Hugging Face's logo

sbrzz
/
TinyLLaVA-SmolLM2-135M-Instruct-dinov2-small-pretrained

Visual Question Answering
Safetensors
English
tinyllava
Model card Files Community
sbrzz commited on
Commit
9cfd357
•
1 Parent(s): 96e5464

Upload 11 files

Browse files
Files changed (11) hide show
  1. config.json +111 -0
  2. generation_config.json +8 -0
  3. merges.txt +0 -0
  4. model.safetensors +3 -0
  5. rng_state.pth +3 -0
  6. scheduler.pt +3 -0
  7. special_tokens_map.json +28 -0
  8. tokenizer_config.json +157 -0
  9. trainer_state.json +3521 -0
  10. vocab.json +0 -0
  11. zero_to_fp32.py +604 -0
config.json ADDED
@@ -0,0 +1,111 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "TinyLlavaForConditionalGeneration"
4
+ ],
5
+ "cache_dir": null,
6
+ "connector_type": "mlp2x_gelu",
7
+ "hidden_size": 576,
8
+ "ignore_index": -100,
9
+ "image_aspect_ratio": "square",
10
+ "image_token_index": -200,
11
+ "llm_model_name_or_path": "HuggingFaceTB/SmolLM2-135M-Instruct",
12
+ "model_type": "tinyllava",
13
+ "num_queries": 128,
14
+ "num_resampler_layers": 3,
15
+ "pad_token": "<|endoftext|>",
16
+ "pad_token_id": 0,
17
+ "resampler_hidden_size": 768,
18
+ "text_config": {
19
+ "_name_or_path": "HuggingFaceTB/SmolLM2-135M-Instruct",
20
+ "architectures": [
21
+ "LlamaForCausalLM"
22
+ ],
23
+ "hidden_size": 576,
24
+ "initializer_range": 0.041666666666666664,
25
+ "intermediate_size": 1536,
26
+ "is_llama_config": true,
27
+ "max_position_embeddings": 8192,
28
+ "mlp_bias": false,
29
+ "model_type": "llama",
30
+ "num_attention_heads": 9,
31
+ "num_hidden_layers": 30,
32
+ "num_key_value_heads": 3,
33
+ "pad_token_id": 2,
34
+ "rms_norm_eps": 1e-05,
35
+ "rope_interleaved": false,
36
+ "rope_theta": 100000,
37
+ "tie_word_embeddings": true,
38
+ "torch_dtype": "float16",
39
+ "transformers.js_config": {
40
+ "kv_cache_dtype": {
41
+ "fp16": "float16",
42
+ "q4f16": "float16"
43
+ }
44
+ },
45
+ "vocab_size": 49152
46
+ },
47
+ "tokenizer_model_max_length": 2048,
48
+ "tokenizer_name_or_path": "HuggingFaceTB/SmolLM2-135M-Instruct",
49
+ "tokenizer_padding_side": "right",
50
+ "tokenizer_use_fast": false,
51
+ "torch_dtype": "float16",
52
+ "transformers_version": "4.39.3",
53
+ "tune_type_connector": "full",
54
+ "tune_type_llm": "frozen",
55
+ "tune_type_vision_tower": "frozen",
56
+ "tune_vision_tower_from_layer": 0,
57
+ "use_cache": false,
58
+ "vision_config": {
59
+ "_name_or_path": "facebook/dinov2-small",
60
+ "apply_layernorm": true,
61
+ "architectures": [
62
+ "Dinov2Model"
63
+ ],
64
+ "attention_probs_dropout_prob": 0.0,
65
+ "drop_path_rate": 0.0,
66
+ "hidden_act": "gelu",
67
+ "hidden_dropout_prob": 0.0,
68
+ "hidden_size": 384,
69
+ "image_size": 518,
70
+ "layer_norm_eps": 1e-06,
71
+ "layerscale_value": 1.0,
72
+ "mlp_ratio": 4,
73
+ "model_name_or_path": "facebook/dinov2-small",
74
+ "model_name_or_path2": "",
75
+ "model_type": "dinov2",
76
+ "num_attention_heads": 6,
77
+ "num_hidden_layers": 12,
78
+ "out_features": [
79
+ "stage12"
80
+ ],
81
+ "out_indices": [
82
+ 12
83
+ ],
84
+ "patch_size": 14,
85
+ "qkv_bias": true,
86
+ "reshape_hidden_states": true,
87
+ "stage_names": [
88
+ "stem",
89
+ "stage1",
90
+ "stage2",
91
+ "stage3",
92
+ "stage4",
93
+ "stage5",
94
+ "stage6",
95
+ "stage7",
96
+ "stage8",
97
+ "stage9",
98
+ "stage10",
99
+ "stage11",
100
+ "stage12"
101
+ ],
102
+ "torch_dtype": "float32",
103
+ "use_swiglu_ffn": false
104
+ },
105
+ "vision_feature_layer": -2,
106
+ "vision_feature_select_strategy": "patch",
107
+ "vision_hidden_size": 384,
108
+ "vision_model_name_or_path": "facebook/dinov2-small",
109
+ "vision_model_name_or_path2": "",
110
+ "vocab_size": 49152
111
+ }
generation_config.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "pad_token_id": 2,
6
+ "transformers_version": "4.39.3",
7
+ "use_cache": false
8
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98b8b99db344e92e6a37dd42a6d4c7874d7e1a990a4c991deb3231cc962b7424
3
+ size 314316256
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:835f869ea325fd6edf27b48b589309fb66641cb92b45f2fc13d1bb6e8814106c
3
+ size 14244
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:577fe24c67df92dec5a643aa1599690ef896ca41cbbb97802177c72073e27add
3
+ size 1064
special_tokens_map.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>"
5
+ ],
6
+ "bos_token": {
7
+ "content": "<|im_start|>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false
12
+ },
13
+ "eos_token": {
14
+ "content": "<|im_end|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false
19
+ },
20
+ "pad_token": "<|endoftext|>",
21
+ "unk_token": {
22
+ "content": "<|endoftext|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false
27
+ }
28
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,157 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "0": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "1": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "2": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "3": {
30
+ "content": "<repo_name>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "4": {
38
+ "content": "<reponame>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "5": {
46
+ "content": "<file_sep>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "6": {
54
+ "content": "<filename>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "7": {
62
+ "content": "<gh_stars>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "8": {
70
+ "content": "<issue_start>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "9": {
78
+ "content": "<issue_comment>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "10": {
86
+ "content": "<issue_closed>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "11": {
94
+ "content": "<jupyter_start>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "12": {
102
+ "content": "<jupyter_text>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "13": {
110
+ "content": "<jupyter_code>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "14": {
118
+ "content": "<jupyter_output>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": true
124
+ },
125
+ "15": {
126
+ "content": "<jupyter_script>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": true
132
+ },
133
+ "16": {
134
+ "content": "<empty_output>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": true
140
+ }
141
+ },
142
+ "additional_special_tokens": [
143
+ "<|im_start|>",
144
+ "<|im_end|>"
145
+ ],
146
+ "bos_token": "<|im_start|>",
147
+ "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful AI assistant named SmolLM, trained by Hugging Face<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
148
+ "clean_up_tokenization_spaces": false,
149
+ "eos_token": "<|im_end|>",
150
+ "errors": "replace",
151
+ "model_max_length": 2048,
152
+ "pad_token": "<|endoftext|>",
153
+ "padding_side": "right",
154
+ "tokenizer_class": "GPT2Tokenizer",
155
+ "unk_token": "<|endoftext|>",
156
+ "vocab_size": 49152
157
+ }
trainer_state.json ADDED
@@ -0,0 +1,3521 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.11466574934067195,
5
+ "eval_steps": 500,
6
+ "global_step": 500,
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": 0.0,
14
+ "learning_rate": 0.0,
15
+ "loss": 7.2451,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.0,
20
+ "grad_norm": 0.0,
21
+ "learning_rate": 0.0,
22
+ "loss": 7.3192,
23
+ "step": 2
24
+ },
25
+ {
26
+ "epoch": 0.0,
27
+ "grad_norm": 14.819024134660662,
28
+ "learning_rate": 1.3333333333333334e-06,
29
+ "loss": 7.2901,
30
+ "step": 3
31
+ },
32
+ {
33
+ "epoch": 0.0,
34
+ "grad_norm": 14.819024134660662,
35
+ "learning_rate": 1.3333333333333334e-06,
36
+ "loss": 7.4127,
37
+ "step": 4
38
+ },
39
+ {
40
+ "epoch": 0.0,
41
+ "grad_norm": 14.199093979418544,
42
+ "learning_rate": 2.666666666666667e-06,
43
+ "loss": 7.4321,
44
+ "step": 5
45
+ },
46
+ {
47
+ "epoch": 0.0,
48
+ "grad_norm": 10.185841979858534,
49
+ "learning_rate": 4.000000000000001e-06,
50
+ "loss": 7.1568,
51
+ "step": 6
52
+ },
53
+ {
54
+ "epoch": 0.0,
55
+ "grad_norm": 39.91704954913313,
56
+ "learning_rate": 5.333333333333334e-06,
57
+ "loss": 7.2837,
58
+ "step": 7
59
+ },
60
+ {
61
+ "epoch": 0.0,
62
+ "grad_norm": 18.930408398058322,
63
+ "learning_rate": 6.666666666666667e-06,
64
+ "loss": 7.2329,
65
+ "step": 8
66
+ },
67
+ {
68
+ "epoch": 0.0,
69
+ "grad_norm": 8.865495224195184,
70
+ "learning_rate": 8.000000000000001e-06,
71
+ "loss": 7.1651,
72
+ "step": 9
73
+ },
74
+ {
75
+ "epoch": 0.0,
76
+ "grad_norm": 35.166574749663724,
77
+ "learning_rate": 9.333333333333334e-06,
78
+ "loss": 7.1455,
79
+ "step": 10
80
+ },
81
+ {
82
+ "epoch": 0.0,
83
+ "grad_norm": 14.929376279495122,
84
+ "learning_rate": 1.0666666666666667e-05,
85
+ "loss": 7.2938,
86
+ "step": 11
87
+ },
88
+ {
89
+ "epoch": 0.0,
90
+ "grad_norm": 14.323589623470497,
91
+ "learning_rate": 1.2e-05,
92
+ "loss": 7.136,
93
+ "step": 12
94
+ },
95
+ {
96
+ "epoch": 0.0,
97
+ "grad_norm": 12.033497318446193,
98
+ "learning_rate": 1.3333333333333333e-05,
99
+ "loss": 7.2873,
100
+ "step": 13
101
+ },
102
+ {
103
+ "epoch": 0.0,
104
+ "grad_norm": 12.713515670882014,
105
+ "learning_rate": 1.4666666666666666e-05,
106
+ "loss": 7.1778,
107
+ "step": 14
108
+ },
109
+ {
110
+ "epoch": 0.0,
111
+ "grad_norm": 16.40476702849706,
112
+ "learning_rate": 1.6000000000000003e-05,
113
+ "loss": 7.3031,
114
+ "step": 15
115
+ },
116
+ {
117
+ "epoch": 0.0,
118
+ "grad_norm": 16.463860488428647,
119
+ "learning_rate": 1.7333333333333336e-05,
120
+ "loss": 6.958,
121
+ "step": 16
122
+ },
123
+ {
124
+ "epoch": 0.0,
125
+ "grad_norm": 11.114739607650908,
126
+ "learning_rate": 1.866666666666667e-05,
127
+ "loss": 7.328,
128
+ "step": 17
129
+ },
130
+ {
131
+ "epoch": 0.0,
132
+ "grad_norm": 8.560556168903155,
133
+ "learning_rate": 2e-05,
134
+ "loss": 6.8535,
135
+ "step": 18
136
+ },
137
+ {
138
+ "epoch": 0.0,
139
+ "grad_norm": 11.543963578227485,
140
+ "learning_rate": 1.999979021001399e-05,
141
+ "loss": 7.1039,
142
+ "step": 19
143
+ },
144
+ {
145
+ "epoch": 0.0,
146
+ "grad_norm": 13.024267824258278,
147
+ "learning_rate": 1.999916084885832e-05,
148
+ "loss": 7.03,
149
+ "step": 20
150
+ },
151
+ {
152
+ "epoch": 0.0,
153
+ "grad_norm": 24.347656967541983,
154
+ "learning_rate": 1.9998111942939727e-05,
155
+ "loss": 7.1369,
156
+ "step": 21
157
+ },
158
+ {
159
+ "epoch": 0.01,
160
+ "grad_norm": 13.978477853833892,
161
+ "learning_rate": 1.9996643536268202e-05,
162
+ "loss": 7.1052,
163
+ "step": 22
164
+ },
165
+ {
166
+ "epoch": 0.01,
167
+ "grad_norm": 9.01222616499256,
168
+ "learning_rate": 1.9994755690455154e-05,
169
+ "loss": 7.2016,
170
+ "step": 23
171
+ },
172
+ {
173
+ "epoch": 0.01,
174
+ "grad_norm": 8.338853855265398,
175
+ "learning_rate": 1.99924484847108e-05,
176
+ "loss": 7.0447,
177
+ "step": 24
178
+ },
179
+ {
180
+ "epoch": 0.01,
181
+ "grad_norm": 14.599634725835617,
182
+ "learning_rate": 1.998972201584088e-05,
183
+ "loss": 7.0037,
184
+ "step": 25
185
+ },
186
+ {
187
+ "epoch": 0.01,
188
+ "grad_norm": 8.332198683157735,
189
+ "learning_rate": 1.9986576398242566e-05,
190
+ "loss": 7.1784,
191
+ "step": 26
192
+ },
193
+ {
194
+ "epoch": 0.01,
195
+ "grad_norm": 14.23890995116397,
196
+ "learning_rate": 1.9983011763899674e-05,
197
+ "loss": 6.8922,
198
+ "step": 27
199
+ },
200
+ {
201
+ "epoch": 0.01,
202
+ "grad_norm": 8.923950492978154,
203
+ "learning_rate": 1.997902826237712e-05,
204
+ "loss": 6.9407,
205
+ "step": 28
206
+ },
207
+ {
208
+ "epoch": 0.01,
209
+ "grad_norm": 8.06378020962777,
210
+ "learning_rate": 1.997462606081465e-05,
211
+ "loss": 7.1025,
212
+ "step": 29
213
+ },
214
+ {
215
+ "epoch": 0.01,
216
+ "grad_norm": 11.24345210949927,
217
+ "learning_rate": 1.9969805343919822e-05,
218
+ "loss": 6.9762,
219
+ "step": 30
220
+ },
221
+ {
222
+ "epoch": 0.01,
223
+ "grad_norm": 8.092214633553768,
224
+ "learning_rate": 1.9964566313960265e-05,
225
+ "loss": 7.0578,
226
+ "step": 31
227
+ },
228
+ {
229
+ "epoch": 0.01,
230
+ "grad_norm": 12.87228727478449,
231
+ "learning_rate": 1.995890919075519e-05,
232
+ "loss": 6.8295,
233
+ "step": 32
234
+ },
235
+ {
236
+ "epoch": 0.01,
237
+ "grad_norm": 7.111093086376143,
238
+ "learning_rate": 1.995283421166614e-05,
239
+ "loss": 6.7973,
240
+ "step": 33
241
+ },
242
+ {
243
+ "epoch": 0.01,
244
+ "grad_norm": 8.632087862294554,
245
+ "learning_rate": 1.9946341631587086e-05,
246
+ "loss": 6.7752,
247
+ "step": 34
248
+ },
249
+ {
250
+ "epoch": 0.01,
251
+ "grad_norm": 8.994822315110934,
252
