step": 6360
2236
+ },
2237
+ {
2238
+ "epoch": 0.75,
2239
+ "grad_norm": 3.2350295137123313,
2240
+ "learning_rate": 1.8996796695813256e-05,
2241
+ "loss": 0.261,
2242
+ "step": 6380
2243
+ },
2244
+ {
2245
+ "epoch": 0.75,
2246
+ "grad_norm": 3.2500906711548123,
2247
+ "learning_rate": 1.8988333769326564e-05,
2248
+ "loss": 0.2641,
2249
+ "step": 6400
2250
+ },
2251
+ {
2252
+ "epoch": 0.75,
2253
+ "grad_norm": 2.815087548041305,
2254
+ "learning_rate": 1.8979837197966415e-05,
2255
+ "loss": 0.2583,
2256
+ "step": 6420
2257
+ },
2258
+ {
2259
+ "epoch": 0.75,
2260
+ "grad_norm": 3.7397965536014257,
2261
+ "learning_rate": 1.8971307013540464e-05,
2262
+ "loss": 0.2485,
2263
+ "step": 6440
2264
+ },
2265
+ {
2266
+ "epoch": 0.76,
2267
+ "grad_norm": 2.7592373806083423,
2268
+ "learning_rate": 1.8962743247982205e-05,
2269
+ "loss": 0.2641,
2270
+ "step": 6460
2271
+ },
2272
+ {
2273
+ "epoch": 0.76,
2274
+ "grad_norm": 2.818080452105616,
2275
+ "learning_rate": 1.895414593335085e-05,
2276
+ "loss": 0.2554,
2277
+ "step": 6480
2278
+ },
2279
+ {
2280
+ "epoch": 0.76,
2281
+ "grad_norm": 3.627385505536174,
2282
+ "learning_rate": 1.8945515101831203e-05,
2283
+ "loss": 0.2693,
2284
+ "step": 6500
2285
+ },
2286
+ {
2287
+ "epoch": 0.76,
2288
+ "grad_norm": 3.255509767944512,
2289
+ "learning_rate": 1.893685078573354e-05,
2290
+ "loss": 0.2459,
2291
+ "step": 6520
2292
+ },
2293
+ {
2294
+ "epoch": 0.77,
2295
+ "grad_norm": 4.135182845698409,
2296
+ "learning_rate": 1.8928153017493488e-05,
2297
+ "loss": 0.2499,
2298
+ "step": 6540
2299
+ },
2300
+ {
2301
+ "epoch": 0.77,
2302
+ "grad_norm": 5.892816677202276,
2303
+ "learning_rate": 1.8919421829671904e-05,
2304
+ "loss": 0.2786,
2305
+ "step": 6560
2306
+ },
2307
+ {
2308
+ "epoch": 0.77,
2309
+ "grad_norm": 3.479892875732629,
2310
+ "learning_rate": 1.891065725495476e-05,
2311
+ "loss": 0.265,
2312
+ "step": 6580
2313
+ },
2314
+ {
2315
+ "epoch": 0.77,
2316
+ "grad_norm": 3.8343809741781216,
2317
+ "learning_rate": 1.890185932615301e-05,
2318
+ "loss": 0.2527,
2319
+ "step": 6600
2320
+ },
2321
+ {
2322
+ "epoch": 0.77,
2323
+ "grad_norm": 3.1643033159708107,
2324
+ "learning_rate": 1.8893028076202472e-05,
2325
+ "loss": 0.2741,
2326
+ "step": 6620
2327
+ },
2328
+ {
2329
+ "epoch": 0.78,
2330
+ "grad_norm": 4.477352585232187,
2331
+ "learning_rate": 1.8884163538163708e-05,
2332
+ "loss": 0.2525,
2333
+ "step": 6640
2334
+ },
2335
+ {
2336
+ "epoch": 0.78,
2337
+ "grad_norm": 3.529760400999529,
2338
+ "learning_rate": 1.8875265745221894e-05,
2339
+ "loss": 0.2643,
2340
+ "step": 6660
2341
+ },
2342
+ {
2343
+ "epoch": 0.78,
2344
+ "grad_norm": 2.3110189076619525,
2345
+ "learning_rate": 1.8866334730686703e-05,
2346
+ "loss": 0.2721,
2347
+ "step": 6680
2348
+ },
2349
+ {
2350
+ "epoch": 0.78,
2351
+ "grad_norm": 3.685226774238388,
2352
+ "learning_rate": 1.8857370527992165e-05,
2353
+ "loss": 0.2549,
2354
+ "step": 6700
2355
+ },
2356
+ {
2357
+ "epoch": 0.79,
2358
+ "grad_norm": 3.8074891978039864,
2359
+ "learning_rate": 1.8848373170696567e-05,
2360
+ "loss": 0.2418,
2361
+ "step": 6720
2362
+ },
2363
+ {
2364
+ "epoch": 0.79,
2365
+ "grad_norm": 7.507877790185021,
2366
+ "learning_rate": 1.8839342692482302e-05,
2367
+ "loss": 0.2451,
2368
+ "step": 6740
2369
+ },
2370
+ {
2371
+ "epoch": 0.79,
2372