+ "learning_rate": 1.9939431722933678e-05,
253
+ "loss": 7.0197,
254
+ "step": 35
255
+ },
256
+ {
257
+ "epoch": 0.01,
258
+ "grad_norm": 7.486951908393647,
259
+ "learning_rate": 1.9932104775631847e-05,
260
+ "loss": 6.8283,
261
+ "step": 36
262
+ },
263
+ {
264
+ "epoch": 0.01,
265
+ "grad_norm": 7.4584761693908375,
266
+ "learning_rate": 1.9924361097105624e-05,
267
+ "loss": 6.8002,
268
+ "step": 37
269
+ },
270
+ {
271
+ "epoch": 0.01,
272
+ "grad_norm": 6.261343208572681,
273
+ "learning_rate": 1.9916201012264255e-05,
274
+ "loss": 6.6381,
275
+ "step": 38
276
+ },
277
+ {
278
+ "epoch": 0.01,
279
+ "grad_norm": 18.79112809499115,
280
+ "learning_rate": 1.990762486348855e-05,
281
+ "loss": 7.01,
282
+ "step": 39
283
+ },
284
+ {
285
+ "epoch": 0.01,
286
+ "grad_norm": 7.150915500939425,
287
+ "learning_rate": 1.989863301061654e-05,
288
+ "loss": 6.9012,
289
+ "step": 40
290
+ },
291
+ {
292
+ "epoch": 0.01,
293
+ "grad_norm": 6.738925101776216,
294
+ "learning_rate": 1.9889225830928365e-05,
295
+ "loss": 6.7694,
296
+ "step": 41
297
+ },
298
+ {
299
+ "epoch": 0.01,
300
+ "grad_norm": 6.944529789995812,
301
+ "learning_rate": 1.987940371913044e-05,
302
+ "loss": 6.799,
303
+ "step": 42
304
+ },
305
+ {
306
+ "epoch": 0.01,
307
+ "grad_norm": 7.556349878373683,
308
+ "learning_rate": 1.9869167087338908e-05,
309
+ "loss": 6.6428,
310
+ "step": 43
311
+ },
312
+ {
313
+ "epoch": 0.01,
314
+ "grad_norm": 6.611474483622349,
315
+ "learning_rate": 1.9858516365062334e-05,
316
+ "loss": 6.6243,
317
+ "step": 44
318
+ },
319
+ {
320
+ "epoch": 0.01,
321
+ "grad_norm": 6.916298787571706,
322
+ "learning_rate": 1.9847451999183692e-05,
323
+ "loss": 6.6087,
324
+ "step": 45
325
+ },
326
+ {
327
+ "epoch": 0.01,
328
+ "grad_norm": 8.457565509913838,
329
+ "learning_rate": 1.9835974453941623e-05,
330
+ "loss": 6.7551,
331
+ "step": 46
332
+ },
333
+ {
334
+ "epoch": 0.01,
335
+ "grad_norm": 8.540062078386342,
336
+ "learning_rate": 1.9824084210910924e-05,
337
+ "loss": 6.6972,
338
+ "step": 47
339
+ },
340
+ {
341
+ "epoch": 0.01,
342
+ "grad_norm": 7.4270052163791185,
343
+ "learning_rate": 1.9811781768982392e-05,
344
+ "loss": 6.9304,
345
+ "step": 48
346
+ },
347
+ {
348
+ "epoch": 0.01,
349
+ "grad_norm": 7.31190391936824,
350
+ "learning_rate": 1.9799067644341844e-05,
351
+ "loss": 6.7148,
352
+ "step": 49
353
+ },
354
+ {
355
+ "epoch": 0.01,
356
+ "grad_norm": 6.181290414236328,
357
+ "learning_rate": 1.978594237044849e-05,
358
+ "loss": 6.6086,
359
+ "step": 50
360
+ },
361
+ {
362
+ "epoch": 0.01,
363
+ "grad_norm": 6.059093545289046,
364
+ "learning_rate": 1.977240649801253e-05,
365
+ "loss": 6.5689,
366
+ "step": 51
367
+ },
368
+ {
369
+ "epoch": 0.01,
370
+ "grad_norm": 7.376855166703994,
371
+ "learning_rate": 1.9758460594972068e-05,
372
+ "loss": 6.5767,
373
+ "step": 52
374
+ },
375
+ {
376
+ "epoch": 0.01,
377
+ "grad_norm": 9.315509966882654,
378
+ "learning_rate": 1.9744105246469264e-05,
379
+ "loss": 6.601,
380
+ "step": 53
381
+ },
382
+ {
383
+ "epoch": 0.01,
384
+ "grad_norm": 7.110322978482923,
385
+ "learning_rate": 1.9729341054825783e-05,
386
+ "loss": 6.5206,
387
+ "step": 54
388
+ },
389
+ {
390
+ "epoch": 0.01,
391
+ "grad_norm": 8.348814181636438,
392
+ "learning_rate": 1.9714168639517543e-05,
393
+ "loss": 6.6781,
394
+ "step": 55
395
+ },
396
+ {
397
+ "epoch": 0.01,
398
+ "grad_norm": 7.889124746527842,
399
+ "learning_rate": 1.9698588637148705e-05,
400
+ "loss": 6.363,
401
+ "step": 56
402
+ },
403
+ {
404
+ "epoch": 0.01,
405
+ "grad_norm": 6.048373264000878,
406
+ "learning_rate": 1.9682601701424958e-05,
407
+ "loss": 6.5605,
408
+ "step": 57
409
+ },
410
+ {
411
+ "epoch": 0.01,
412
+ "grad_norm": 6.666923562210865,
413
+ "learning_rate": 1.9666208503126115e-05,
414
+ "loss": 6.5264,
415
+ "step": 58
416
+ },
417
+ {
418
+ "epoch": 0.01,
419
+ "grad_norm": 6.215144195698015,
420
+ "learning_rate": 1.9649409730077934e-05,
421
+ "loss": 6.7055,
422
+ "step": 59
423
+ },
424
+ {
425
+ "epoch": 0.01,
426
+ "grad_norm": 5.851763399842508,
427
+ "learning_rate": 1.9632206087123296e-05,
428
+ "loss": 6.3156,
429
+ "step": 60
430
+ },
431
+ {
432
+ "epoch": 0.01,
433
+ "grad_norm": 6.899818690987695,
434
+ "learning_rate": 1.9614598296092603e-05,
435
+ "loss": 6.5751,
436
+ "step": 61
437
+ },
438
+ {
439
+ "epoch": 0.01,
440
+ "grad_norm": 6.233556175977335,
441
+ "learning_rate": 1.9596587095773496e-05,
442
+ "loss": 6.3861,
443
+ "step": 62
444
+ },
445
+ {
446
+ "epoch": 0.01,
447
+ "grad_norm": 6.677665311671007,
448
+ "learning_rate": 1.957817324187987e-05,
449
+ "loss": 6.5699,
450
+ "step": 63
451
+ },
452
+ {
453
+ "epoch": 0.01,
454
+ "grad_norm": 6.655226097949778,
455
+ "learning_rate": 1.9559357507020163e-05,
456
+ "loss": 6.5021,
457
+ "step": 64
458
+ },
459
+ {
460
+ "epoch": 0.01,
461
+ "grad_norm": 7.47281811823693,
462
+ "learning_rate": 1.9540140680664915e-05,
463
+ "loss": 6.5324,
464
+ "step": 65
465
+ },
466
+ {
467
+ "epoch": 0.02,
468
+ "grad_norm": 5.6360351804666475,
469
+ "learning_rate": 1.952052356911368e-05,
470
+ "loss": 6.6096,
471
+ "step": 66
472
+ },
473
+ {
474
+ "epoch": 0.02,
475
+ "grad_norm": 5.4238841079756615,
476
+ "learning_rate": 1.950050699546116e-05,
477
+ "loss": 6.2913,
478
+ "step": 67
479
+ },
480
+ {
481
+ "epoch": 0.02,
482
+ "grad_norm": 6.691271836321062,
483
+ "learning_rate": 1.9480091799562706e-05,
484
+ "loss": 6.6404,
485
+ "step": 68
486
+ },
487
+ {
488
+ "epoch": 0.02,
489
+ "grad_norm": 5.779952770675111,
490
+ "learning_rate": 1.9459278837999048e-05,
491
+ "loss": 6.5131,
492
+ "step": 69
493
+ },
494
+ {
495
+ "epoch": 0.02,
496
+ "grad_norm": 5.75641781132488,
497
+ "learning_rate": 1.9438068984040366e-05,
498
+ "loss": 6.4102,
499
+ "step": 70
500
+ },
501
+ {
502
+ "epoch": 0.02,
503
+ "grad_norm": 5.473240945440776,
504
+ "learning_rate": 1.9416463127609655e-05,
505
+ "loss": 6.1943,
506
+ "step": 71
507
+ },
508
+ {
509
+ "epoch": 0.02,
510
+ "grad_norm": 5.639227300435135,
511
+ "learning_rate": 1.9394462175245382e-05,
512
+ "loss": 6.2661,
513
+ "step": 72
514
+ },
515
+ {
516
+ "epoch": 0.02,
517
+ "grad_norm": 6.359397311850186,
518
+ "learning_rate": 1.937206705006344e-05,
519
+ "loss": 6.3749,
520
+ "step": 73
521
+ },
522
+ {
523
+ "epoch": 0.02,
524
+ "grad_norm": 5.9168823793145835,
525
+ "learning_rate": 1.9349278691718426e-05,
526
+ "loss": 6.4857,
527
+ "step": 74
528
+ },
529
+ {
530
+ "epoch": 0.02,
531
+ "grad_norm": 5.934252295974776,
532
+ "learning_rate": 1.9326098056364224e-05,
533
+ "loss": 6.4401,
534
+ "step": 75
535
+ },
536
+ {
537
+ "epoch": 0.02,
538
+ "grad_norm": 5.184345891936555,
539
+ "learning_rate": 1.9302526116613863e-05,
540
+ "loss": 6.272,
541
+ "step": 76
542
+ },
543
+ {
544
+ "epoch": 0.02,
545
+ "grad_norm": 6.42618267064823,
546
+ "learning_rate": 1.9278563861498726e-05,
547
+ "loss": 6.4017,
548
+ "step": 77
549
+ },
550
+ {
551
+ "epoch": 0.02,
552
+ "grad_norm": 5.698764300135597,
553
+ "learning_rate": 1.9254212296427043e-05,
554
+ "loss": 6.422,
555
+ "step": 78
556
+ },
557
+ {
558
+ "epoch": 0.02,
559
+ "grad_norm": 5.515269918748001,
560
+ "learning_rate": 1.922947244314172e-05,
561
+ "loss": 6.4018,
562
+ "step": 79
563
+ },
564
+ {
565
+ "epoch": 0.02,
566
+ "grad_norm": 5.186220316136429,
567
+ "learning_rate": 1.9204345339677442e-05,
568
+ "loss": 6.4008,
569
+ "step": 80
570
+ },
571
+ {
572
+ "epoch": 0.02,
573
+ "grad_norm": 6.071075169984412,
574
+ "learning_rate": 1.9178832040317153e-05,
575
+ "loss": 6.236,
576
+ "step": 81
577
+ },
578
+ {
579
+ "epoch": 0.02,
580
+ "grad_norm": 5.211523270674776,
581
+ "learning_rate": 1.91529336155478e-05,
582
+ "loss": 6.2431,
583
+ "step": 82
584
+ },
585
+ {
586
+ "epoch": 0.02,
587
+ "grad_norm": 4.654030406820622,
588
+ "learning_rate": 1.9126651152015404e-05,
589
+ "loss": 6.1377,
590
+ "step": 83
591
+ },
592
+ {
593
+ "epoch": 0.02,
594
+ "grad_norm": 5.542618960137867,
595
+ "learning_rate": 1.9099985752479505e-05,
596
+ "loss": 6.3531,
597
+ "step": 84
598
+ },
599
+ {
600
+ "epoch": 0.02,
601
+ "grad_norm": 4.807919282457549,
602
+ "learning_rate": 1.9072938535766864e-05,
603
+ "loss": 6.2508,
604
+ "step": 85
605
+ },
606
+ {
607
+ "epoch": 0.02,
608
+ "grad_norm": 4.852730067283944,
609
+ "learning_rate": 1.904551063672452e-05,
610
+ "loss": 6.2348,
611
+ "step": 86
612
+ },
613
+ {
614
+ "epoch": 0.02,
615
+ "grad_norm": 4.925798203153842,
616
+ "learning_rate": 1.9017703206172187e-05,
617
+ "loss": 6.2765,
618
+ "step": 87
619
+ },
620
+ {
621
+ "epoch": 0.02,
622
+ "grad_norm": 5.0975569371666865,
623
+ "learning_rate": 1.8989517410853956e-05,
624
+ "loss": 6.2451,
625
+ "step": 88
626
+ },
627
+ {
628
+ "epoch": 0.02,
629
+ "grad_norm": 4.699489385610056,
630
+ "learning_rate": 1.896095443338935e-05,
631
+ "loss": 6.2578,
632
+ "step": 89
633
+ },
634
+ {
635
+ "epoch": 0.02,
636
+ "grad_norm": 5.660080834025456,
637
+ "learning_rate": 1.8932015472223692e-05,
638
+ "loss": 6.2843,
639
+ "step": 90
640
+ },
641
+ {
642
+ "epoch": 0.02,
643
+ "grad_norm": 4.9661954295884225,
644
+ "learning_rate": 1.8902701741577844e-05,
645
+ "loss": 6.1715,
646
+ "step": 91
647
+ },
648
+ {
649
+ "epoch": 0.02,
650
+ "grad_norm": 4.888648288627258,
651
+ "learning_rate": 1.8873014471397225e-05,
652
+ "loss": 6.2495,
653
+ "step": 92
654
+ },
655
+ {
656
+ "epoch": 0.02,
657
+ "grad_norm": 4.710120588181978,
658
+ "learning_rate": 1.8842954907300236e-05,
659
+ "loss": 5.9913,
660
+ "step": 93
661
+ },
662
+ {
663
+ "epoch": 0.02,
664
+ "grad_norm": 5.155284987315298,
665
+ "learning_rate": 1.881252431052599e-05,
666
+ "loss": 6.3151,
667
+ "step": 94
668
+ },
669
+ {
670
+ "epoch": 0.02,
671
+ "grad_norm": 5.013674756768704,
672
+ "learning_rate": 1.8781723957881374e-05,
673
+ "loss": 6.2286,
674
+ "step": 95
675
+ },
676
+ {
677
+ "epoch": 0.02,
678
+ "grad_norm": 4.011825376784628,
679
+ "learning_rate": 1.87505551416875e-05,
680
+ "loss": 6.2459,
681
+ "step": 96
682
+ },
683
+ {
684
+ "epoch": 0.02,
685
+ "grad_norm": 4.835636123417131,
686
+ "learning_rate": 1.871901916972547e-05,
687
+ "loss": 6.2033,
688
+ "step": 97
689
+ },
690
+ {
691
+ "epoch": 0.02,
692
+ "grad_norm": 5.171704259527532,
693
+ "learning_rate": 1.8687117365181514e-05,
694
+ "loss": 6.2608,
695
+ "step": 98
696
+ },
697
+ {
698
+ "epoch": 0.02,
699
+ "grad_norm": 5.181750952717401,
700
+ "learning_rate": 1.865485106659145e-05,
701
+ "loss": 6.0338,
702
+ "step": 99
703
+ },
704
+ {
705
+ "epoch": 0.02,
706
+ "grad_norm": 4.687033388755913,
707
+ "learning_rate": 1.862222162778454e-05,
708
+ "loss": 6.1161,
709
+ "step": 100
710
+ },
711
+ {
712
+ "epoch": 0.02,
713
+ "grad_norm": 5.061452310176261,
714
+ "learning_rate": 1.85892304178267e-05,
715
+ "loss": 6.1018,
716
+ "step": 101
717
+ },
718
+ {
719
+ "epoch": 0.02,
720
+ "grad_norm": 4.742767764167209,
721
+ "learning_rate": 1.8555878820963014e-05,
722
+ "loss": 6.2337,
723
+ "step": 102
724
+ },
725
+ {
726
+ "epoch": 0.02,
727
+ "grad_norm": 5.294660524784012,
728
+ "learning_rate": 1.8522168236559693e-05,
729
+ "loss": 6.2453,
730
+ "step": 103
731
+ },
732
+ {
733
+ "epoch": 0.02,
734
+ "grad_norm": 4.1375219719939915,
735
+ "learning_rate": 1.8488100079045345e-05,
736
+ "loss": 6.1595,
737
+ "step": 104
738
+ },
739
+ {
740
+ "epoch": 0.02,
741
+ "grad_norm": 4.537495039210601,
742
+ "learning_rate": 1.8453675777851627e-05,
743
+ "loss": 6.1449,
744
+ "step": 105
745
+ },
746
+ {
747
+ "epoch": 0.02,
748
+ "grad_norm": 5.402914196313115,
749
+ "learning_rate": 1.8418896777353272e-05,
750
+ "loss": 6.2237,
751
+ "step": 106
752
+ },
753
+ {
754
+ "epoch": 0.02,
755
+ "grad_norm": 4.1771424603623615,
756
+ "learning_rate": 1.8383764536807486e-05,
757
+ "loss": 6.1728,
758
+ "step": 107
759
+ },
760
+ {
761
+ "epoch": 0.02,
762
+ "grad_norm": 4.198851233644863,
763
+ "learning_rate": 1.8348280530292712e-05,
764
+ "loss": 6.0628,
765
+ "step": 108
766
+ },
767
+ {
768
+ "epoch": 0.02,
769
+ "grad_norm": 4.559214073625153,
770
+ "learning_rate": 1.831244624664681e-05,
771
+ "loss": 6.2743,
772
+ "step": 109
773
+ },
774
+ {
775
+ "epoch": 0.03,
776
+ "grad_norm": 3.97241616226669,
777
+ "learning_rate": 1.827626318940454e-05,
778
+ "loss": 6.1513,
779
+ "step": 110
780
+ },
781
+ {
782
+ "epoch": 0.03,
783
+ "grad_norm": 4.029464327696206,
784
+ "learning_rate": 1.8239732876734525e-05,
785
+ "loss": 6.1743,
786
+ "step": 111
787
+ },
788
+ {
789
+ "epoch": 0.03,
790
+ "grad_norm": 4.9362284691818905,
791