+ "grad_norm": 2.7359733051983186,
2373
+ "learning_rate": 1.8830279127155764e-05,
2374
+ "loss": 0.2484,
2375
+ "step": 6760
2376
+ },
2377
+ {
2378
+ "epoch": 0.79,
2379
+ "grad_norm": 2.4635290638854657,
2380
+ "learning_rate": 1.88211825086472e-05,
2381
+ "loss": 0.2407,
2382
+ "step": 6780
2383
+ },
2384
+ {
2385
+ "epoch": 0.8,
2386
+ "grad_norm": 3.040130192327906,
2387
+ "learning_rate": 1.8812052871010605e-05,
2388
+ "loss": 0.2544,
2389
+ "step": 6800
2390
+ },
2391
+ {
2392
+ "epoch": 0.8,
2393
+ "grad_norm": 5.610778632978326,
2394
+ "learning_rate": 1.8802890248423574e-05,
2395
+ "loss": 0.2397,
2396
+ "step": 6820
2397
+ },
2398
+ {
2399
+ "epoch": 0.8,
2400
+ "grad_norm": 4.156547822535594,
2401
+ "learning_rate": 1.8793694675187197e-05,
2402
+ "loss": 0.2635,
2403
+ "step": 6840
2404
+ },
2405
+ {
2406
+ "epoch": 0.8,
2407
+ "grad_norm": 4.057314806159278,
2408
+ "learning_rate": 1.8784466185725902e-05,
2409
+ "loss": 0.2598,
2410
+ "step": 6860
2411
+ },
2412
+ {
2413
+ "epoch": 0.81,
2414
+ "grad_norm": 8.314874066232434,
2415
+ "learning_rate": 1.877520481458736e-05,
2416
+ "loss": 0.2706,
2417
+ "step": 6880
2418
+ },
2419
+ {
2420
+ "epoch": 0.81,
2421
+ "grad_norm": 3.7767818313058195,
2422
+ "learning_rate": 1.8765910596442316e-05,
2423
+ "loss": 0.2483,
2424
+ "step": 6900
2425
+ },
2426
+ {
2427
+ "epoch": 0.81,
2428
+ "grad_norm": 3.53905010431742,
2429
+ "learning_rate": 1.8756583566084502e-05,
2430
+ "loss": 0.2683,
2431
+ "step": 6920
2432
+ },
2433
+ {
2434
+ "epoch": 0.81,
2435
+ "grad_norm": 3.2053331160944625,
2436
+ "learning_rate": 1.8747223758430467e-05,
2437
+ "loss": 0.2567,
2438
+ "step": 6940
2439
+ },
2440
+ {
2441
+ "epoch": 0.81,
2442
+ "grad_norm": 3.655977418308093,
2443
+ "learning_rate": 1.8737831208519478e-05,
2444
+ "loss": 0.2433,
2445
+ "step": 6960
2446
+ },
2447
+ {
2448
+ "epoch": 0.82,
2449
+ "grad_norm": 3.334646935711793,
2450
+ "learning_rate": 1.8728405951513376e-05,
2451
+ "loss": 0.2442,
2452
+ "step": 6980
2453
+ },
2454
+ {
2455
+ "epoch": 0.82,
2456
+ "grad_norm": 3.3678493336120505,
2457
+ "learning_rate": 1.8718948022696427e-05,
2458
+ "loss": 0.2634,
2459
+ "step": 7000
2460
+ },
2461
+ {
2462
+ "epoch": 0.82,
2463
+ "grad_norm": 2.9941614875638956,
2464
+ "learning_rate": 1.8709457457475224e-05,
2465
+ "loss": 0.2719,
2466
+ "step": 7020
2467
+ },
2468
+ {
2469
+ "epoch": 0.82,
2470
+ "grad_norm": 3.3212764283242495,
2471
+ "learning_rate": 1.8699934291378534e-05,
2472
+ "loss": 0.2616,
2473
+ "step": 7040
2474
+ },
2475
+ {
2476
+ "epoch": 0.83,
2477
+ "grad_norm": 3.7449113015848963,
2478
+ "learning_rate": 1.8690378560057165e-05,
2479
+ "loss": 0.251,
2480
+ "step": 7060
2481
+ },
2482
+ {
2483
+ "epoch": 0.83,
2484
+ "grad_norm": 2.588981565454773,
2485
+ "learning_rate": 1.868079029928383e-05,
2486
+ "loss": 0.2345,
2487
+ "step": 7080
2488
+ },
2489
+ {
2490
+ "epoch": 0.83,
2491
+ "grad_norm": 4.188807496871825,
2492
+ "learning_rate": 1.8671169544953033e-05,
2493
+ "loss": 0.2537,
2494
+ "step": 7100
2495
+ },
2496
+ {
2497
+ "epoch": 0.83,
2498
+ "grad_norm": 4.66518430781191,
2499
+ "learning_rate": 1.866151633308091e-05,
2500
+ "loss": 0.2397,
2501
+ "step": 7120
2502
+ },
2503
+ {
2504
+ "epoch": 0.84,
2505
+ "grad_norm": 4.197676860816234,
2506
+ "learning_rate": 1.8651830699805108e-05,
2507
+ "loss": 0.2446,
2508
+ "step": 7140
2509
+ },
2510
+ {
2511
+ "epoch": 0.84,
2512
+ "grad_norm": 3.6434672643458543,
2513
+ "learning_rate": 1.8642112681384645e-05,
2514
+ "loss": 0.2402,
2515
+ "step": 7160
2516
+ },
2517
+ {
2518
+ "epoch": 0.84,
2519
+ "grad_norm": 3.950387723218738,
2520
+ "learning_rate": 1.8632362314199778e-05,
2521
+ "loss": 0.2536,
2522
+ "step": 7180
2523
+ },
2524
+ {
2525
+ "epoch": 0.84,
2526
+ "grad_norm": 9.658211616844094,
2527
+ "learning_rate": 1.8622579634751864e-05,
2528
+ "loss": 0.2541,
2529
+ "step": 7200
2530
+ },
2531
+ {
2532
+ "epoch": 0.84,
2533
+ "grad_norm": 3.799337836345952,
2534
+ "learning_rate": 1.8612764679663222e-05,
2535
+ "loss": 0.2586,
2536
+ "step": 7220
2537
+ },
2538
+ {
2539
+ "epoch": 0.85,
2540
+ "grad_norm": 3.737906856782671,
2541
+ "learning_rate": 1.8602917485677e-05,
2542
+ "loss": 0.2539,
2543
+ "step": 7240
2544
+ },
2545
+ {
2546
+ "epoch": 0.85,
2547
+ "grad_norm": 3.2570157254536016,
2548
+ "learning_rate": 1.859303808965703e-05,
2549
+ "loss": 0.2533,
2550
+ "step": 7260
2551
+ },
2552
+ {
2553
+ "epoch": 0.85,
2554
+ "grad_norm": 2.8736525985585226,
2555
+ "learning_rate": 1.85831265285877e-05,
2556
+ "loss": 0.2532,
2557
+ "step": 7280
2558
+ },
2559
+ {
2560
+ "epoch": 0.85,
2561
+ "grad_norm": 3.5447389329516503,
2562
+ "learning_rate": 1.857318283957381e-05,
2563
+ "loss": 0.2387,
2564
+ "step": 7300
2565
+ },
2566
+ {
2567
+ "epoch": 0.86,
2568
+ "grad_norm": 2.7468496831393403,
2569
+ "learning_rate": 1.8563207059840436e-05,
2570
+ "loss": 0.2455,
2571
+ "step": 7320
2572
+ },
2573
+ {
2574
+ "epoch": 0.86,
2575
+ "grad_norm": 3.483830293921577,
2576
+ "learning_rate": 1.8553199226732784e-05,
2577
+ "loss": 0.2456,
2578
+ "step": 7340
2579
+ },
2580
+ {
2581
+ "epoch": 0.86,
2582
+ "grad_norm": 3.3132796899404635,
2583
+ "learning_rate": 1.8543159377716054e-05,
2584
+ "loss": 0.2564,
2585
+ "step": 7360
2586
+ },
2587
+ {
2588
+ "epoch": 0.86,
2589
+ "grad_norm": 3.3795508326649575,
2590
+ "learning_rate": 1.8533087550375302e-05,
2591
+ "loss": 0.2527,
2592
+ "step": 7380
2593
+ },
2594
+ {
2595
+ "epoch": 0.87,
2596
+ "grad_norm": 4.001941448174156,
2597
+ "learning_rate": 1.8522983782415306e-05,
2598
+ "loss": 0.2433,
2599
+ "step": 7400
2600
+ },
2601
+ {
2602
+ "epoch": 0.87,
2603
+ "grad_norm": 3.4205382361927117,
2604
+ "learning_rate": 1.85128481116604e-05,
2605
+ "loss": 0.2538,
2606
+ "step": 7420
2607
+ },
2608
+ {
2609
+ "epoch": 0.87,
2610
+ "grad_norm": 3.896330422632185,
2611
+ "learning_rate": 1.8502680576054365e-05,
2612
+ "loss": 0.2478,
2613
+ "step": 7440
2614
+ },
2615
+ {
2616
+ "epoch": 0.87,
2617
+ "grad_norm": 5.020228094868969,
2618
+ "learning_rate": 1.849248121366026e-05,
2619
+ "loss": 0.2427,
2620
+ "step": 7460
2621
+ },
2622
+ {
2623
+ "epoch": 0.88,
2624
+ "grad_norm": 3.574276782956127,
2625
+ "learning_rate": 1.8482250062660293e-05,
2626
+ "loss": 0.2389,
2627
+ "step": 7480
2628
+ },
2629
+ {
2630
+ "epoch": 0.88,
2631
+ "grad_norm": 3.5309579103911184,
2632
+ "learning_rate": 1.847198716135568e-05,
2633
+ "loss": 0.2567,
2634
+ "step": 7500
2635
+ },
2636
+ {
2637
+ "epoch": 0.88,
2638
+ "grad_norm": 3.1760414458004,
2639
+ "learning_rate": 1.846169254816649e-05,
2640
+ "loss": 0.2291,
2641
+ "step": 7520
2642