+ "learning_rate": 1.8202856841375517e-05,
792
+ "loss": 6.0945,
793
+ "step": 112
794
+ },
795
+ {
796
+ "epoch": 0.03,
797
+ "grad_norm": 5.525591589138074,
798
+ "learning_rate": 1.816563663057211e-05,
799
+ "loss": 6.2648,
800
+ "step": 113
801
+ },
802
+ {
803
+ "epoch": 0.03,
804
+ "grad_norm": 3.913078134221513,
805
+ "learning_rate": 1.81280738060098e-05,
806
+ "loss": 6.0767,
807
+ "step": 114
808
+ },
809
+ {
810
+ "epoch": 0.03,
811
+ "grad_norm": 4.539776653253226,
812
+ "learning_rate": 1.8090169943749477e-05,
813
+ "loss": 6.2008,
814
+ "step": 115
815
+ },
816
+ {
817
+ "epoch": 0.03,
818
+ "grad_norm": 7.07561003185471,
819
+ "learning_rate": 1.8051926634161282e-05,
820
+ "loss": 6.1817,
821
+ "step": 116
822
+ },
823
+ {
824
+ "epoch": 0.03,
825
+ "grad_norm": 4.4932679628420376,
826
+ "learning_rate": 1.8013345481857903e-05,
827
+ "loss": 6.0112,
828
+ "step": 117
829
+ },
830
+ {
831
+ "epoch": 0.03,
832
+ "grad_norm": 4.728104224718154,
833
+ "learning_rate": 1.797442810562721e-05,
834
+ "loss": 6.1642,
835
+ "step": 118
836
+ },
837
+ {
838
+ "epoch": 0.03,
839
+ "grad_norm": 5.010653662783029,
840
+ "learning_rate": 1.793517613836437e-05,
841
+ "loss": 6.0345,
842
+ "step": 119
843
+ },
844
+ {
845
+ "epoch": 0.03,
846
+ "grad_norm": 4.58293874507086,
847
+ "learning_rate": 1.7895591227003316e-05,
848
+ "loss": 6.0006,
849
+ "step": 120
850
+ },
851
+ {
852
+ "epoch": 0.03,
853
+ "grad_norm": 4.357889872637869,
854
+ "learning_rate": 1.7855675032447648e-05,
855
+ "loss": 6.0577,
856
+ "step": 121
857
+ },
858
+ {
859
+ "epoch": 0.03,
860
+ "grad_norm": 4.095191120284703,
861
+ "learning_rate": 1.7815429229500946e-05,
862
+ "loss": 5.9336,
863
+ "step": 122
864
+ },
865
+ {
866
+ "epoch": 0.03,
867
+ "grad_norm": 4.736888424569404,
868
+ "learning_rate": 1.7774855506796497e-05,
869
+ "loss": 5.961,
870
+ "step": 123
871
+ },
872
+ {
873
+ "epoch": 0.03,
874
+ "grad_norm": 9.32933189803368,
875
+ "learning_rate": 1.7733955566726438e-05,
876
+ "loss": 5.8729,
877
+ "step": 124
878
+ },
879
+ {
880
+ "epoch": 0.03,
881
+ "grad_norm": 4.352013234575603,
882
+ "learning_rate": 1.7692731125370355e-05,
883
+ "loss": 5.9631,
884
+ "step": 125
885
+ },
886
+ {
887
+ "epoch": 0.03,
888
+ "grad_norm": 4.0044514035329986,
889
+ "learning_rate": 1.7651183912423228e-05,
890
+ "loss": 6.0128,
891
+ "step": 126
892
+ },
893
+ {
894
+ "epoch": 0.03,
895
+ "grad_norm": 4.158183905120493,
896
+ "learning_rate": 1.7609315671122912e-05,
897
+ "loss": 6.0247,
898
+ "step": 127
899
+ },
900
+ {
901
+ "epoch": 0.03,
902
+ "grad_norm": 4.257943967689872,
903
+ "learning_rate": 1.7567128158176955e-05,
904
+ "loss": 6.2873,
905
+ "step": 128
906
+ },
907
+ {
908
+ "epoch": 0.03,
909
+ "grad_norm": 4.1763697860525655,
910
+ "learning_rate": 1.7524623143688905e-05,
911
+ "loss": 5.9392,
912
+ "step": 129
913
+ },
914
+ {
915
+ "epoch": 0.03,
916
+ "grad_norm": 3.9607955760999056,
917
+ "learning_rate": 1.748180241108404e-05,
918
+ "loss": 5.8487,
919
+ "step": 130
920
+ },
921
+ {
922
+ "epoch": 0.03,
923
+ "grad_norm": 4.0896919573576875,
924
+ "learning_rate": 1.7438667757034547e-05,
925
+ "loss": 6.0042,
926
+ "step": 131
927
+ },
928
+ {
929
+ "epoch": 0.03,
930
+ "grad_norm": 4.282944803020933,
931
+ "learning_rate": 1.739522099138411e-05,
932
+ "loss": 6.1652,
933
+ "step": 132
934
+ },
935
+ {
936
+ "epoch": 0.03,
937
+ "grad_norm": 4.13081574380951,
938
+ "learning_rate": 1.7351463937072008e-05,
939
+ "loss": 6.0233,
940
+ "step": 133
941
+ },
942
+ {
943
+ "epoch": 0.03,
944
+ "grad_norm": 4.96960839892131,
945
+ "learning_rate": 1.7307398430056595e-05,
946
+ "loss": 5.7853,
947
+ "step": 134
948
+ },
949
+ {
950
+ "epoch": 0.03,
951
+ "grad_norm": 4.459520148451556,
952
+ "learning_rate": 1.72630263192383e-05,
953
+ "loss": 5.9414,
954
+ "step": 135
955
+ },
956
+ {
957
+ "epoch": 0.03,
958
+ "grad_norm": 3.7882596771040893,
959
+ "learning_rate": 1.7218349466382024e-05,
960
+ "loss": 5.8131,
961
+ "step": 136
962
+ },
963
+ {
964
+ "epoch": 0.03,
965
+ "grad_norm": 4.207888044625232,
966
+ "learning_rate": 1.7173369746039026e-05,
967
+ "loss": 5.8724,
968
+ "step": 137
969
+ },
970
+ {
971
+ "epoch": 0.03,
972
+ "grad_norm": 3.6658140789192695,
973
+ "learning_rate": 1.7128089045468294e-05,
974
+ "loss": 5.9003,
975
+ "step": 138
976
+ },
977
+ {
978
+ "epoch": 0.03,
979
+ "grad_norm": 3.7722042701323493,
980
+ "learning_rate": 1.7082509264557333e-05,
981
+ "loss": 5.8756,
982
+ "step": 139
983
+ },
984
+ {
985
+ "epoch": 0.03,
986
+ "grad_norm": 4.8972909846842185,
987
+ "learning_rate": 1.7036632315742464e-05,
988
+ "loss": 6.0623,
989
+ "step": 140
990
+ },
991
+ {
992
+ "epoch": 0.03,
993
+ "grad_norm": 3.703243911637958,
994
+ "learning_rate": 1.6990460123928577e-05,
995
+ "loss": 5.9731,
996
+ "step": 141
997
+ },
998
+ {
999
+ "epoch": 0.03,
1000
+ "grad_norm": 4.040351904365242,
1001
+ "learning_rate": 1.6943994626408365e-05,
1002
+ "loss": 6.0629,
1003
+ "step": 142
1004
+ },
1005
+ {
1006
+ "epoch": 0.03,
1007
+ "grad_norm": 7.5014912670407865,
1008
+ "learning_rate": 1.6897237772781046e-05,
1009
+ "loss": 5.945,
1010
+ "step": 143
1011
+ },
1012
+ {
1013
+ "epoch": 0.03,
1014
+ "grad_norm": 4.183607320000293,
1015
+ "learning_rate": 1.6850191524870548e-05,
1016
+ "loss": 5.906,
1017
+ "step": 144
1018
+ },
1019
+ {
1020
+ "epoch": 0.03,
1021
+ "grad_norm": 4.441179437405616,
1022
+ "learning_rate": 1.6802857856643214e-05,
1023
+ "loss": 5.9791,
1024
+ "step": 145
1025
+ },
1026
+ {
1027
+ "epoch": 0.03,
1028
+ "grad_norm": 3.5946652188893187,
1029
+ "learning_rate": 1.6755238754124965e-05,
1030
+ "loss": 5.8481,
1031
+ "step": 146
1032
+ },
1033
+ {
1034
+ "epoch": 0.03,
1035
+ "grad_norm": 5.030328438406491,
1036
+ "learning_rate": 1.6707336215317968e-05,
1037
+ "loss": 6.047,
1038
+ "step": 147
1039
+ },
1040
+ {
1041
+ "epoch": 0.03,
1042
+ "grad_norm": 4.048177117370896,
1043
+ "learning_rate": 1.665915225011681e-05,
1044
+ "loss": 5.9285,
1045
+ "step": 148
1046
+ },
1047
+ {
1048
+ "epoch": 0.03,
1049
+ "grad_norm": 4.299948019967649,
1050
+ "learning_rate": 1.6610688880224178e-05,
1051
+ "loss": 5.8888,
1052
+ "step": 149
1053
+ },
1054
+ {
1055
+ "epoch": 0.03,
1056
+ "grad_norm": 4.095923631849489,
1057
+ "learning_rate": 1.6561948139065997e-05,
1058
+ "loss": 6.2453,
1059
+ "step": 150
1060
+ },
1061
+ {
1062
+ "epoch": 0.03,
1063
+ "grad_norm": 3.481920812397553,
1064
+ "learning_rate": 1.6512932071706153e-05,
1065
+ "loss": 6.0019,
1066
+ "step": 151
1067
+ },
1068
+ {
1069
+ "epoch": 0.03,
1070
+ "grad_norm": 4.3296848974625135,
1071
+ "learning_rate": 1.646364273476067e-05,
1072
+ "loss": 5.8587,
1073
+ "step": 152
1074
+ },
1075
+ {
1076
+ "epoch": 0.04,
1077
+ "grad_norm": 3.947321316567111,
1078
+ "learning_rate": 1.6414082196311402e-05,
1079
+ "loss": 5.928,
1080
+ "step": 153
1081
+ },
1082
+ {
1083
+ "epoch": 0.04,
1084
+ "grad_norm": 4.7243925766736705,
1085
+ "learning_rate": 1.6364252535819284e-05,
1086
+ "loss": 5.9007,
1087
+ "step": 154
1088
+ },
1089
+ {
1090
+ "epoch": 0.04,
1091
+ "grad_norm": 13.974458344598341,
1092
+ "learning_rate": 1.6314155844037074e-05,
1093
+ "loss": 5.9754,
1094
+ "step": 155
1095
+ },
1096
+ {
1097
+ "epoch": 0.04,
1098
+ "grad_norm": 4.17123392978259,
1099
+ "learning_rate": 1.626379422292162e-05,
1100
+ "loss": 5.8421,
1101
+ "step": 156
1102
+ },
1103
+ {
1104
+ "epoch": 0.04,
1105
+ "grad_norm": 4.043820315204433,
1106
+ "learning_rate": 1.6213169785545688e-05,
1107
+ "loss": 5.9519,
1108
+ "step": 157
1109
+ },
1110
+ {
1111
+ "epoch": 0.04,
1112
+ "grad_norm": 3.526011441047756,
1113
+ "learning_rate": 1.6162284656009276e-05,
1114
+ "loss": 5.8076,
1115
+ "step": 158
1116
+ },
1117
+ {
1118
+ "epoch": 0.04,
1119
+ "grad_norm": 3.8459147008292724,
1120
+ "learning_rate": 1.6111140969350504e-05,
1121
+ "loss": 5.7672,
1122
+ "step": 159
1123
+ },
1124
+ {
1125
+ "epoch": 0.04,
1126
+ "grad_norm": 3.562828950103298,
1127
+ "learning_rate": 1.6059740871456035e-05,
1128
+ "loss": 5.898,
1129
+ "step": 160
1130
+ },
1131
+ {
1132
+ "epoch": 0.04,
1133
+ "grad_norm": 4.431960095480538,
1134
+ "learning_rate": 1.6008086518971037e-05,
1135
+ "loss": 5.7521,
1136
+ "step": 161
1137
+ },
1138
+ {
1139
+ "epoch": 0.04,
1140
+ "grad_norm": 3.6428660471557737,
1141
+ "learning_rate": 1.5956180079208684e-05,
1142
+ "loss": 5.953,
1143
+ "step": 162
1144
+ },
1145
+ {
1146
+ "epoch": 0.04,
1147
+ "grad_norm": 3.767661352257826,
1148
+ "learning_rate": 1.5904023730059227e-05,
1149
+ "loss": 6.0916,
1150
+ "step": 163
1151
+ },
1152
+ {
1153
+ "epoch": 0.04,
1154
+ "grad_norm": 3.660701500822884,
1155
+ "learning_rate": 1.5851619659898623e-05,
1156
+ "loss": 5.8085,
1157
+ "step": 164
1158
+ },
1159
+ {
1160
+ "epoch": 0.04,
1161
+ "grad_norm": 3.6310726747437143,
1162
+ "learning_rate": 1.57989700674967e-05,
1163
+ "loss": 5.8102,
1164
+ "step": 165
1165
+ },
1166
+ {
1167
+ "epoch": 0.04,
1168
+ "grad_norm": 4.0023266213419495,
1169
+ "learning_rate": 1.5746077161924905e-05,
1170
+ "loss": 5.8879,
1171
+ "step": 166
1172
+ },
1173
+ {
1174
+ "epoch": 0.04,
1175
+ "grad_norm": 3.99237095864865,
1176
+ "learning_rate": 1.5692943162463628e-05,
1177
+ "loss": 5.8366,
1178
+ "step": 167
1179
+ },
1180
+ {
1181
+ "epoch": 0.04,
1182
+ "grad_norm": 3.5149347856437703,
1183
+ "learning_rate": 1.5639570298509067e-05,
1184
+ "loss": 5.7649,
1185
+ "step": 168
1186
+ },
1187
+ {
1188
+ "epoch": 0.04,
1189
+ "grad_norm": 3.301730499203444,
1190
+ "learning_rate": 1.5585960809479698e-05,
1191
+ "loss": 5.9754,
1192
+ "step": 169
1193
+ },
1194
+ {
1195
+ "epoch": 0.04,
1196
+ "grad_norm": 3.462115038986119,
1197
+ "learning_rate": 1.5532116944722308e-05,
1198
+ "loss": 5.6883,
1199
+ "step": 170
1200
+ },
1201
+ {
1202
+ "epoch": 0.04,
1203
+ "grad_norm": 4.004970215448472,
1204
+ "learning_rate": 1.547804096341763e-05,
1205
+ "loss": 5.8396,
1206
+ "step": 171
1207
+ },
1208
+ {
1209
+ "epoch": 0.04,
1210
+ "grad_norm": 3.944647631447267,
1211
+ "learning_rate": 1.5423735134485537e-05,
1212
+ "loss": 5.9988,
1213
+ "step": 172
1214
+ },
1215
+ {
1216
+ "epoch": 0.04,
1217
+ "grad_norm": 3.5900984593685723,
1218
+ "learning_rate": 1.536920173648984e-05,
1219
+ "loss": 5.7239,
1220
+ "step": 173
1221
+ },
1222
+ {
1223
+ "epoch": 0.04,
1224
+ "grad_norm": 3.6285904810079574,
1225
+ "learning_rate": 1.5314443057542703e-05,
1226
+ "loss": 5.8721,
1227
+ "step": 174
1228
+ },
1229
+ {
1230
+ "epoch": 0.04,
1231
+ "grad_norm": 3.8894726151488843,
1232
+ "learning_rate": 1.5259461395208628e-05,
1233
+ "loss": 5.8299,
1234
+ "step": 175
1235
+ },
1236
+ {
1237
+ "epoch": 0.04,
1238
+ "grad_norm": 3.5924238969723072,
1239
+ "learning_rate": 1.5204259056408046e-05,
1240
+ "loss": 5.9516,
1241
+ "step": 176
1242
+ },
1243
+ {
1244
+ "epoch": 0.04,
1245
+ "grad_norm": 5.538554836605968,
1246
+ "learning_rate": 1.5148838357320537e-05,
1247
+ "loss": 5.8963,
1248
+ "step": 177
1249
+ },
1250
+ {
1251
+ "epoch": 0.04,
1252
+ "grad_norm": 3.560612079271743,
1253
+ "learning_rate": 1.5093201623287631e-05,
1254
+ "loss": 5.888,
1255
+ "step": 178
1256
+ },
1257
+ {
1258
+ "epoch": 0.04,
1259
+ "grad_norm": 3.467820770014584,
1260
+ "learning_rate": 1.5037351188715265e-05,
1261
+ "loss": 5.7864,
1262
+ "step": 179
1263
+ },
1264
+ {
1265
+ "epoch": 0.04,
1266
+ "grad_norm": 4.172673845567331,
1267
+ "learning_rate": 1.4981289396975818e-05,
1268
+ "loss": 5.8498,
1269
+ "step": 180
1270
+ },
1271
+ {
1272
+ "epoch": 0.04,
1273
+ "grad_norm": 3.1779411060929816,
1274
+ "learning_rate": 1.4925018600309784e-05,
1275
+ "loss": 5.6926,
1276
+ "step": 181
1277
+ },
1278
+ {
1279
+ "epoch": 0.04,
1280
+ "grad_norm": 3.782675685637412,
1281
+ "learning_rate": 1.4868541159727097e-05,
1282
+ "loss": 5.6674,
1283
+ "step": 182
1284
+ },
1285
+ {
1286
+ "epoch": 0.04,
1287
+ "grad_norm": 3.7799393833784296,
1288
+ "learning_rate": 1.4811859444908053e-05,
1289
+ "loss": 5.8661,
1290
+ "step": 183
1291
+ },
1292
+ {
1293
+ "epoch": 0.04,
1294
+ "grad_norm": 4.487305708943541,
1295
+ "learning_rate": 1.4754975834103877e-05,
1296
+ "loss": 5.6999,
1297
+ "step": 184
1298
+ },
1299
+ {
1300
+ "epoch": 0.04,
1301
+ "grad_norm": 6.107952099039134,
1302
+ "learning_rate": 1.4697892714036959e-05,
1303
+ "loss": 5.9596,
1304
+ "step": 185
1305
+ },
1306
+ {
1307
+ "epoch": 0.04,
1308
+ "grad_norm": 3.897845235029772,
1309
+ "learning_rate": 1.4640612479800686e-05,
1310
+ "loss": 5.8336,
1311
+ "step": 186
1312
+ },
1313