+ },
2643
+ {
2644
+ "epoch": 0.88,
2645
+ "grad_norm": 4.214334621561301,
2646
+ "learning_rate": 1.8451366261631514e-05,
2647
+ "loss": 0.2624,
2648
+ "step": 7540
2649
+ },
2650
+ {
2651
+ "epoch": 0.88,
2652
+ "grad_norm": 2.977221959069712,
2653
+ "learning_rate": 1.8441008340408115e-05,
2654
+ "loss": 0.2425,
2655
+ "step": 7560
2656
+ },
2657
+ {
2658
+ "epoch": 0.89,
2659
+ "grad_norm": 3.78383188925913,
2660
+ "learning_rate": 1.8430618823272076e-05,
2661
+ "loss": 0.2437,
2662
+ "step": 7580
2663
+ },
2664
+ {
2665
+ "epoch": 0.89,
2666
+ "grad_norm": 3.2432934448936224,
2667
+ "learning_rate": 1.842019774911747e-05,
2668
+ "loss": 0.2628,
2669
+ "step": 7600
2670
+ },
2671
+ {
2672
+ "epoch": 0.89,
2673
+ "grad_norm": 3.05382414419798,
2674
+ "learning_rate": 1.8409745156956503e-05,
2675
+ "loss": 0.2332,
2676
+ "step": 7620
2677
+ },
2678
+ {
2679
+ "epoch": 0.89,
2680
+ "grad_norm": 3.098744070302099,
2681
+ "learning_rate": 1.839926108591937e-05,
2682
+ "loss": 0.2359,
2683
+ "step": 7640
2684
+ },
2685
+ {
2686
+ "epoch": 0.9,
2687
+ "grad_norm": 3.427098715766002,
2688
+ "learning_rate": 1.8388745575254117e-05,
2689
+ "loss": 0.2429,
2690
+ "step": 7660
2691
+ },
2692
+ {
2693
+ "epoch": 0.9,
2694
+ "grad_norm": 5.630711008711337,
2695
+ "learning_rate": 1.8378198664326474e-05,
2696
+ "loss": 0.2416,
2697
+ "step": 7680
2698
+ },
2699
+ {
2700
+ "epoch": 0.9,
2701
+ "grad_norm": 3.1144723181358622,
2702
+ "learning_rate": 1.836762039261974e-05,
2703
+ "loss": 0.2406,
2704
+ "step": 7700
2705
+ },
2706
+ {
2707
+ "epoch": 0.9,
2708
+ "grad_norm": 4.070301722184627,
2709
+ "learning_rate": 1.8357010799734588e-05,
2710
+ "loss": 0.234,
2711
+ "step": 7720
2712
+ },
2713
+ {
2714
+ "epoch": 0.91,
2715
+ "grad_norm": 3.255428182680105,
2716
+ "learning_rate": 1.834636992538898e-05,
2717
+ "loss": 0.2365,
2718
+ "step": 7740
2719
+ },
2720
+ {
2721
+ "epoch": 0.91,
2722
+ "grad_norm": 3.4823775104086576,
2723
+ "learning_rate": 1.8335697809417952e-05,
2724
+ "loss": 0.2376,
2725
+ "step": 7760
2726
+ },
2727
+ {
2728
+ "epoch": 0.91,
2729
+ "grad_norm": 5.528269163918348,
2730
+ "learning_rate": 1.8324994491773507e-05,
2731
+ "loss": 0.2679,
2732
+ "step": 7780
2733
+ },
2734
+ {
2735
+ "epoch": 0.91,
2736
+ "grad_norm": 2.7221100312640467,
2737
+ "learning_rate": 1.8314260012524462e-05,
2738
+ "loss": 0.2544,
2739
+ "step": 7800
2740
+ },
2741
+ {
2742
+ "epoch": 0.92,
2743
+ "grad_norm": 3.224515621271091,
2744
+ "learning_rate": 1.8303494411856278e-05,
2745
+ "loss": 0.2586,
2746
+ "step": 7820
2747
+ },
2748
+ {
2749
+ "epoch": 0.92,
2750
+ "grad_norm": 2.798714485709818,
2751
+ "learning_rate": 1.829269773007092e-05,
2752
+ "loss": 0.2277,
2753
+ "step": 7840
2754
+ },
2755
+ {
2756
+ "epoch": 0.92,
2757
+ "grad_norm": 3.890849401904209,
2758
+ "learning_rate": 1.8281870007586735e-05,
2759
+ "loss": 0.2565,
2760
+ "step": 7860
2761
+ },
2762
+ {
2763
+ "epoch": 0.92,
2764
+ "grad_norm": 3.701194549732654,
2765
+ "learning_rate": 1.8271011284938233e-05,
2766
+ "loss": 0.2307,
2767
+ "step": 7880
2768
+ },
2769
+ {
2770
+ "epoch": 0.92,
2771
+ "grad_norm": 3.9057568048025875,
2772
+ "learning_rate": 1.8260121602776003e-05,
2773
+ "loss": 0.2464,
2774
+ "step": 7900
2775
+ },
2776
+ {
2777
+ "epoch": 0.93,
2778
+ "grad_norm": 3.397196786583742,