+ {
1314
+ "epoch": 0.04,
1315
+ "grad_norm": 3.574618642064676,
1316
+ "learning_rate": 1.4583137534758968e-05,
1317
+ "loss": 5.678,
1318
+ "step": 187
1319
+ },
1320
+ {
1321
+ "epoch": 0.04,
1322
+ "grad_norm": 3.2767605854680717,
1323
+ "learning_rate": 1.4525470290445392e-05,
1324
+ "loss": 5.7779,
1325
+ "step": 188
1326
+ },
1327
+ {
1328
+ "epoch": 0.04,
1329
+ "grad_norm": 3.3494797594069343,
1330
+ "learning_rate": 1.4467613166462024e-05,
1331
+ "loss": 5.6117,
1332
+ "step": 189
1333
+ },
1334
+ {
1335
+ "epoch": 0.04,
1336
+ "grad_norm": 3.4560187160140114,
1337
+ "learning_rate": 1.4409568590377918e-05,
1338
+ "loss": 5.9744,
1339
+ "step": 190
1340
+ },
1341
+ {
1342
+ "epoch": 0.04,
1343
+ "grad_norm": 5.124637343157441,
1344
+ "learning_rate": 1.4351338997627233e-05,
1345
+ "loss": 5.8884,
1346
+ "step": 191
1347
+ },
1348
+ {
1349
+ "epoch": 0.04,
1350
+ "grad_norm": 4.802869734676735,
1351
+ "learning_rate": 1.429292683140706e-05,
1352
+ "loss": 5.8834,
1353
+ "step": 192
1354
+ },
1355
+ {
1356
+ "epoch": 0.04,
1357
+ "grad_norm": 4.830747863900574,
1358
+ "learning_rate": 1.4234334542574906e-05,
1359
+ "loss": 5.8741,
1360
+ "step": 193
1361
+ },
1362
+ {
1363
+ "epoch": 0.04,
1364
+ "grad_norm": 3.0222266779531313,
1365
+ "learning_rate": 1.4175564589545853e-05,
1366
+ "loss": 5.7957,
1367
+ "step": 194
1368
+ },
1369
+ {
1370
+ "epoch": 0.04,
1371
+ "grad_norm": 3.7517853962161003,
1372
+ "learning_rate": 1.411661943818944e-05,
1373
+ "loss": 5.773,
1374
+ "step": 195
1375
+ },
1376
+ {
1377
+ "epoch": 0.04,
1378
+ "grad_norm": 3.481751429815316,
1379
+ "learning_rate": 1.4057501561726157e-05,
1380
+ "loss": 5.79,
1381
+ "step": 196
1382
+ },
1383
+ {
1384
+ "epoch": 0.05,
1385
+ "grad_norm": 4.032738430652106,
1386
+ "learning_rate": 1.3998213440623691e-05,
1387
+ "loss": 5.8233,
1388
+ "step": 197
1389
+ },
1390
+ {
1391
+ "epoch": 0.05,
1392
+ "grad_norm": 4.3048433576019045,
1393
+ "learning_rate": 1.3938757562492873e-05,
1394
+ "loss": 5.7213,
1395
+ "step": 198
1396
+ },
1397
+ {
1398
+ "epoch": 0.05,
1399
+ "grad_norm": 3.42093480136207,
1400
+ "learning_rate": 1.3879136421983265e-05,
1401
+ "loss": 5.6734,
1402
+ "step": 199
1403
+ },
1404
+ {
1405
+ "epoch": 0.05,
1406
+ "grad_norm": 3.372940690507083,
1407
+ "learning_rate": 1.3819352520678519e-05,
1408
+ "loss": 5.8204,
1409
+ "step": 200
1410
+ },
1411
+ {
1412
+ "epoch": 0.05,
1413
+ "grad_norm": 3.866414604706453,
1414
+ "learning_rate": 1.3759408366991391e-05,
1415
+ "loss": 5.7685,
1416
+ "step": 201
1417
+ },
1418
+ {
1419
+ "epoch": 0.05,
1420
+ "grad_norm": 3.0437497252782015,
1421
+ "learning_rate": 1.3699306476058523e-05,
1422
+ "loss": 5.9321,
1423
+ "step": 202
1424
+ },
1425
+ {
1426
+ "epoch": 0.05,
1427
+ "grad_norm": 3.5806152918043153,
1428
+ "learning_rate": 1.3639049369634878e-05,
1429
+ "loss": 5.6817,
1430
+ "step": 203
1431
+ },
1432
+ {
1433
+ "epoch": 0.05,
1434
+ "grad_norm": 3.652624509453939,
1435
+ "learning_rate": 1.357863957598796e-05,
1436
+ "loss": 5.6123,
1437
+ "step": 204
1438
+ },
1439
+ {
1440
+ "epoch": 0.05,
1441
+ "grad_norm": 3.2444221493972063,
1442
+ "learning_rate": 1.3518079629791725e-05,
1443
+ "loss": 5.5523,
1444
+ "step": 205
1445
+ },
1446
+ {
1447
+ "epoch": 0.05,
1448
+ "grad_norm": 7.46063417126853,
1449
+ "learning_rate": 1.345737207202023e-05,
1450
+ "loss": 5.6855,
1451
+ "step": 206
1452
+ },
1453
+ {
1454
+ "epoch": 0.05,
1455
+ "grad_norm": 3.3218441227109787,
1456
+ "learning_rate": 1.3396519449841006e-05,
1457
+ "loss": 5.7366,
1458
+ "step": 207
1459
+ },
1460
+ {
1461
+ "epoch": 0.05,
1462
+ "grad_norm": 4.5944315284742725,
1463
+ "learning_rate": 1.3335524316508208e-05,
1464
+ "loss": 5.7278,
1465
+ "step": 208
1466
+ },
1467
+ {
1468
+ "epoch": 0.05,
1469
+ "grad_norm": 3.0069349104091603,
1470
+ "learning_rate": 1.3274389231255466e-05,
1471
+ "loss": 5.7355,
1472
+ "step": 209
1473
+ },
1474
+ {
1475
+ "epoch": 0.05,
1476
+ "grad_norm": 3.9317190820703756,
1477
+ "learning_rate": 1.3213116759188525e-05,
1478
+ "loss": 5.6806,
1479
+ "step": 210
1480
+ },
1481
+ {
1482
+ "epoch": 0.05,
1483
+ "grad_norm": 3.33285104772104,
1484
+ "learning_rate": 1.3151709471177589e-05,
1485
+ "loss": 5.7708,
1486
+ "step": 211
1487
+ },
1488
+ {
1489
+ "epoch": 0.05,
1490
+ "grad_norm": 3.6380751570471332,
1491
+ "learning_rate": 1.3090169943749475e-05,
1492
+ "loss": 5.8772,
1493
+ "step": 212
1494
+ },
1495
+ {
1496
+ "epoch": 0.05,
1497
+ "grad_norm": 3.742526748849025,
1498
+ "learning_rate": 1.3028500758979507e-05,
1499
+ "loss": 5.6913,
1500
+ "step": 213
1501
+ },
1502
+ {
1503
+ "epoch": 0.05,
1504
+ "grad_norm": 3.3393134329516196,
1505
+ "learning_rate": 1.296670450438317e-05,
1506
+ "loss": 5.5817,
1507
+ "step": 214
1508
+ },
1509
+ {
1510
+ "epoch": 0.05,
1511
+ "grad_norm": 3.887935011194972,
1512
+ "learning_rate": 1.2904783772807534e-05,
1513
+ "loss": 5.6364,
1514
+ "step": 215
1515
+ },
1516
+ {
1517
+ "epoch": 0.05,
1518
+ "grad_norm": 3.5742955487496175,
1519
+ "learning_rate": 1.2842741162322487e-05,
1520
+ "loss": 5.6486,
1521
+ "step": 216
1522
+ },
1523
+ {
1524
+ "epoch": 0.05,
1525
+ "grad_norm": 3.2697917373453245,
1526
+ "learning_rate": 1.2780579276111702e-05,
1527
+ "loss": 5.7063,
1528
+ "step": 217
1529
+ },
1530
+ {
1531
+ "epoch": 0.05,
1532
+ "grad_norm": 3.401421083396212,
1533
+ "learning_rate": 1.2718300722363431e-05,
1534
+ "loss": 5.7316,
1535
+ "step": 218
1536
+ },
1537
+ {
1538
+ "epoch": 0.05,
1539
+ "grad_norm": 3.6121121716149887,
1540
+ "learning_rate": 1.2655908114161053e-05,
1541
+ "loss": 5.7758,
1542
+ "step": 219
1543
+ },
1544
+ {
1545
+ "epoch": 0.05,
1546
+ "grad_norm": 3.2222903292023406,
1547
+ "learning_rate": 1.2593404069373452e-05,
1548
+ "loss": 5.7089,
1549
+ "step": 220
1550
+ },
1551
+ {
1552
+ "epoch": 0.05,
1553
+ "grad_norm": 3.025802281634462,
1554
+ "learning_rate": 1.2530791210545163e-05,
1555
+ "loss": 5.6953,
1556
+ "step": 221
1557
+ },
1558
+ {
1559
+ "epoch": 0.05,
1560
+ "grad_norm": 3.4673520664847075,
1561
+ "learning_rate": 1.2468072164786342e-05,
1562
+ "loss": 5.8728,
1563
+ "step": 222
1564
+ },
1565
+ {
1566
+ "epoch": 0.05,
1567
+ "grad_norm": 3.6806603608994735,
1568
+ "learning_rate": 1.2405249563662539e-05,
1569
+ "loss": 5.669,
1570
+ "step": 223
1571
+ },
1572
+ {
1573
+ "epoch": 0.05,
1574
+ "grad_norm": 4.64700420741913,
1575
+ "learning_rate": 1.2342326043084268e-05,
1576
+ "loss": 5.7352,
1577
+ "step": 224
1578
+ },
1579
+ {
1580
+ "epoch": 0.05,
1581
+ "grad_norm": 3.0790982186864504,
1582
+ "learning_rate": 1.2279304243196438e-05,
1583
+ "loss": 5.5999,
1584
+ "step": 225
1585
+ },
1586
+ {
1587
+ "epoch": 0.05,
1588
+ "grad_norm": 3.24646862661049,
1589
+ "learning_rate": 1.2216186808267544e-05,
1590
+ "loss": 5.668,
1591
+ "step": 226
1592
+ },
1593
+ {
1594
+ "epoch": 0.05,
1595
+ "grad_norm": 3.1831196277858864,
1596
+ "learning_rate": 1.215297638657875e-05,
1597
+ "loss": 5.7959,
1598
+ "step": 227
1599
+ },
1600
+ {
1601
+ "epoch": 0.05,
1602
+ "grad_norm": 3.4421555634435035,
1603
+ "learning_rate": 1.2089675630312755e-05,
1604
+ "loss": 5.6636,
1605
+ "step": 228
1606
+ },
1607
+ {
1608
+ "epoch": 0.05,
1609
+ "grad_norm": 2.833448990477623,
1610
+ "learning_rate": 1.2026287195442503e-05,
1611
+ "loss": 5.5742,
1612
+ "step": 229
1613
+ },
1614
+ {
1615
+ "epoch": 0.05,
1616
+ "grad_norm": 3.221649072609624,
1617
+ "learning_rate": 1.1962813741619777e-05,
1618
+ "loss": 5.7345,
1619
+ "step": 230
1620
+ },
1621
+ {
1622
+ "epoch": 0.05,
1623
+ "grad_norm": 4.5576181124301,
1624
+ "learning_rate": 1.189925793206357e-05,
1625
+ "loss": 5.8758,
1626
+ "step": 231
1627
+ },
1628
+ {
1629
+ "epoch": 0.05,
1630
+ "grad_norm": 3.1198515608939252,
1631
+ "learning_rate": 1.1835622433448361e-05,
1632
+ "loss": 5.7026,
1633
+ "step": 232
1634
+ },
1635
+ {
1636
+ "epoch": 0.05,
1637
+ "grad_norm": 3.2391737753303063,
1638
+ "learning_rate": 1.177190991579223e-05,
1639
+ "loss": 5.8245,
1640
+ "step": 233
1641
+ },
1642
+ {
1643
+ "epoch": 0.05,
1644
+ "grad_norm": 3.0415855629042006,
1645
+ "learning_rate": 1.1708123052344803e-05,
1646
+ "loss": 5.7423,
1647
+ "step": 234
1648
+ },
1649
+ {
1650
+ "epoch": 0.05,
1651
+ "grad_norm": 3.417436664351836,
1652
+ "learning_rate": 1.164426451947513e-05,
1653
+ "loss": 5.7224,
1654
+ "step": 235
1655
+ },
1656
+ {
1657
+ "epoch": 0.05,
1658
+ "grad_norm": 3.0428028654941963,
1659
+ "learning_rate": 1.1580336996559343e-05,
1660
+ "loss": 5.5215,
1661
+ "step": 236
1662
+ },
1663
+ {
1664
+ "epoch": 0.05,
1665
+ "grad_norm": 3.2412878897803843,
1666
+ "learning_rate": 1.151634316586828e-05,
1667
+ "loss": 5.6753,
1668
+ "step": 237
1669
+ },
1670
+ {
1671
+ "epoch": 0.05,
1672
+ "grad_norm": 2.9715981798900173,
1673
+ "learning_rate": 1.1452285712454905e-05,
1674
+ "loss": 5.6633,
1675
+ "step": 238
1676
+ },
1677
+ {
1678
+ "epoch": 0.05,
1679
+ "grad_norm": 3.3507364284363375,
1680
+ "learning_rate": 1.138816732404167e-05,
1681
+ "loss": 5.7531,
1682
+ "step": 239
1683
+ },
1684
+ {
1685
+ "epoch": 0.06,
1686
+ "grad_norm": 3.695763982565934,
1687
+ "learning_rate": 1.1323990690907734e-05,
1688
+ "loss": 5.6431,
1689
+ "step": 240
1690
+ },
1691
+ {
1692
+ "epoch": 0.06,
1693
+ "grad_norm": 3.442076843360669,
1694
+ "learning_rate": 1.1259758505776092e-05,
1695
+ "loss": 5.763,
1696
+ "step": 241
1697
+ },
1698
+ {
1699
+ "epoch": 0.06,
1700
+ "grad_norm": 3.4221454818185797,
1701
+ "learning_rate": 1.119547346370059e-05,
1702
+ "loss": 5.7121,
1703
+ "step": 242
1704
+ },
1705
+ {
1706
+ "epoch": 0.06,
1707
+ "grad_norm": 32.81366795375805,
1708
+ "learning_rate": 1.1131138261952845e-05,
1709
+ "loss": 5.7642,
1710
+ "step": 243
1711
+ },
1712
+ {
1713
+ "epoch": 0.06,
1714
+ "grad_norm": 2.824713348661609,
1715
+ "learning_rate": 1.1066755599909065e-05,
1716
+ "loss": 5.5872,
1717
+ "step": 244
1718
+ },
1719
+ {
1720
+ "epoch": 0.06,
1721
+ "grad_norm": 3.2122840787719356,
1722
+ "learning_rate": 1.1002328178936813e-05,
1723
+ "loss": 5.6909,
1724
+ "step": 245
1725
+ },
1726
+ {
1727
+ "epoch": 0.06,
1728
+ "grad_norm": 2.914139469492162,
1729
+ "learning_rate": 1.0937858702281631e-05,
1730
+ "loss": 5.5502,
1731
+ "step": 246
1732
+ },
1733
+ {
1734
+ "epoch": 0.06,
1735
+ "grad_norm": 3.372671623706637,
1736
+ "learning_rate": 1.087334987495364e-05,
1737
+ "loss": 5.638,
1738
+ "step": 247
1739
+ },
1740
+ {
1741
+ "epoch": 0.06,
1742
+ "grad_norm": 2.9794994417982705,
1743
+ "learning_rate": 1.0808804403614044e-05,
1744
+ "loss": 5.7327,
1745
+ "step": 248
1746
+ },
1747
+ {
1748
+ "epoch": 0.06,
1749
+ "grad_norm": 3.545219321353326,
1750
+ "learning_rate": 1.0744224996461541e-05,
1751
+ "loss": 5.773,
1752
+ "step": 249
1753
+ },
1754
+ {
1755
+ "epoch": 0.06,
1756
+ "grad_norm": 3.0509122712537904,
1757
+ "learning_rate": 1.0679614363118718e-05,
1758
+ "loss": 5.4833,
1759
+ "step": 250
1760
+ },
1761
+ {
1762
+ "epoch": 0.06,
1763
+ "grad_norm": 3.1328946212391573,
1764
+ "learning_rate": 1.061497521451835e-05,
1765
+ "loss": 5.5846,
1766
+ "step": 251
1767
+ },
1768
+ {
1769
+ "epoch": 0.06,
1770
+ "grad_norm": 2.912951012800422,
1771
+ "learning_rate": 1.055031026278965e-05,
1772
+ "loss": 5.7325,
1773
+ "step": 252
1774
+ },
1775
+ {
1776
+ "epoch": 0.06,
1777
+ "grad_norm": 3.190063789344824,
1778
+ "learning_rate": 1.0485622221144485e-05,
1779
+ "loss": 5.6504,
1780
+ "step": 253
1781
+ },
1782
+ {
1783
+ "epoch": 0.06,
1784
+ "grad_norm": 4.059074907434253,
1785
+ "learning_rate": 1.0420913803763522e-05,
1786
+ "loss": 5.5256,
1787
+ "step": 254
1788
+ },
1789
+ {
1790
+ "epoch": 0.06,
1791
+ "grad_norm": 3.2665977095269763,
1792
+ "learning_rate": 1.0356187725682359e-05,
1793
+ "loss": 5.6545,
1794
+ "step": 255
1795
+ },
1796
+ {
1797
+ "epoch": 0.06,
1798
+ "grad_norm": 2.8731971309590723,
1799
+ "learning_rate": 1.0291446702677598e-05,
1800
+ "loss": 5.4238,
1801
+ "step": 256
1802
+ },
1803
+ {
1804
+ "epoch": 0.06,
1805
+ "grad_norm": 5.960836799262505,
1806
+ "learning_rate": 1.02266934511529e-05,
1807
+ "loss": 5.601,
1808
+ "step": 257
1809
+ },
1810
+ {
1811
+ "epoch": 0.06,
1812
+ "grad_norm": 3.479663130288781,
1813
+ "learning_rate": 1.0161930688025018e-05,
1814
+ "loss": 5.49,
1815
+ "step": 258
1816
+ },
1817
+ {
1818
+ "epoch": 0.06,
1819
+ "grad_norm": 3.458943691824299,
1820
+ "learning_rate": 1.0097161130609774e-05,
1821
+ "loss": 5.7461,
1822
+ "step": 259
1823
+ },
1824
+ {
1825
+ "epoch": 0.06,
1826
+ "grad_norm": 3.279222709074338,
1827
+ "learning_rate": 1.003238749650809e-05,
1828
+ "loss": 5.5439,
1829
+ "step": 260
1830
+ },
1831
+ {
1832
+ "epoch": 0.06,
1833
+ "grad_norm": 5.311427762294207,
1834
+ "learning_rate": 9.967612503491915e-06,
1835
+ "loss": 5.5148,
1836
+ "step": 261
1837
+ },
1838
+ {
1839
+ "epoch": 0.06,
1840
+ "grad_norm": 2.887421444069423,
1841
+ "learning_rate": 9.90283886939023e-06,
1842
+ "loss": 5.4186,
1843
+ "step": 262
1844
+ },
1845
+ {
1846
+ "epoch": 0.06,
1847
+ "grad_norm": 4.401386334484841,
1848
+ "learning_rate": 9.838069311974986e-06,
1849
+ "loss": 5.6778,
1850
+ "step": 263
1851
+ },
1852
+ {
1853
+ "epoch": 0.06,
1854
+ "grad_norm": 2.6023206995102135,
1855
+ "learning_rate": 9.773306548847102e-06,
1856
+ "loss": 5.6155,
1857
+ "step": 264
1858
+ },
1859
+ {
1860
+ "epoch": 0.06,
1861
+ "grad_norm": 3.593734709591131,
1862
+ "learning_rate": 9.708553297322407e-06,
1863
+ "loss": 5.6777,
1864
+ "step": 265
1865
+ },
1866
+ {
1867
+ "epoch": 0.06,
1868
+ "grad_norm": 3.477325665085784,
1869
+ "learning_rate": 9.643812274317644e-06,
1870
+ "loss": 5.5781,
1871
+ "step": 266
1872
+ },
1873
+ {
1874
+ "epoch": 0.06,
1875
+ "grad_norm": 2.8887444330887266,
1876
+ "learning_rate": 9.579086196236483e-06,
1877
+ "loss": 5.7172,
1878
+ "step": 267
1879
+ },
1880
+ {
1881
+ "epoch": 0.06,
1882
+ "grad_norm": 3.8360288485375484,
1883
+ "learning_rate": 9.514377778855521e-06,
1884
+ "loss": 5.5803,
1885
+ "step": 268
1886
+ },
1887
+ {
1888
+ "epoch": 0.06,
1889
+ "grad_norm": 3.135597210733372,
1890
+ "learning_rate": 9.449689737210352e-06,
1891
+ "loss": 5.7742,
1892
+ "step": 269
1893
+ },
1894
+ {
1895
+ "epoch": 0.06,
1896
+ "grad_norm": 3.0578215352523745,
1897
+ "learning_rate": 9.385024785481653e-06,
1898
+ "loss": 5.5618,
1899
+ "step": 270
1900
+ },
1901
+ {
1902
+ "epoch": 0.06,
1903
+ "grad_norm": 2.925737489172214,
1904
+ "learning_rate": 9.320385636881283e-06,
1905
+ "loss": 5.6728,
1906
+ "step": 271
1907
+ },
1908
+ {
1909
+ "epoch": 0.06,
1910
+ "grad_norm": 3.429796358511194,
1911
+ "learning_rate": 9.255775003538462e-06,
1912
+ "loss": 5.7857,
1913
+ "step": 272
1914
+ },
1915
+ {
1916
+ "epoch": 0.06,
1917
+ "grad_norm": 3.1866968840925742,
1918
+ "learning_rate": 9.19119559638596e-06,
1919
+ "loss": 5.504,
1920
+ "step": 273
1921
+ },
1922
+ {
1923
+ "epoch": 0.06,
1924
+ "grad_norm": 3.199919489689139,
1925
+ "learning_rate": 9.126650125046361e-06,
1926
+ "loss": 5.6548,
1927
+ "step": 274
1928
+ },
1929
+ {
1930
+ "epoch": 0.06,
1931
+ "grad_norm": 6.644972124408221,
1932
+ "learning_rate": 9.062141297718372e-06,
1933
+ "loss": 5.7688,
1934
+ "step": 275
1935
+ },
1936
+ {
1937
+ "epoch": 0.06,
1938
+ "grad_norm": 2.840251387463206,
1939
+ "learning_rate": 8.99767182106319e-06,
1940
+ "loss": 5.5356,
1941
+ "step": 276
1942
+ },
1943
+ {
1944
+ "epoch": 0.06,
1945
+ "grad_norm": 2.847024195497787,
1946
+ "learning_rate": 8.933244400090937e-06,
1947
+ "loss": 5.538,
1948
+ "step": 277
1949
+ },
1950
+ {
1951
+ "epoch": 0.06,
1952
+ "grad_norm": 4.300138770028974,
1953
+ "learning_rate": 8.868861738047158e-06,
1954
+ "loss": 5.5466,
1955
+ "step": 278
1956
+ },
1957
+ {
1958
+ "epoch": 0.06,
1959
+ "grad_norm": 3.2598992072099806,
1960
+ "learning_rate": 8.804526536299413e-06,
1961
+ "loss": 5.5438,
1962
+ "step": 279
1963
+ },
1964
+ {
1965
+ "epoch": 0.06,
1966
+ "grad_norm": 2.9223921302516205,
1967
+ "learning_rate": 8.740241494223911e-06,
1968
+ "loss": 5.5199,
1969
+ "step": 280
1970
+ },
1971
+ {
1972
+ "epoch": 0.06,
1973
+ "grad_norm": 2.7228943907903944,
1974
+ "learning_rate": 8.676009309092273e-06,
1975
+ "loss": 5.316,
1976
+ "step": 281
1977
+ },
1978
+ {
1979
+ "epoch": 0.06,
1980
+ "grad_norm": 3.151284881347976,
1981
+ "learning_rate": 8.611832675958335e-06,
1982
+ "loss": 5.605,
1983
+ "step": 282
1984
+ },
1985
+ {
1986
+ "epoch": 0.06,
1987
+ "grad_norm": 2.971020945163525,
1988
+ "learning_rate": 8.5477142875451e-06,
1989
+ "loss": 5.5074,
1990
+ "step": 283
1991
+ },
1992
+ {
1993
+ "epoch": 0.07,
1994
+ "grad_norm": 3.383598237834473,
1995
+ "learning_rate": 8.48365683413172e-06,
1996
+ "loss": 5.6454,
1997
+ "step": 284
1998
+ },
1999
+ {
2000
+ "epoch": 0.07,
2001
+ "grad_norm": 3.1708795210472065,
2002
+ "learning_rate": 8.419663003440657e-06,
2003
+ "loss": 5.6293,
2004
+ "step": 285
2005
+ },
2006
+ {
2007
+ "epoch": 0.07,
2008
+ "grad_norm": 3.0060809846941523,
2009
+ "learning_rate": 8.355735480524874e-06,
2010
+ "loss": 5.7011,
2011
+ "step": 286
2012
+ },
2013
+ {
2014
+ "epoch": 0.07,
2015
+ "grad_norm": 3.082555191182687,
2016
+ "learning_rate": 8.291876947655197e-06,
2017
+ "loss": 5.5058,
2018
+ "step": 287
2019
+ },
2020
+ {
2021
+ "epoch": 0.07,
2022
+ "grad_norm": 3.1290655800212988,
2023
+ "learning_rate": 8.228090084207773e-06,
2024
+ "loss": 5.6645,
2025
+ "step": 288
2026
+ },
2027
+ {
2028
+ "epoch": 0.07,
2029
+ "grad_norm": 3.6512062285821107,
2030
+ "learning_rate": 8.16437756655164e-06,
2031
+ "loss": 5.632,
2032
+ "step": 289
2033
+ },
2034
+ {
2035
+ "epoch": 0.07,
2036
+ "grad_norm": 5.580032474883606,
2037
+ "learning_rate": 8.100742067936432e-06,
2038
+ "loss": 5.6529,
2039
+ "step": 290
2040
+ },
2041
+ {
2042
+ "epoch": 0.07,
2043
+ "grad_norm": 2.8279735010912828,
2044
+ "learning_rate": 8.037186258380226e-06,
2045
+ "loss": 5.653,
2046
+ "step": 291
2047
+ },
2048
+ {
2049
+ "epoch": 0.07,
2050
+ "grad_norm": 3.0485365662315185,
2051
+ "learning_rate": 7.9737128045575e-06,
2052
+ "loss": 5.6271,
2053
+ "step": 292
2054
+ },
2055
+ {
2056
+ "epoch": 0.07,
2057
+ "grad_norm": 2.9000686534385065,
2058
+ "learning_rate": 7.91032436968725e-06,
2059
+ "loss": 5.497,
2060
+ "step": 293
2061
+ },
2062
+ {
2063
+ "epoch": 0.07,
2064
+ "grad_norm": 3.7477894288698823,
2065
+ "learning_rate": 7.847023613421251e-06,
2066
+ "loss": 5.6454,
2067
+ "step": 294
2068
+ },
2069
+ {
2070
+ "epoch": 0.07,
2071
+ "grad_norm": 3.1406780416239988,
2072
+ "learning_rate": 7.78381319173246e-06,
2073
+ "loss": 5.702,
2074
+ "step": 295
2075
+ },
2076
+ {
2077
+ "epoch": 0.07,
2078
+ "grad_norm": 2.8819040313896664,
2079
+ "learning_rate": 7.720695756803569e-06,
2080
+ "loss": 5.5487,
2081
+ "step": 296
2082
+ },
2083
+ {
2084
+ "epoch": 0.07,
2085
+ "grad_norm": 3.1093053507455433,
2086
+ "learning_rate": 7.657673956915735e-06,
2087
+ "loss": 5.6263,
2088
+ "step": 297
2089
+ },
2090
+ {
2091
+ "epoch": 0.07,
2092
+ "grad_norm": 3.532080902497016,
2093
+ "learning_rate": 7.594750436337467e-06,
2094
+ "loss": 5.7086,
2095
+ "step": 298
2096
+ },
2097
+ {
2098
+ "epoch": 0.07,
2099
+ "grad_norm": 2.9271604989314417,
2100
+ "learning_rate": 7.531927835213657e-06,
2101
+ "loss": 5.5651,
2102
+ "step": 299
2103
+ },
2104
+ {
2105
+ "epoch": 0.07,
2106
+ "grad_norm": 3.9573677159990344,
2107
+ "learning_rate": 7.469208789454838e-06,
2108
+ "loss": 5.6534,
2109
+ "step": 300
2110
+ },
2111
+ {
2112
+ "epoch": 0.07,
2113
+ "grad_norm": 2.813452933446664,
2114
+ "learning_rate": 7.40659593062655e-06,
2115
+ "loss": 5.5948,
2116
+ "step": 301
2117
+ },
2118
+ {
2119
+ "epoch": 0.07,
2120
+ "grad_norm": 3.64918124403021,
2121
+ "learning_rate": 7.344091885838949e-06,
2122
+ "loss": 5.833,
2123
+ "step": 302
2124
+ },
2125
+ {
2126
+ "epoch": 0.07,
2127
+ "grad_norm": 8.447026127108531,
2128
+ "learning_rate": 7.2816992776365714e-06,
2129
+ "loss": 5.6345,
2130
+ "step": 303
2131
+ },
2132
+ {
2133
+ "epoch": 0.07,
2134
+ "grad_norm": 3.487783377440865,
2135
+ "learning_rate": 7.219420723888301e-06,
2136
+ "loss": 5.5796,
2137
+ "step": 304
2138
+ },
2139
+ {
2140
+ "epoch": 0.07,
2141
+ "grad_norm": 3.3746557421600545,
2142
+ "learning_rate": 7.157258837677514e-06,
2143
+ "loss": 5.4475,
2144
+ "step": 305
2145
+ },
2146
+ {
2147
+ "epoch": 0.07,
2148
+ "grad_norm": 3.109170816456849,
2149
+ "learning_rate": 7.095216227192467e-06,
2150
+ "loss": 5.4312,
2151
+ "step": 306
2152
+ },
2153
+ {
2154
+ "epoch": 0.07,
2155
+ "grad_norm": 3.5209993320764634,
2156
+ "learning_rate": 7.033295495616834e-06,
2157
+ "loss": 5.6655,
2158
+ "step": 307
2159
+ },
2160
+ {
2161
+ "epoch": 0.07,
2162
+ "grad_norm": 3.7885132212833215,
2163
+ "learning_rate": 6.971499241020495e-06,
2164
+ "loss": 5.5506,
2165
+ "step": 308
2166
+ },
2167
+ {
2168
+ "epoch": 0.07,
2169
+ "grad_norm": 2.8035280155793902,
2170
+ "learning_rate": 6.909830056250527e-06,
2171
+ "loss": 5.5899,
2172
+ "step": 309
2173
+ },
2174
+ {
2175
+ "epoch": 0.07,
2176
+ "grad_norm": 3.3096990722045665,
2177
+ "learning_rate": 6.848290528822417e-06,
2178
+ "loss": 5.5894,
2179
+ "step": 310
2180
+ },
2181
+ {
2182
+ "epoch": 0.07,
2183
+ "grad_norm": 3.1035770416615245,
2184
+ "learning_rate": 6.786883240811479e-06,
2185
+ "loss": 5.5137,
2186
+ "step": 311
2187
+ },
2188
+ {
2189
+ "epoch": 0.07,
2190
+ "grad_norm": 2.934891742489131,
2191
+ "learning_rate": 6.725610768744535e-06,
2192
+ "loss": 5.5028,
2193
+ "step": 312
2194
+ },
2195
+ {
2196
+ "epoch": 0.07,
2197
+ "grad_norm": 3.371896210185273,
2198
+ "learning_rate": 6.664475683491797e-06,
2199
+ "loss": 5.5001,
2200
+ "step": 313
2201
+ },
2202
+ {
2203
+ "epoch": 0.07,
2204
+ "grad_norm": 2.8035512763165222,
2205
+ "learning_rate": 6.603480550158995e-06,
2206
+ "loss": 5.6237,
2207
+ "step": 314
2208
+ },
2209
+ {
2210
+ "epoch": 0.07,
2211
+ "grad_norm": 4.14158466178092,
2212
+ "learning_rate": 6.542627927979772e-06,
2213
+ "loss": 5.6736,
2214
+ "step": 315
2215
+ },
2216
+ {
2217
+ "epoch": 0.07,
2218
+ "grad_norm": 2.972559988613375,
2219
+ "learning_rate": 6.481920370208274e-06,
2220
+ "loss": 5.6447,
2221
+ "step": 316
2222
+ },
2223
+ {
2224
+ "epoch": 0.07,
2225
+ "grad_norm": 3.411613112123063,
2226
+ "learning_rate": 6.421360424012039e-06,
2227
+ "loss": 5.7263,
2228
+ "step": 317
2229
+ },
2230
+ {
2231
+ "epoch": 0.07,
2232
+ "grad_norm": 2.8412153634814308,
2233
+ "learning_rate": 6.360950630365126e-06,
2234
+ "loss": 5.5072,
2235
+ "step": 318
2236
+ },
2237
+ {
2238
+ "epoch": 0.07,
2239
+ "grad_norm": 3.376692187351766,
2240
+ "learning_rate": 6.300693523941481e-06,
2241
+ "loss": 5.6328,
2242
+ "step": 319
2243
+ },
2244
+ {
2245
+ "epoch": 0.07,
2246
+ "grad_norm": 2.950706203004661,
2247
+ "learning_rate": 6.2405916330086106e-06,
2248
+ "loss": 5.606,
2249
+ "step": 320
2250
+ },
2251
+ {
2252
+ "epoch": 0.07,
2253
+ "grad_norm": 3.6164997018204152,
2254
+ "learning_rate": 6.180647479321484e-06,
2255
+ "loss": 5.7159,
2256
+ "step": 321
2257
+ },
2258
+ {
2259
+ "epoch": 0.07,
2260
+ "grad_norm": 2.6776792490163492,
2261
+ "learning_rate": 6.120863578016736e-06,
2262
+ "loss": 5.4034,
2263
+ "step": 322
2264
+ },
2265
+ {
2266
+ "epoch": 0.07,
2267
+ "grad_norm": 2.8823500476883783,
2268
+ "learning_rate": 6.061242437507131e-06,
2269
+ "loss": 5.5799,
2270
+ "step": 323
2271
+ },
2272
+ {
2273
+ "epoch": 0.07,
2274
+ "grad_norm": 2.7310153497627305,
2275
+ "learning_rate": 6.00178655937631e-06,
2276
+ "loss": 5.4403,
2277
+ "step": 324
2278
+ },
2279
+ {
2280
+ "epoch": 0.07,
2281
+ "grad_norm": 2.8043161109289105,
2282
+ "learning_rate": 5.942498438273849e-06,
2283
+ "loss": 5.5201,
2284
+ "step": 325
2285
+ },
2286
+ {
2287
+ "epoch": 0.07,
2288
+ "grad_norm": 2.6257376922078577,
2289
+ "learning_rate": 5.8833805618105635e-06,
2290
+ "loss": 5.5161,
2291
+ "step": 326
2292
+ },
2293
+ {
2294
+ "epoch": 0.07,
2295
+ "grad_norm": 3.100365890406591,
2296
+ "learning_rate": 5.82443541045415e-06,
2297
+ "loss": 5.4591,
2298
+ "step": 327
2299
+ },
2300
+ {
2301
+ "epoch": 0.08,
2302
+ "grad_norm": 3.164224570414931,
2303
+ "learning_rate": 5.765665457425102e-06,
2304
+ "loss": 5.5032,
2305
+ "step": 328
2306
+ },
2307
+ {
2308
+ "epoch": 0.08,
2309
+ "grad_norm": 2.8688071534260713,
2310
+ "learning_rate": 5.707073168592943e-06,
2311
+ "loss": 5.6875,
2312
+ "step": 329
2313
+ },
2314
+ {
2315
+ "epoch": 0.08,
2316
+ "grad_norm": 2.9334355868792925,
2317
+ "learning_rate": 5.648661002372769e-06,
2318
+ "loss": 5.4918,
2319
+ "step": 330
2320
+ },
2321
+ {
2322
+ "epoch": 0.08,
2323
+ "grad_norm": 2.618203487365493,
2324
+ "learning_rate": 5.590431409622081e-06,
2325
+ "loss": 5.6324,
2326
+ "step": 331
2327
+ },
2328
+ {
2329
+ "epoch": 0.08,
2330
+ "grad_norm": 2.7468055145560473,
2331
+ "learning_rate": 5.5323868335379775e-06,
2332
+ "loss": 5.5696,
2333
+ "step": 332
2334
+ },
2335
+ {
2336
+ "epoch": 0.08,
2337
+ "grad_norm": 2.945148993510563,
2338
+ "learning_rate": 5.4745297095546125e-06,
2339
+ "loss": 5.6865,
2340
+ "step": 333
2341