2779
+ "learning_rate": 1.8249201001866533e-05,
2780
+ "loss": 0.25,
2781
+ "step": 7920
2782
+ },
2783
+ {
2784
+ "epoch": 0.93,
2785
+ "grad_norm": 3.6136942859367496,
2786
+ "learning_rate": 1.8238249523092046e-05,
2787
+ "loss": 0.2398,
2788
+ "step": 7940
2789
+ },
2790
+ {
2791
+ "epoch": 0.93,
2792
+ "grad_norm": 3.6009293204428765,
2793
+ "learning_rate": 1.8227267207450363e-05,
2794
+ "loss": 0.2356,
2795
+ "step": 7960
2796
+ },
2797
+ {
2798
+ "epoch": 0.93,
2799
+ "grad_norm": 3.1462270393488585,
2800
+ "learning_rate": 1.8216254096054747e-05,
2801
+ "loss": 0.2292,
2802
+ "step": 7980
2803
+ },
2804
+ {
2805
+ "epoch": 0.94,
2806
+ "grad_norm": 3.453766085807792,
2807
+ "learning_rate": 1.8205210230133756e-05,
2808
+ "loss": 0.2532,
2809
+ "step": 8000
2810
+ },
2811
+ {
2812
+ "epoch": 0.94,
2813
+ "grad_norm": 4.718419953986993,
2814
+ "learning_rate": 1.8194135651031063e-05,
2815
+ "loss": 0.2379,
2816
+ "step": 8020
2817
+ },
2818
+ {
2819
+ "epoch": 0.94,
2820
+ "grad_norm": 4.074122309800583,
2821
+ "learning_rate": 1.8183030400205335e-05,
2822
+ "loss": 0.247,
2823
+ "step": 8040
2824
+ },
2825
+ {
2826
+ "epoch": 0.94,
2827
+ "grad_norm": 3.597629476607082,
2828
+ "learning_rate": 1.817189451923005e-05,
2829
+ "loss": 0.2563,
2830
+ "step": 8060
2831
+ },
2832
+ {
2833
+ "epoch": 0.95,
2834
+ "grad_norm": 3.2550443969136436,
2835
+ "learning_rate": 1.8160728049793357e-05,
2836
+ "loss": 0.2425,
2837
+ "step": 8080
2838
+ },
2839
+ {
2840
+ "epoch": 0.95,
2841
+ "grad_norm": 2.6027515890006505,
2842
+ "learning_rate": 1.8149531033697913e-05,
2843
+ "loss": 0.2329,
2844
+ "step": 8100
2845
+ },
2846
+ {
2847
+ "epoch": 0.95,
2848
+ "grad_norm": 3.705349797095471,
2849
+ "learning_rate": 1.8138303512860737e-05,
2850
+ "loss": 0.2521,
2851
+ "step": 8120
2852
+ },
2853
+ {
2854
+ "epoch": 0.95,
2855
+ "grad_norm": 3.510055266811937,
2856
+ "learning_rate": 1.812704552931303e-05,
2857
+ "loss": 0.2413,
2858
+ "step": 8140
2859
+ },
2860
+ {
2861
+ "epoch": 0.95,
2862
+ "grad_norm": 3.2863076604207997,
2863
+ "learning_rate": 1.8115757125200054e-05,
2864
+ "loss": 0.2471,
2865
+ "step": 8160
2866
+ },
2867
+ {
2868
+ "epoch": 0.96,
2869
+ "grad_norm": 3.174018473859266,
2870
+ "learning_rate": 1.810443834278094e-05,
2871
+ "loss": 0.2429,
2872
+ "step": 8180
2873
+ },
2874
+ {
2875
+ "epoch": 0.96,
2876
+ "grad_norm": 4.78684161965386,
2877
+ "learning_rate": 1.8093089224428533e-05,
2878
+ "loss": 0.21,
2879
+ "step": 8200
2880
+ },
2881
+ {
2882
+ "epoch": 0.96,
2883
+ "grad_norm": 3.778736256263061,
2884
+ "learning_rate": 1.8081709812629265e-05,
2885
+ "loss": 0.2438,
2886
+ "step": 8220
2887
+ },
2888
+ {
2889
+ "epoch": 0.96,
2890
+ "grad_norm": 2.011964890319697,
2891
+ "learning_rate": 1.807030014998296e-05,
2892
+ "loss": 0.2235,
2893
+ "step": 8240
2894
+ },
2895
+ {
2896
+ "epoch": 0.97,
2897
+ "grad_norm": 2.469231474076661,
2898
+ "learning_rate": 1.8058860279202698e-05,
2899
+ "loss": 0.2361,
2900
+ "step": 8260
2901
+ },
2902
+ {
2903
+ "epoch": 0.97,
2904
+ "grad_norm": 4.194473608952758,
2905
+ "learning_rate": 1.804739024311463e-05,
2906
+ "loss": 0.2485,
2907
+ "step": 8280
2908
+ },
2909
+ {
2910
+ "epoch": 0.97,
2911
+ "grad_norm": 3.648802255045255,
2912
+ "learning_rate": 1.8035890084657854e-05,
2913