+ },
2342
+ {
2343
+ "epoch": 0.08,
2344
+ "grad_norm": 4.27286874096374,
2345
+ "learning_rate": 5.416862465241033e-06,
2346
+ "loss": 5.574,
2347
+ "step": 334
2348
+ },
2349
+ {
2350
+ "epoch": 0.08,
2351
+ "grad_norm": 3.075467269510141,
2352
+ "learning_rate": 5.359387520199317e-06,
2353
+ "loss": 5.4386,
2354
+ "step": 335
2355
+ },
2356
+ {
2357
+ "epoch": 0.08,
2358
+ "grad_norm": 12.979333973830173,
2359
+ "learning_rate": 5.302107285963045e-06,
2360
+ "loss": 5.5171,
2361
+ "step": 336
2362
+ },
2363
+ {
2364
+ "epoch": 0.08,
2365
+ "grad_norm": 2.946796474047621,
2366
+ "learning_rate": 5.245024165896126e-06,
2367
+ "loss": 5.4337,
2368
+ "step": 337
2369
+ },
2370
+ {
2371
+ "epoch": 0.08,
2372
+ "grad_norm": 47.491660622837905,
2373
+ "learning_rate": 5.18814055509195e-06,
2374
+ "loss": 5.3093,
2375
+ "step": 338
2376
+ },
2377
+ {
2378
+ "epoch": 0.08,
2379
+ "grad_norm": 2.6684688936423306,
2380
+ "learning_rate": 5.131458840272905e-06,
2381
+ "loss": 5.4504,
2382
+ "step": 339
2383
+ },
2384
+ {
2385
+ "epoch": 0.08,
2386
+ "grad_norm": 2.9524813775023664,
2387
+ "learning_rate": 5.074981399690219e-06,
2388
+ "loss": 5.6183,
2389
+ "step": 340
2390
+ },
2391
+ {
2392
+ "epoch": 0.08,
2393
+ "grad_norm": 3.071692317071865,
2394
+ "learning_rate": 5.018710603024187e-06,
2395
+ "loss": 5.8018,
2396
+ "step": 341
2397
+ },
2398
+ {
2399
+ "epoch": 0.08,
2400
+ "grad_norm": 2.8943426237790604,
2401
+ "learning_rate": 4.9626488112847384e-06,
2402
+ "loss": 5.5124,
2403
+ "step": 342
2404
+ },
2405
+ {
2406
+ "epoch": 0.08,
2407
+ "grad_norm": 3.643486862482141,
2408
+ "learning_rate": 4.9067983767123736e-06,
2409
+ "loss": 5.6324,
2410
+ "step": 343
2411
+ },
2412
+ {
2413
+ "epoch": 0.08,
2414
+ "grad_norm": 3.0446879911014793,
2415
+ "learning_rate": 4.851161642679466e-06,
2416
+ "loss": 5.5722,
2417
+ "step": 344
2418
+ },
2419
+ {
2420
+ "epoch": 0.08,
2421
+ "grad_norm": 3.0223614188577583,
2422
+ "learning_rate": 4.795740943591955e-06,
2423
+ "loss": 5.6997,
2424
+ "step": 345
2425
+ },
2426
+ {
2427
+ "epoch": 0.08,
2428
+ "grad_norm": 3.139884095364713,
2429
+ "learning_rate": 4.740538604791371e-06,
2430
+ "loss": 5.6098,
2431
+ "step": 346
2432
+ },
2433
+ {
2434
+ "epoch": 0.08,
2435
+ "grad_norm": 3.0355695744448905,
2436
+ "learning_rate": 4.685556942457296e-06,
2437
+ "loss": 5.5642,
2438
+ "step": 347
2439
+ },
2440
+ {
2441
+ "epoch": 0.08,
2442
+ "grad_norm": 3.2148124769621145,
2443
+ "learning_rate": 4.630798263510162e-06,
2444
+ "loss": 5.4686,
2445
+ "step": 348
2446
+ },
2447
+ {
2448
+ "epoch": 0.08,
2449
+ "grad_norm": 6.599100924296892,
2450
+ "learning_rate": 4.576264865514467e-06,
2451
+ "loss": 5.6283,
2452
+ "step": 349
2453
+ },
2454
+ {
2455
+ "epoch": 0.08,
2456
+ "grad_norm": 3.176903096047862,
2457
+ "learning_rate": 4.521959036582372e-06,
2458
+ "loss": 5.6207,
2459
+ "step": 350
2460
+ },
2461
+ {
2462
+ "epoch": 0.08,
2463
+ "grad_norm": 2.7882598761964217,
2464
+ "learning_rate": 4.467883055277696e-06,
2465
+ "loss": 5.4235,
2466
+ "step": 351
2467
+ },
2468
+ {
2469
+ "epoch": 0.08,
2470
+ "grad_norm": 3.4794420663108894,
2471
+ "learning_rate": 4.414039190520308e-06,
2472
+ "loss": 5.5631,
2473
+ "step": 352
2474
+ },
2475
+ {
2476
+ "epoch": 0.08,
2477
+ "grad_norm": 2.903622461402484,
2478
+ "learning_rate": 4.360429701490935e-06,
2479
+ "loss": 5.6318,
2480
+ "step": 353
2481
+ },
2482
+ {
2483
+ "epoch": 0.08,
2484
+ "grad_norm": 7.699211582040187,
2485
+ "learning_rate": 4.307056837536373e-06,
2486
+ "loss": 5.7193,
2487
+ "step": 354
2488
+ },
2489
+ {
2490
+ "epoch": 0.08,
2491
+ "grad_norm": 3.637017761567808,
2492
+ "learning_rate": 4.2539228380750955e-06,
2493
+ "loss": 5.6718,
2494
+ "step": 355
2495
+ },
2496
+ {
2497
+ "epoch": 0.08,
2498
+ "grad_norm": 3.871928827914661,
2499
+ "learning_rate": 4.201029932503303e-06,
2500
+ "loss": 5.5911,
2501
+ "step": 356
2502
+ },
2503
+ {
2504
+ "epoch": 0.08,
2505
+ "grad_norm": 3.373623961352284,
2506
+ "learning_rate": 4.14838034010138e-06,
2507
+ "loss": 5.6983,
2508
+ "step": 357
2509
+ },
2510
+ {
2511
+ "epoch": 0.08,
2512
+ "grad_norm": 2.7039462572520025,
2513
+ "learning_rate": 4.095976269940777e-06,
2514
+ "loss": 5.4657,
2515
+ "step": 358
2516
+ },
2517
+ {
2518
+ "epoch": 0.08,
2519
+ "grad_norm": 2.906259270274048,
2520
+ "learning_rate": 4.043819920791322e-06,
2521
+ "loss": 5.6059,
2522
+ "step": 359
2523
+ },
2524
+ {
2525
+ "epoch": 0.08,
2526
+ "grad_norm": 2.80215901884808,
2527
+ "learning_rate": 3.991913481028965e-06,
2528
+ "loss": 5.3867,
2529
+ "step": 360
2530
+ },
2531
+ {
2532
+ "epoch": 0.08,
2533
+ "grad_norm": 2.7611740391342763,
2534
+ "learning_rate": 3.940259128543967e-06,
2535
+ "loss": 5.6265,
2536
+ "step": 361
2537
+ },
2538
+ {
2539
+ "epoch": 0.08,
2540
+ "grad_norm": 3.496220876703193,
2541
+ "learning_rate": 3.888859030649498e-06,
2542
+ "loss": 5.7369,
2543
+ "step": 362
2544
+ },
2545
+ {
2546
+ "epoch": 0.08,
2547
+ "grad_norm": 3.069627191877829,
2548
+ "learning_rate": 3.837715343990727e-06,
2549
+ "loss": 5.6878,
2550
+ "step": 363
2551
+ },
2552
+ {
2553
+ "epoch": 0.08,
2554
+ "grad_norm": 2.8861264690211628,
2555
+ "learning_rate": 3.7868302144543146e-06,
2556
+ "loss": 5.6452,
2557
+ "step": 364
2558
+ },
2559
+ {
2560
+ "epoch": 0.08,
2561
+ "grad_norm": 2.973826787518843,
2562
+ "learning_rate": 3.736205777078381e-06,
2563
+ "loss": 5.4028,
2564
+ "step": 365
2565
+ },
2566
+ {
2567
+ "epoch": 0.08,
2568
+ "grad_norm": 2.9173591380715704,
2569
+ "learning_rate": 3.685844155962931e-06,
2570
+ "loss": 5.5656,
2571
+ "step": 366
2572
+ },
2573
+ {
2574
+ "epoch": 0.08,
2575
+ "grad_norm": 3.147705831696598,
2576
+ "learning_rate": 3.63574746418072e-06,
2577
+ "loss": 5.5783,
2578
+ "step": 367
2579
+ },
2580
+ {
2581
+ "epoch": 0.08,
2582
+ "grad_norm": 3.072248624553244,
2583
+ "learning_rate": 3.585917803688603e-06,
2584
+ "loss": 5.539,
2585
+ "step": 368
2586
+ },
2587
+ {
2588
+ "epoch": 0.08,
2589
+ "grad_norm": 2.7322222965861336,
2590
+ "learning_rate": 3.536357265239333e-06,
2591
+ "loss": 5.4976,
2592
+ "step": 369
2593
+ },
2594
+ {
2595
+ "epoch": 0.08,
2596
+ "grad_norm": 2.658079903087022,
2597
+ "learning_rate": 3.487067928293848e-06,
2598
+ "loss": 5.4221,
2599
+ "step": 370
2600
+ },
2601
+ {
2602
+ "epoch": 0.09,
2603
+ "grad_norm": 2.680370794084422,
2604
+ "learning_rate": 3.4380518609340076e-06,
2605
+ "loss": 5.5785,
2606
+ "step": 371
2607
+ },
2608
+ {
2609
+ "epoch": 0.09,
2610
+ "grad_norm": 2.89315062272381,
2611
+ "learning_rate": 3.3893111197758276e-06,
2612
+ "loss": 5.4699,
2613
+ "step": 372
2614
+ },
2615
+ {
2616
+ "epoch": 0.09,
2617
+ "grad_norm": 3.088114743299729,
2618
+ "learning_rate": 3.3408477498831917e-06,
2619
+ "loss": 5.5848,
2620
+ "step": 373
2621
+ },
2622
+ {
2623
+ "epoch": 0.09,
2624
+ "grad_norm": 3.3744405750175783,
2625
+ "learning_rate": 3.2926637846820366e-06,
2626
+ "loss": 5.5855,
2627
+ "step": 374
2628
+ },
2629
+ {
2630
+ "epoch": 0.09,
2631
+ "grad_norm": 3.0949682435479815,
2632
+ "learning_rate": 3.2447612458750365e-06,
2633
+ "loss": 5.5665,
2634
+ "step": 375
2635
+ },
2636
+ {
2637
+ "epoch": 0.09,
2638
+ "grad_norm": 3.5298918745049064,
2639
+ "learning_rate": 3.197142143356787e-06,
2640
+ "loss": 5.4282,
2641
+ "step": 376
2642
+ },
2643
+ {
2644
+ "epoch": 0.09,
2645
+ "grad_norm": 2.3415078853618483,
2646
+ "learning_rate": 3.1498084751294523e-06,
2647
+ "loss": 5.548,
2648
+ "step": 377
2649
+ },
2650
+ {
2651
+ "epoch": 0.09,
2652
+ "grad_norm": 2.6530303000947377,
2653
+ "learning_rate": 3.1027622272189572e-06,
2654
+ "loss": 5.4614,
2655
+ "step": 378
2656
+ },
2657
+ {
2658
+ "epoch": 0.09,
2659
+ "grad_norm": 2.6198793225005095,
2660
+ "learning_rate": 3.0560053735916372e-06,
2661
+ "loss": 5.3387,
2662
+ "step": 379
2663
+ },
2664
+ {
2665
+ "epoch": 0.09,
2666
+ "grad_norm": 3.253733908459485,
2667
+ "learning_rate": 3.009539876071427e-06,
2668
+ "loss": 5.5892,
2669
+ "step": 380
2670
+ },
2671
+ {
2672
+ "epoch": 0.09,
2673
+ "grad_norm": 2.893078721591448,
2674
+ "learning_rate": 2.9633676842575386e-06,
2675
+ "loss": 5.5098,
2676
+ "step": 381
2677
+ },
2678
+ {
2679
+ "epoch": 0.09,
2680
+ "grad_norm": 3.2332247538663927,
2681
+ "learning_rate": 2.9174907354426696e-06,
2682
+ "loss": 5.611,
2683
+ "step": 382
2684
+ },
2685
+ {
2686
+ "epoch": 0.09,
2687
+ "grad_norm": 3.717694710480861,
2688
+ "learning_rate": 2.8719109545317102e-06,
2689
+ "loss": 5.7531,
2690
+ "step": 383
2691
+ },
2692
+ {
2693
+ "epoch": 0.09,
2694
+ "grad_norm": 5.973350020162763,
2695
+ "learning_rate": 2.8266302539609747e-06,
2696
+ "loss": 5.5053,
2697
+ "step": 384
2698
+ },
2699
+ {
2700
+ "epoch": 0.09,
2701
+ "grad_norm": 3.106939977808064,
2702
+ "learning_rate": 2.78165053361798e-06,
2703
+ "loss": 5.5299,
2704
+ "step": 385
2705
+ },
2706
+ {
2707
+ "epoch": 0.09,
2708
+ "grad_norm": 3.743314132309298,
2709
+ "learning_rate": 2.736973680761702e-06,
2710
+ "loss": 5.6108,
2711
+ "step": 386
2712
+ },
2713
+ {
2714
+ "epoch": 0.09,
2715
+ "grad_norm": 3.020284676653917,
2716
+ "learning_rate": 2.692601569943407e-06,
2717
+ "loss": 5.5949,
2718
+ "step": 387
2719
+ },
2720
+ {
2721
+ "epoch": 0.09,
2722
+ "grad_norm": 2.8688896218347466,
2723
+ "learning_rate": 2.648536062927999e-06,
2724
+ "loss": 5.6542,
2725
+ "step": 388
2726
+ },
2727
+ {
2728
+ "epoch": 0.09,
2729
+ "grad_norm": 2.8774214420614292,
2730
+ "learning_rate": 2.604779008615895e-06,
2731
+ "loss": 5.6006,
2732
+ "step": 389
2733
+ },
2734
+ {
2735
+ "epoch": 0.09,
2736
+ "grad_norm": 2.5974922903934443,
2737
+ "learning_rate": 2.5613322429654573e-06,
2738
+ "loss": 5.4603,
2739
+ "step": 390
2740
+ },
2741
+ {
2742
+ "epoch": 0.09,
2743
+ "grad_norm": 3.1182016556730034,
2744
+ "learning_rate": 2.5181975889159615e-06,
2745
+ "loss": 5.4617,
2746
+ "step": 391
2747
+ },
2748
+ {
2749
+ "epoch": 0.09,
2750
+ "grad_norm": 3.02296875164515,
2751
+ "learning_rate": 2.475376856311097e-06,
2752
+ "loss": 5.5758,
2753
+ "step": 392
2754
+ },
2755
+ {
2756
+ "epoch": 0.09,
2757
+ "grad_norm": 3.1443470043442066,
2758
+ "learning_rate": 2.432871841823047e-06,
2759
+ "loss": 5.4938,
2760
+ "step": 393
2761
+ },
2762
+ {
2763
+ "epoch": 0.09,
2764
+ "grad_norm": 5.21437085093465,
2765
+ "learning_rate": 2.390684328877089e-06,
2766
+ "loss": 5.543,
2767
+ "step": 394
2768
+ },
2769
+ {
2770
+ "epoch": 0.09,
2771
+ "grad_norm": 3.0664362463556603,
2772
+ "learning_rate": 2.3488160875767717e-06,
2773
+ "loss": 5.6436,
2774
+ "step": 395
2775
+ },
2776
+ {
2777
+ "epoch": 0.09,
2778
+ "grad_norm": 2.793841886787081,
2779
+ "learning_rate": 2.307268874629649e-06,
2780
+ "loss": 5.3901,
2781
+ "step": 396
2782
+ },
2783
+ {
2784
+ "epoch": 0.09,
2785
+ "grad_norm": 2.6568460536060137,
2786
+ "learning_rate": 2.266044433273562e-06,
2787
+ "loss": 5.5585,
2788
+ "step": 397
2789
+ },
2790
+ {
2791
+ "epoch": 0.09,
2792
+ "grad_norm": 3.3923148492013504,
2793
+ "learning_rate": 2.2251444932035094e-06,
2794
+ "loss": 5.3515,
2795
+ "step": 398
2796
+ },
2797
+ {
2798
+ "epoch": 0.09,
2799
+ "grad_norm": 4.187296832849823,
2800
+ "learning_rate": 2.184570770499056e-06,
2801
+ "loss": 5.4722,
2802
+ "step": 399
2803
+ },
2804
+ {
2805
+ "epoch": 0.09,
2806
+ "grad_norm": 2.8884881509409555,
2807
+ "learning_rate": 2.1443249675523536e-06,
2808
+ "loss": 5.537,
2809
+ "step": 400
2810
+ },
2811
+ {
2812
+ "epoch": 0.09,
2813
+ "grad_norm": 3.051680492849398,
2814
+ "learning_rate": 2.1044087729966856e-06,
2815
+ "loss": 5.4369,
2816
+ "step": 401
2817
+ },
2818
+ {
2819
+ "epoch": 0.09,
2820
+ "grad_norm": 2.8280027110122608,
2821
+ "learning_rate": 2.064823861635633e-06,
2822
+ "loss": 5.5705,
2823
+ "step": 402
2824
+ },
2825
+ {
2826
+ "epoch": 0.09,
2827
+ "grad_norm": 3.4037253520135216,
2828
+ "learning_rate": 2.025571894372794e-06,
2829
+ "loss": 5.5584,
2830
+ "step": 403
2831
+ },
2832
+ {
2833
+ "epoch": 0.09,
2834
+ "grad_norm": 4.319540102434969,
2835
+ "learning_rate": 1.9866545181421016e-06,
2836
+ "loss": 5.4076,
2837
+ "step": 404
2838
+ },
2839
+ {
2840
+ "epoch": 0.09,
2841
+ "grad_norm": 2.808044157619595,
2842
+ "learning_rate": 1.9480733658387175e-06,
2843
+ "loss": 5.4129,
2844
+ "step": 405
2845
+ },
2846
+ {
2847
+ "epoch": 0.09,
2848
+ "grad_norm": 2.7351469976881306,
2849
+ "learning_rate": 1.9098300562505266e-06,
2850
+ "loss": 5.6271,
2851
+ "step": 406
2852
+ },
2853
+ {
2854
+ "epoch": 0.09,
2855
+ "grad_norm": 2.7165917157682737,
2856
+ "learning_rate": 1.8719261939902023e-06,
2857
+ "loss": 5.4525,
2858
+ "step": 407
2859
+ },
2860
+ {
2861
+ "epoch": 0.09,
2862
+ "grad_norm": 3.040380201846354,
2863
+ "learning_rate": 1.8343633694278895e-06,
2864
+ "loss": 5.5268,
2865
+ "step": 408
2866
+ },
2867
+ {
2868
+ "epoch": 0.09,
2869
+ "grad_norm": 2.8167171485840035,
2870
+ "learning_rate": 1.7971431586244814e-06,
2871
+ "loss": 5.4247,
2872
+ "step": 409
2873
+ },
2874
+ {
2875
+ "epoch": 0.09,
2876
+ "grad_norm": 2.8317598164442632,
2877
+ "learning_rate": 1.7602671232654755e-06,
2878
+ "loss": 5.4478,
2879
+ "step": 410
2880
+ },
2881
+ {
2882
+ "epoch": 0.09,
2883
+ "grad_norm": 2.8994461696949276,
2884
+ "learning_rate": 1.723736810595461e-06,
2885
+ "loss": 5.6048,
2886
+ "step": 411
2887
+ },
2888
+ {
2889
+ "epoch": 0.09,
2890
+ "grad_norm": 2.848027250263409,
2891
+ "learning_rate": 1.687553753353195e-06,
2892
+ "loss": 5.6236,
2893
+ "step": 412
2894
+ },
2895
+ {
2896
+ "epoch": 0.09,
2897
+ "grad_norm": 2.7979223820102126,
2898
+ "learning_rate": 1.6517194697072903e-06,
2899
+ "loss": 5.5997,
2900
+ "step": 413
2901
+ },
2902
+ {
2903
+ "epoch": 0.09,
2904
+ "grad_norm": 4.406985290020591,
2905
+ "learning_rate": 1.6162354631925203e-06,
2906
+ "loss": 5.6047,
2907
+ "step": 414
2908
+ },
2909
+ {
2910
+ "epoch": 0.1,
2911
+ "grad_norm": 3.2716585943989362,
2912
+ "learning_rate": 1.5811032226467304e-06,
2913
+ "loss": 5.5763,
2914
+ "step": 415
2915
+ },
2916
+ {
2917
+ "epoch": 0.1,
2918
+ "grad_norm": 3.2119530719491967,
2919
+ "learning_rate": 1.5463242221483742e-06,
2920
+ "loss": 5.5576,
2921
+ "step": 416
2922
+ },
2923
+ {
2924
+ "epoch": 0.1,
2925
+ "grad_norm": 2.7525960595562,
2926
+ "learning_rate": 1.511899920954656e-06,
2927
+ "loss": 5.6124,
2928
+ "step": 417
2929
+ },
2930
+ {
2931
+ "epoch": 0.1,
2932
+ "grad_norm": 2.575626661151134,
2933
+ "learning_rate": 1.4778317634403082e-06,
2934
+ "loss": 5.5585,
2935
+ "step": 418
2936
+ },
2937
+ {
2938
+ "epoch": 0.1,
2939
+ "grad_norm": 3.109772746844713,
2940
+ "learning_rate": 1.4441211790369892e-06,
2941
+ "loss": 5.6023,
2942
+ "step": 419
2943
+ },
2944
+ {
2945
+ "epoch": 0.1,
2946
+ "grad_norm": 3.48770216871105,
2947
+ "learning_rate": 1.4107695821733026e-06,
2948
+ "loss": 5.6905,
2949
+ "step": 420
2950
+ },
2951
+ {
2952
+ "epoch": 0.1,
2953
+ "grad_norm": 2.776399817120803,
2954
+ "learning_rate": 1.3777783722154603e-06,
2955
+ "loss": 5.5346,
2956
+ "step": 421
2957
+ },
2958
+ {
2959
+ "epoch": 0.1,
2960
+ "grad_norm": 2.689995264190314,
2961
+ "learning_rate": 1.3451489334085555e-06,
2962
+ "loss": 5.776,
2963
+ "step": 422
2964
+ },
2965
+ {
2966
+ "epoch": 0.1,
2967
+ "grad_norm": 2.9589020357519855,
2968
+ "learning_rate": 1.3128826348184886e-06,
2969
+ "loss": 5.5889,
2970
+ "step": 423
2971
+ },
2972
+ {
2973
+ "epoch": 0.1,
2974
+ "grad_norm": 3.926701278155943,
2975
+ "learning_rate": 1.2809808302745298e-06,
2976
+ "loss": 5.4377,
2977
+ "step": 424
2978
+ },
2979
+ {
2980
+ "epoch": 0.1,
2981
+ "grad_norm": 8.614907788515115,
2982
+ "learning_rate": 1.249444858312502e-06,
2983
+ "loss": 5.3811,
2984
+ "step": 425
2985
+ },
2986
+ {
2987
+ "epoch": 0.1,
2988
+ "grad_norm": 2.844919169807355,
2989
+ "learning_rate": 1.218276042118629e-06,
2990
+ "loss": 5.3662,
2991
+ "step": 426
2992
+ },
2993
+ {
2994
+ "epoch": 0.1,
2995
+ "grad_norm": 3.200788157807446,
2996
+ "learning_rate": 1.1874756894740137e-06,
2997
+ "loss": 5.5491,
2998
+ "step": 427
2999
+ },
3000
+ {
3001
+ "epoch": 0.1,
3002
+ "grad_norm": 2.8493126527965655,
3003
+ "learning_rate": 1.1570450926997657e-06,
3004
+ "loss": 5.6191,
3005
+ "step": 428
3006
+ },
3007
+ {
3008
+ "epoch": 0.1,
3009
+ "grad_norm": 5.199235991842511,
3010
+ "learning_rate": 1.1269855286027798e-06,
3011
+ "loss": 5.6003,
3012
+ "step": 429
3013
+ },
3014
+ {
3015
+ "epoch": 0.1,
3016
+ "grad_norm": 2.577166179252903,
3017
+ "learning_rate": 1.0972982584221592e-06,
3018
+ "loss": 5.5332,
3019
+ "step": 430
3020
+ },
3021
+ {
3022
+ "epoch": 0.1,
3023
+ "grad_norm": 2.5080571141871837,
3024
+ "learning_rate": 1.067984527776309e-06,
3025
+ "loss": 5.5226,
3026
+ "step": 431
3027
+ },
3028
+ {
3029
+ "epoch": 0.1,
3030
+ "grad_norm": 2.9156280169840563,
3031
+ "learning_rate": 1.0390455666106547e-06,
3032
+ "loss": 5.4879,
3033
+ "step": 432
3034
+ },
3035
+ {
3036
+ "epoch": 0.1,
3037
+ "grad_norm": 3.3439938592575094,
3038
+ "learning_rate": 1.010482589146048e-06,
3039
+ "loss": 5.5742,
3040
+ "step": 433
3041
+ },
3042
+ {
3043
+ "epoch": 0.1,
3044
+ "grad_norm": 2.70223178081651,
3045
+ "learning_rate": 9.822967938278172e-07,
3046
+ "loss": 5.53,
3047
+ "step": 434
3048
+ },
3049
+ {
3050
+ "epoch": 0.1,
3051
+ "grad_norm": 2.910070747579829,
3052
+ "learning_rate": 9.544893632754816e-07,
3053
+ "loss": 5.5037,
3054
+ "step": 435
3055
+ },
3056
+ {
3057
+ "epoch": 0.1,
3058
+ "grad_norm": 2.859773242243526,
3059
+ "learning_rate": 9.270614642331377e-07,
3060
+ "loss": 5.5309,
3061
+ "step": 436
3062
+ },
3063
+ {
3064
+ "epoch": 0.1,
3065
+ "grad_norm": 2.990603251521546,
3066
+ "learning_rate": 9.000142475204965e-07,
3067
+ "loss": 5.4596,
3068
+ "step": 437
3069
+ },
3070
+ {
3071
+ "epoch": 0.1,
3072
+ "grad_norm": 11.13467016828935,
3073
+ "learning_rate": 8.733488479845997e-07,
3074
+ "loss": 5.4875,
3075
+ "step": 438
3076
+ },
3077
+ {
3078
+ "epoch": 0.1,
3079
+ "grad_norm": 3.499523656417884,
3080
+ "learning_rate": 8.470663844522053e-07,
3081
+ "loss": 5.5061,
3082
+ "step": 439
3083
+ },
3084
+ {
3085
+ "epoch": 0.1,
3086
+ "grad_norm": 3.2234243851466418,
3087
+ "learning_rate": 8.211679596828481e-07,
3088
+ "loss": 5.5508,
3089
+ "step": 440
3090
+ },
3091
+ {
3092
+ "epoch": 0.1,
3093
+ "grad_norm": 3.689783836568065,
3094
+ "learning_rate": 7.956546603225601e-07,
3095
+ "loss": 5.4979,
3096
+ "step": 441
3097
+ },
3098
+ {
3099
+ "epoch": 0.1,
3100
+ "grad_norm": 3.550751273965273,
3101
+ "learning_rate": 7.705275568582848e-07,
3102
+ "loss": 5.6264,
3103
+ "step": 442
3104
+ },
3105
+ {
3106
+ "epoch": 0.1,
3107
+ "grad_norm": 2.721521824889205,
3108
+ "learning_rate": 7.457877035729588e-07,
3109
+ "loss": 5.5113,
3110
+ "step": 443
3111
+ },
3112
+ {
3113
+ "epoch": 0.1,
3114
+ "grad_norm": 2.768024715912084,
3115
+ "learning_rate": 7.21436138501278e-07,
3116
+ "loss": 5.5389,
3117
+ "step": 444
3118
+ },
3119
+ {
3120
+ "epoch": 0.1,
3121
+ "grad_norm": 2.6088969137232456,
3122
+ "learning_rate": 6.974738833861383e-07,
3123
+ "loss": 5.4611,
3124
+ "step": 445
3125
+ },
3126
+ {
3127
+ "epoch": 0.1,
3128
+ "grad_norm": 2.693525018419924,
3129
+ "learning_rate": 6.739019436357774e-07,
3130
+ "loss": 5.5317,
3131
+ "step": 446
3132
+ },
3133
+ {
3134
+ "epoch": 0.1,
3135
+ "grad_norm": 2.5671066935530367,
3136
+ "learning_rate": 6.507213082815745e-07,
3137
+ "loss": 5.5331,
3138
+ "step": 447
3139
+ },
3140
+ {
3141
+ "epoch": 0.1,
3142
+ "grad_norm": 3.4210107522987996,
3143
+ "learning_rate": 6.279329499365649e-07,
3144
+ "loss": 5.6906,
3145
+ "step": 448
3146
+ },
3147
+ {
3148
+ "epoch": 0.1,
3149
+ "grad_norm": 2.772985182537546,
3150
+ "learning_rate": 6.055378247546217e-07,
3151
+ "loss": 5.6865,
3152
+ "step": 449
3153
+ },
3154
+ {
3155
+ "epoch": 0.1,
3156
+ "grad_norm": 3.298986027174084,
3157
+ "learning_rate": 5.835368723903456e-07,
3158
+ "loss": 5.5719,
3159
+ "step": 450
3160
+ },
3161
+ {
3162
+ "epoch": 0.1,
3163
+ "grad_norm": 2.715794146885051,
3164
+ "learning_rate": 5.619310159596358e-07,
3165
+ "loss": 5.6766,
3166
+ "step": 451
3167
+ },
3168
+ {
3169
+ "epoch": 0.1,
3170
+ "grad_norm": 3.0933470216619243,
3171
+ "learning_rate": 5.407211620009545e-07,
3172
+ "loss": 5.5423,
3173
+ "step": 452
3174
+ },
3175
+ {
3176
+ "epoch": 0.1,
3177
+ "grad_norm": 4.143643936396307,
3178
+ "learning_rate": 5.199082004372958e-07,
3179
+ "loss": 5.3755,
3180
+ "step": 453
3181
+ },
3182
+ {
3183
+ "epoch": 0.1,
3184
+ "grad_norm": 2.868440457281347,
3185
+ "learning_rate": 4.994930045388414e-07,
3186
+ "loss": 5.5415,
3187
+ "step": 454
3188
+ },
3189
+ {
3190
+ "epoch": 0.1,
3191
+ "grad_norm": 2.9560966126847372,
3192
+ "learning_rate": 4.794764308863242e-07,
3193
+ "loss": 5.636,
3194
+ "step": 455
3195
+ },
3196
+ {
3197
+ "epoch": 0.1,
3198
+ "grad_norm": 3.658181273470064,
3199
+ "learning_rate": 4.5985931933508757e-07,
3200
+ "loss": 5.6803,
3201
+ "step": 456
3202
+ },
3203
+ {
3204
+ "epoch": 0.1,
3205
+ "grad_norm": 3.423368722688064,
3206
+ "learning_rate": 4.406424929798403e-07,
3207
+ "loss": 5.4089,
3208
+ "step": 457
3209
+ },
3210
+ {
3211
+ "epoch": 0.11,
3212
+ "grad_norm": 3.0534636300165188,
3213
+ "learning_rate": 4.218267581201296e-07,
3214
+ "loss": 5.5947,
3215
+ "step": 458
3216
+ },
3217
+ {
3218
+ "epoch": 0.11,
3219
+ "grad_norm": 4.345779139508837,
3220
+ "learning_rate": 4.034129042265067e-07,
3221
+ "loss": 5.5494,
3222
+ "step": 459
3223
+ },
3224
+ {
3225
+ "epoch": 0.11,
3226
+ "grad_norm": 2.733287289238342,
3227
+ "learning_rate": 3.8540170390740097e-07,
3228
+ "loss": 5.3952,
3229
+ "step": 460
3230
+ },
3231
+ {
3232
+ "epoch": 0.11,
3233
+ "grad_norm": 2.578794557649385,
3234
+ "learning_rate": 3.67793912876705e-07,
3235
+ "loss": 5.4544,
3236
+ "step": 461
3237
+ },
3238
+ {
3239
+ "epoch": 0.11,
3240
+ "grad_norm": 6.462067447086064,
3241
+ "learning_rate": 3.5059026992206645e-07,
3242
+ "loss": 5.4354,
3243
+ "step": 462
3244
+ },
3245
+ {
3246
+ "epoch": 0.11,
3247
+ "grad_norm": 2.6125800075217755,
3248
+ "learning_rate": 3.3379149687388866e-07,
3249
+ "loss": 5.3488,
3250
+ "step": 463
3251
+ },
3252
+ {
3253
+ "epoch": 0.11,
3254
+ "grad_norm": 2.7049049281222506,
3255
+ "learning_rate": 3.1739829857504235e-07,
3256
+ "loss": 5.5727,
3257
+ "step": 464
3258
+ },
3259
+ {
3260
+ "epoch": 0.11,
3261
+ "grad_norm": 3.620043130272865,
3262
+ "learning_rate": 3.0141136285129825e-07,
3263
+ "loss": 5.6706,
3264
+ "step": 465
3265
+ },
3266
+ {
3267
+ "epoch": 0.11,
3268
+ "grad_norm": 3.0809997678670187,
3269
+ "learning_rate": 2.8583136048245697e-07,
3270
+ "loss": 5.7061,
3271
+ "step": 466
3272
+ },
3273
+ {
3274
+ "epoch": 0.11,
3275
+ "grad_norm": 3.1616078894179482,
3276
+ "learning_rate": 2.706589451742181e-07,
3277
+ "loss": 5.6098,
3278
+ "step": 467
3279
+ },
3280
+ {
3281
+ "epoch": 0.11,
3282
+ "grad_norm": 2.6779003366529275,
3283
+ "learning_rate": 2.5589475353073987e-07,
3284
+ "loss": 5.4707,
3285
+ "step": 468
3286
+ },
3287
+ {
3288
+ "epoch": 0.11,
3289
+ "grad_norm": 6.707485770015231,
3290
+ "learning_rate": 2.4153940502793185e-07,
3291
+ "loss": 5.4335,
3292
+ "step": 469
3293
+ },
3294
+ {
3295
+ "epoch": 0.11,
3296
+ "grad_norm": 3.1605981177662383,
3297
+ "learning_rate": 2.2759350198746978e-07,
3298
+ "loss": 5.5259,
3299
+ "step": 470
3300
+ },
3301
+ {
3302
+ "epoch": 0.11,
3303
+ "grad_norm": 3.014186395148381,
3304
+ "learning_rate": 2.1405762955151178e-07,
3305
+ "loss": 5.6467,
3306
+ "step": 471
3307
+ },
3308
+ {
3309
+ "epoch": 0.11,
3310
+ "grad_norm": 2.889242425776768,
3311
+ "learning_rate": 2.009323556581566e-07,
3312
+ "loss": 5.4662,
3313
+ "step": 472
3314
+ },
3315
+ {
3316
+ "epoch": 0.11,
3317
+ "grad_norm": 3.095182172474526,
3318
+ "learning_rate": 1.8821823101760949e-07,
3319
+ "loss": 5.5991,
3320
+ "step": 473
3321
+ },
3322
+ {
3323
+ "epoch": 0.11,
3324
+ "grad_norm": 2.8025822861133,
3325
+ "learning_rate": 1.7591578908907724e-07,
3326
+ "loss": 5.6508,
3327
+ "step": 474
3328
+ },
3329
+ {
3330
+ "epoch": 0.11,
3331
+ "grad_norm": 3.052158040260495,
3332
+ "learning_rate": 1.6402554605838173e-07,
3333
+ "loss": 5.6078,
3334
+ "step": 475
3335
+ },
3336
+ {
3337
+ "epoch": 0.11,
3338
+ "grad_norm": 3.0334351565612603,
3339
+ "learning_rate": 1.5254800081630828e-07,
3340
+ "loss": 5.6521,
3341
+ "step": 476
3342
+ },
3343
+ {
3344
+ "epoch": 0.11,
3345
+ "grad_norm": 3.2390067488864203,
3346
+ "learning_rate": 1.4148363493766803e-07,
3347
+ "loss": 5.5932,
3348
+ "step": 477
3349
+ },
3350
+ {
3351
+ "epoch": 0.11,
3352
+ "grad_norm": 2.6513576443759246,
3353
+ "learning_rate": 1.30832912661093e-07,
3354
+ "loss": 5.4863,
3355
+ "step": 478
3356
+ },
3357
+ {
3358
+ "epoch": 0.11,
3359
+ "grad_norm": 2.8858856433117337,
3360
+ "learning_rate": 1.2059628086956044e-07,
3361
+ "loss": 5.6502,
3362
+ "step": 479
3363
+ },
3364
+ {
3365
+ "epoch": 0.11,
3366
+ "grad_norm": 6.026494893009913,
3367
+ "learning_rate": 1.1077416907163573e-07,
3368
+ "loss": 5.6161,
3369
+ "step": 480
3370
+ },
3371
+ {
3372