+ "loss": 0.2448,
2914
+ "step": 8300
2915
+ },
2916
+ {
2917
+ "epoch": 0.97,
2918
+ "grad_norm": 2.8256525011924443,
2919
+ "learning_rate": 1.8024359846884213e-05,
2920
+ "loss": 0.2688,
2921
+ "step": 8320
2922
+ },
2923
+ {
2924
+ "epoch": 0.98,
2925
+ "grad_norm": 4.409589867929299,
2926
+ "learning_rate": 1.8012799572958173e-05,
2927
+ "loss": 0.2405,
2928
+ "step": 8340
2929
+ },
2930
+ {
2931
+ "epoch": 0.98,
2932
+ "grad_norm": 5.024619525530542,
2933
+ "learning_rate": 1.8001209306156634e-05,
2934
+ "loss": 0.2284,
2935
+ "step": 8360
2936
+ },
2937
+ {
2938
+ "epoch": 0.98,
2939
+ "grad_norm": 3.5087106485202026,
2940
+ "learning_rate": 1.7989589089868774e-05,
2941
+ "loss": 0.2298,
2942
+ "step": 8380
2943
+ },
2944
+ {
2945
+ "epoch": 0.98,
2946
+ "grad_norm": 4.40165070000877,
2947
+ "learning_rate": 1.7977938967595897e-05,
2948
+ "loss": 0.2507,
2949
+ "step": 8400
2950
+ },
2951
+ {
2952
+ "epoch": 0.99,
2953
+ "grad_norm": 3.9660729217673705,
2954
+ "learning_rate": 1.7966258982951255e-05,
2955
+ "loss": 0.2459,
2956
+ "step": 8420
2957
+ },
2958
+ {
2959
+ "epoch": 0.99,
2960
+ "grad_norm": 4.665706172372416,
2961
+ "learning_rate": 1.7954549179659903e-05,
2962
+ "loss": 0.2458,
2963
+ "step": 8440
2964
+ },
2965
+ {
2966
+ "epoch": 0.99,
2967
+ "grad_norm": 3.2908069759526333,
2968
+ "learning_rate": 1.794280960155851e-05,
2969
+ "loss": 0.2251,
2970
+ "step": 8460
2971
+ },
2972
+ {
2973
+ "epoch": 0.99,
2974
+ "grad_norm": 6.7504752662905725,
2975
+ "learning_rate": 1.7931040292595227e-05,
2976
+ "loss": 0.254,
2977
+ "step": 8480
2978
+ },
2979
+ {
2980
+ "epoch": 0.99,
2981
+ "grad_norm": 2.2047073417435787,
2982
+ "learning_rate": 1.791924129682949e-05,
2983
+ "loss": 0.2456,
2984
+ "step": 8500
2985
+ },
2986
+ {
2987
+ "epoch": 1.0,
2988
+ "grad_norm": 3.9048911210632236,
2989
+ "learning_rate": 1.7907412658431877e-05,
2990
+ "loss": 0.2307,
2991
+ "step": 8520
2992
+ },
2993
+ {
2994
+ "epoch": 1.0,
2995
+ "grad_norm": 3.33290909610624,
2996
+ "learning_rate": 1.7895554421683938e-05,
2997
+ "loss": 0.2514,
2998
+ "step": 8540
2999
+ }
3000
+ ],
3001
+ "logging_steps": 20,
3002
+ "max_steps": 34184,
3003
+ "num_input_tokens_seen": 0,
3004
+ "num_train_epochs": 4,
3005
+ "save_steps": 500,
3006
+ "total_flos": 0.0,
3007
+ "train_batch_size": 1,
3008
+ "trial_name": null,
3009
+ "trial_params": null
3010
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f33bda593f07c6124a4429ccae3e94c7c86a2cfcc673b73aaba2e1ac11baeef
3
+ size 6840
zero_to_fp32.py ADDED
@@ -0,0 +1,592 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
215
+ elif zero_stage == 3:
216
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
217
+
218
+
219
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
220
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
221
+ return
222
+
223
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
224
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
225
+
226
+ if debug:
227
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
228
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
229
+
230
+ wanted_params = len(frozen_param_shapes)
231
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
232
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
233
+ print(f'Frozen params: Have {avail_numel} numels to process.')
234
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
235
+
236
+ total_params = 0
237
+ total_numel = 0
238
+ for name, shape in frozen_param_shapes.items():
239
+ total_params += 1
240
+ unpartitioned_numel = shape.numel()
241
+ total_numel += unpartitioned_numel
242
+
243
+ state_dict[name] = frozen_param_fragments[name]
244
+
245
+ if debug:
246
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
247
+
248
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
249
+
250
+
251
+ def _has_callable(obj, fn):
252
+ attr = getattr(obj, fn, None)
253
+ return callable(attr)
254
+
255
+
256
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
257
+ param_shapes = zero_model_states[0].param_shapes
258
+
259
+ # Reconstruction protocol:
260
+ #
261
+ # XXX: document this
262
+
263
+ if debug:
264
+ for i in range(world_size):
265
+ for j in range(len(fp32_flat_groups[0])):
266
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
267
+
268
+ # XXX: memory usage doubles here (zero2)
269
+ num_param_groups = len(fp32_flat_groups[0])
270
+ merged_single_partition_of_fp32_groups = []
271
+ for i in range(num_param_groups):
272
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
273
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
274
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
275
+ avail_numel = sum(
276
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
277
+
278
+ if debug:
279
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
280
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
281
+ # not asserting if there is a mismatch due to possible padding
282
+ print(f"Have {avail_numel} numels to process.")
283
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
284
+
285
+ # params
286
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
287
+ # out-of-core computing solution
288
+ total_numel = 0
289
+ total_params = 0
290
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
291
+ offset = 0
292
+ avail_numel = full_single_fp32_vector.numel()
293
+ for name, shape in shapes.items():
294
+
295
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
296
+ total_numel += unpartitioned_numel
297
+ total_params += 1
298
+
299
+ if debug:
300
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
301
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
302
+ offset += unpartitioned_numel
303
+
304
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
305
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
306
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
307
+ # live optimizer object, so we are checking that the numbers are within the right range
308
+ align_to = 2 * world_size
309
+
310
+ def zero2_align(x):
311
+ return align_to * math.ceil(x / align_to)
312
+
313
+ if debug:
314
+ print(f"original offset={offset}, avail_numel={avail_numel}")
315
+
316
+ offset = zero2_align(offset)
317
+ avail_numel = zero2_align(avail_numel)
318
+
319
+ if debug:
320
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
321
+
322
+ # Sanity check
323
+ if offset != avail_numel:
324
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
325
+
326
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
327
+
328
+
329
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
330
+ state_dict = OrderedDict()
331
+
332
+ # buffers
333
+ buffers = zero_model_states[0].buffers
334
+ state_dict.update(buffers)
335
+ if debug:
336
+ print(f"added {len(buffers)} buffers")
337
+
338
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
339
+
340
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
341
+
342
+ # recover shared parameters
343
+ for pair in zero_model_states[0].shared_params:
344
+ if pair[1] in state_dict:
345
+ state_dict[pair[0]] = state_dict[pair[1]]
346
+
347
+ return state_dict
348
+
349
+
350
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
351
+ remainder = unpartitioned_numel % world_size
352
+ padding_numel = (world_size - remainder) if remainder else 0
353
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
354
+ return partitioned_numel, padding_numel
355
+
356
+
357
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
358
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
359
+ return
360
+
361
+ if debug:
362
+ for i in range(world_size):
363
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
364
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
365
+
366
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
367
+ wanted_params = len(frozen_param_shapes)
368
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
369
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
370
+ print(f'Frozen params: Have {avail_numel} numels to process.')
371
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
372
+
373
+ total_params = 0
374
+ total_numel = 0
375
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
376
+ total_params += 1
377
+ unpartitioned_numel = shape.numel()
378
+ total_numel += unpartitioned_numel
379
+
380
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
381
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
382
+
383
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
384
+
385
+ if debug:
386
+ print(
387
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
388
+ )
389
+
390
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
391
+
392
+
393
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
394
+ param_shapes = zero_model_states[0].param_shapes
395
+ avail_numel = fp32_flat_groups[0].numel() * world_size
396
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
397
+ # param, re-consolidating each param, while dealing with padding if any
398
+
399
+ # merge list of dicts, preserving order
400
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
401
+
402
+ if debug:
403
+ for i in range(world_size):
404
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
405
+
406
+ wanted_params = len(param_shapes)
407
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
408
+ # not asserting if there is a mismatch due to possible padding
409
+ avail_numel = fp32_flat_groups[0].numel() * world_size
410
+ print(f"Trainable params: Have {avail_numel} numels to process.")
411
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
412
+
413
+ # params
414
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
415
+ # out-of-core computing solution
416
+ offset = 0
417
+ total_numel = 0
418
+ total_params = 0
419
+ for name, shape in param_shapes.items():
420
+
421
+ unpartitioned_numel = shape.numel()
422
+ total_numel += unpartitioned_numel
423
+ total_params += 1
424
+
425
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
426
+
427
+ if debug:
428
+ print(
429
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
430
+ )
431
+
432
+ # XXX: memory usage doubles here
433
+ state_dict[name] = torch.cat(
434
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
435
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
436
+ offset += partitioned_numel
437
+
438
+ offset *= world_size
439
+
440
+ # Sanity check
441
+ if offset != avail_numel:
442
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
443
+
444
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
445
+
446
+
447
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
448
+ state_dict = OrderedDict()
449
+
450
+ # buffers
451
+ buffers = zero_model_states[0].buffers
452
+ state_dict.update(buffers)
453
+ if debug:
454
+ print(f"added {len(buffers)} buffers")
455
+
456
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
457
+
458
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
459
+
460
+ # recover shared parameters
461
+ for pair in zero_model_states[0].shared_params:
462
+ if pair[1] in state_dict:
463
+ state_dict[pair[0]] = state_dict[pair[1]]
464
+
465
+ return state_dict
466
+
467
+
468
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
469
+ """
470
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
471
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
472
+ via a model hub.
473
+
474
+ Args:
475
+ - ``checkpoint_dir``: path to the desired checkpoint folder
476
+ - ``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``
477
+
478
+ Returns:
479
+ - pytorch ``state_dict``
480
+
481
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
482
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
483
+ the checkpoint.
484
+
485
+ A typical usage might be ::
486
+
487
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
488
+ # do the training and checkpoint saving
489
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
490
+ model = model.cpu() # move to cpu
491
+ model.load_state_dict(state_dict)
492
+ # submit to model hub or save the model to share with others
493
+
494
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
495
+ application. i.e. you will need to re-initialize the deepspeed engine, since
496
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
497
+
498
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
499
+
500
+ """
501
+ if tag is None:
502
+ latest_path = os.path.join(checkpoint_dir, 'latest')
503
+ if os.path.isfile(latest_path):
504
+ with open(latest_path, 'r') as fd:
505
+ tag = fd.read().strip()
506
+ else:
507
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
508
+
509
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
510
+
511
+ if not os.path.isdir(ds_checkpoint_dir):
512
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
513
+
514
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
515
+
516
+
517
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
518
+ """
519
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
520
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
521
+
522
+ Args:
523
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
524
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
525
+ - ``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``
526
+ """
527
+
528
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
529
+ print(f"Saving fp32 state dict to {output_file}")
530
+ torch.save(state_dict, output_file)
531
+
532
+
533
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
534
+ """
535
+ 1. Put the provided model to cpu
536
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
537
+ 3. Load it into the provided model
538
+
539
+ Args:
540
+ - ``model``: the model object to update
541
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
542
+ - ``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``
543
+
544
+ Returns:
545
+ - ``model`: modified model
546
+
547
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
548
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
549
+ conveniently placed for you in the checkpoint folder.
550
+
551
+ A typical usage might be ::
552
+
553
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
554
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
555
+ # submit to model hub or save the model to share with others
556
+
557
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
558
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
559
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
560
+
561
+ """
562
+ logger.info(f"Extracting fp32 weights")
563
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
564
+
565
+ logger.info(f"Overwriting model with fp32 weights")
566
+ model = model.cpu()
567
+ model.load_state_dict(state_dict, strict=False)
568
+
569
+ return model
570
+
571
+
572
+ if __name__ == "__main__":
573
+
574
+ parser = argparse.ArgumentParser()
575
+ parser.add_argument("checkpoint_dir",
576
+ type=str,
577
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
578
+ parser.add_argument(
579
+ "output_file",
580
+ type=str,
581
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
582
+ parser.add_argument("-t",
583
+ "--tag",
584
+ type=str,
585
+ default=None,
586
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
587
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
588
+ args = parser.parse_args()
589
+
590
+ debug = args.debug
591
+
592
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)