+ "epoch": 0.11,
3373
+ "grad_norm": 3.269750470433105,
3374
+ "learning_rate": 1.0136698938346012e-07,
3375
+ "loss": 5.6146,
3376
+ "step": 481
3377
+ },
3378
+ {
3379
+ "epoch": 0.11,
3380
+ "grad_norm": 2.915176554260305,
3381
+ "learning_rate": 9.237513651145224e-08,
3382
+ "loss": 5.4934,
3383
+ "step": 482
3384
+ },
3385
+ {
3386
+ "epoch": 0.11,
3387
+ "grad_norm": 2.703171232414342,
3388
+ "learning_rate": 8.379898773574924e-08,
3389
+ "loss": 5.4725,
3390
+ "step": 483
3391
+ },
3392
+ {
3393
+ "epoch": 0.11,
3394
+ "grad_norm": 2.955963983229779,
3395
+ "learning_rate": 7.563890289437825e-08,
3396
+ "loss": 5.7339,
3397
+ "step": 484
3398
+ },
3399
+ {
3400
+ "epoch": 0.11,
3401
+ "grad_norm": 3.089224348103812,
3402
+ "learning_rate": 6.78952243681541e-08,
3403
+ "loss": 5.5206,
3404
+ "step": 485
3405
+ },
3406
+ {
3407
+ "epoch": 0.11,
3408
+ "grad_norm": 3.1319980991851524,
3409
+ "learning_rate": 6.056827706632185e-08,
3410
+ "loss": 5.3931,
3411
+ "step": 486
3412
+ },
3413
+ {
3414
+ "epoch": 0.11,
3415
+ "grad_norm": 3.2761777556196376,
3416
+ "learning_rate": 5.365836841291439e-08,
3417
+ "loss": 5.4341,
3418
+ "step": 487
3419
+ },
3420
+ {
3421
+ "epoch": 0.11,
3422
+ "grad_norm": 2.8804231914855665,
3423
+ "learning_rate": 4.716578833386054e-08,
3424
+ "loss": 5.3863,
3425
+ "step": 488
3426
+ },
3427
+ {
3428
+ "epoch": 0.11,
3429
+ "grad_norm": 2.2873446601496052,
3430
+ "learning_rate": 4.109080924481479e-08,
3431
+ "loss": 5.4484,
3432
+ "step": 489
3433
+ },
3434
+ {
3435
+ "epoch": 0.11,
3436
+ "grad_norm": 3.110318580535298,
3437
+ "learning_rate": 3.543368603973529e-08,
3438
+ "loss": 5.5471,
3439
+ "step": 490
3440
+ },
3441
+ {
3442
+ "epoch": 0.11,
3443
+ "grad_norm": 3.791270347470137,
3444
+ "learning_rate": 3.019465608018024e-08,
3445
+ "loss": 5.5891,
3446
+ "step": 491
3447
+ },
3448
+ {
3449
+ "epoch": 0.11,
3450
+ "grad_norm": 2.6716828750765056,
3451
+ "learning_rate": 2.537393918535358e-08,
3452
+ "loss": 5.5102,
3453
+ "step": 492
3454
+ },
3455
+ {
3456
+ "epoch": 0.11,
3457
+ "grad_norm": 4.590128972359705,
3458
+ "learning_rate": 2.0971737622883515e-08,
3459
+ "loss": 5.6497,
3460
+ "step": 493
3461
+ },
3462
+ {
3463
+ "epoch": 0.11,
3464
+ "grad_norm": 2.491730237188739,
3465
+ "learning_rate": 1.698823610032929e-08,
3466
+ "loss": 5.5464,
3467
+ "step": 494
3468
+ },
3469
+ {
3470
+ "epoch": 0.11,
3471
+ "grad_norm": 2.541793206348526,
3472
+ "learning_rate": 1.3423601757436289e-08,
3473
+ "loss": 5.5446,
3474
+ "step": 495
3475
+ },
3476
+ {
3477
+ "epoch": 0.11,
3478
+ "grad_norm": 4.43180997277326,
3479
+ "learning_rate": 1.0277984159122734e-08,
3480
+ "loss": 5.5588,
3481
+ "step": 496
3482
+ },
3483
+ {
3484
+ "epoch": 0.11,
3485
+ "grad_norm": 2.7657573951340706,
3486
+ "learning_rate": 7.551515289203615e-09,
3487
+ "loss": 5.6594,
3488
+ "step": 497
3489
+ },
3490
+ {
3491
+ "epoch": 0.11,
3492
+ "grad_norm": 2.8695545506323783,
3493
+ "learning_rate": 5.2443095448506674e-09,
3494
+ "loss": 5.625,
3495
+ "step": 498
3496
+ },
3497
+ {
3498
+ "epoch": 0.11,
3499
+ "grad_norm": 3.165811767433061,
3500
+ "learning_rate": 3.3564637317984318e-09,
3501
+ "loss": 5.4195,
3502
+ "step": 499
3503
+ },
3504
+ {
3505
+ "epoch": 0.11,
3506
+ "grad_norm": 2.9299117263430903,
3507
+ "learning_rate": 1.888057060274173e-09,
3508
+ "loss": 5.6383,
3509
+ "step": 500
3510
+ }
3511
+ ],
3512
+ "logging_steps": 1.0,
3513
+ "max_steps": 500,
3514
+ "num_input_tokens_seen": 0,
3515
+ "num_train_epochs": 1,
3516
+ "save_steps": 25,
3517
+ "total_flos": 2968550903808.0,
3518
+ "train_batch_size": 32,
3519
+ "trial_name": null,
3520
+ "trial_params": null
3521
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
zero_to_fp32.py ADDED
@@ -0,0 +1,604 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
215
+ exclude_frozen_parameters)
216
+ elif zero_stage == 3:
217
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
218
+ exclude_frozen_parameters)
219
+
220
+
221
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
222
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
223
+ return
224
+
225
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
226
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
227
+
228
+ if debug:
229
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
230
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
231
+
232
+ wanted_params = len(frozen_param_shapes)
233
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
234
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
235
+ print(f'Frozen params: Have {avail_numel} numels to process.')
236
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
237
+
238
+ total_params = 0
239
+ total_numel = 0
240
+ for name, shape in frozen_param_shapes.items():
241
+ total_params += 1
242
+ unpartitioned_numel = shape.numel()
243
+ total_numel += unpartitioned_numel
244
+
245
+ state_dict[name] = frozen_param_fragments[name]
246
+
247
+ if debug:
248
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
249
+
250
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
251
+
252
+
253
+ def _has_callable(obj, fn):
254
+ attr = getattr(obj, fn, None)
255
+ return callable(attr)
256
+
257
+
258
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
259
+ param_shapes = zero_model_states[0].param_shapes
260
+
261
+ # Reconstruction protocol:
262
+ #
263
+ # XXX: document this
264
+
265
+ if debug:
266
+ for i in range(world_size):
267
+ for j in range(len(fp32_flat_groups[0])):
268
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
269
+
270
+ # XXX: memory usage doubles here (zero2)
271
+ num_param_groups = len(fp32_flat_groups[0])
272
+ merged_single_partition_of_fp32_groups = []
273
+ for i in range(num_param_groups):
274
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
275
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
276
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
277
+ avail_numel = sum(
278
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
279
+
280
+ if debug:
281
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
282
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
283
+ # not asserting if there is a mismatch due to possible padding
284
+ print(f"Have {avail_numel} numels to process.")
285
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
286
+
287
+ # params
288
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
289
+ # out-of-core computing solution
290
+ total_numel = 0
291
+ total_params = 0
292
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
293
+ offset = 0
294
+ avail_numel = full_single_fp32_vector.numel()
295
+ for name, shape in shapes.items():
296
+
297
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
298
+ total_numel += unpartitioned_numel
299
+ total_params += 1
300
+
301
+ if debug:
302
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
303
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
304
+ offset += unpartitioned_numel
305
+
306
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
307
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
308
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
309
+ # live optimizer object, so we are checking that the numbers are within the right range
310
+ align_to = 2 * world_size
311
+
312
+ def zero2_align(x):
313
+ return align_to * math.ceil(x / align_to)
314
+
315
+ if debug:
316
+ print(f"original offset={offset}, avail_numel={avail_numel}")
317
+
318
+ offset = zero2_align(offset)
319
+ avail_numel = zero2_align(avail_numel)
320
+
321
+ if debug:
322
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
323
+
324
+ # Sanity check
325
+ if offset != avail_numel:
326
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
327
+
328
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
329
+
330
+
331
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
332
+ exclude_frozen_parameters):
333
+ state_dict = OrderedDict()
334
+
335
+ # buffers
336
+ buffers = zero_model_states[0].buffers
337
+ state_dict.update(buffers)
338
+ if debug:
339
+ print(f"added {len(buffers)} buffers")
340
+
341
+ if not exclude_frozen_parameters:
342
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
343
+
344
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
345
+
346
+ # recover shared parameters
347
+ for pair in zero_model_states[0].shared_params:
348
+ if pair[1] in state_dict:
349
+ state_dict[pair[0]] = state_dict[pair[1]]
350
+
351
+ return state_dict
352
+
353
+
354
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
355
+ remainder = unpartitioned_numel % world_size
356
+ padding_numel = (world_size - remainder) if remainder else 0
357
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
358
+ return partitioned_numel, padding_numel
359
+
360
+
361
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
362
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
363
+ return
364
+
365
+ if debug:
366
+ for i in range(world_size):
367
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
368
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
369
+
370
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
371
+ wanted_params = len(frozen_param_shapes)
372
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
373
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
374
+ print(f'Frozen params: Have {avail_numel} numels to process.')
375
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
376
+
377
+ total_params = 0
378
+ total_numel = 0
379
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
380
+ total_params += 1
381
+ unpartitioned_numel = shape.numel()
382
+ total_numel += unpartitioned_numel
383
+
384
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
385
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
386
+
387
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
388
+
389
+ if debug:
390
+ print(
391
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
392
+ )
393
+
394
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
395
+
396
+
397
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
398
+ param_shapes = zero_model_states[0].param_shapes
399
+ avail_numel = fp32_flat_groups[0].numel() * world_size
400
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
401
+ # param, re-consolidating each param, while dealing with padding if any
402
+
403
+ # merge list of dicts, preserving order
404
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
405
+
406
+ if debug:
407
+ for i in range(world_size):
408
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
409
+
410
+ wanted_params = len(param_shapes)
411
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
412
+ # not asserting if there is a mismatch due to possible padding
413
+ avail_numel = fp32_flat_groups[0].numel() * world_size
414
+ print(f"Trainable params: Have {avail_numel} numels to process.")
415
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
416
+
417
+ # params
418
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
419
+ # out-of-core computing solution
420
+ offset = 0
421
+ total_numel = 0
422
+ total_params = 0
423
+ for name, shape in param_shapes.items():
424
+
425
+ unpartitioned_numel = shape.numel()
426
+ total_numel += unpartitioned_numel
427
+ total_params += 1
428
+
429
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
430
+
431
+ if debug:
432
+ print(
433
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
434
+ )
435
+
436
+ # XXX: memory usage doubles here
437
+ state_dict[name] = torch.cat(
438
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
439
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
440
+ offset += partitioned_numel
441
+
442
+ offset *= world_size
443
+
444
+ # Sanity check
445
+ if offset != avail_numel:
446
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
447
+
448
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
449
+
450
+
451
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
452
+ exclude_frozen_parameters):
453
+ state_dict = OrderedDict()
454
+
455
+ # buffers
456
+ buffers = zero_model_states[0].buffers
457
+ state_dict.update(buffers)
458
+ if debug:
459
+ print(f"added {len(buffers)} buffers")
460
+
461
+ if not exclude_frozen_parameters:
462
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
463
+
464
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
465
+
466
+ # recover shared parameters
467
+ for pair in zero_model_states[0].shared_params:
468
+ if pair[1] in state_dict:
469
+ state_dict[pair[0]] = state_dict[pair[1]]
470
+
471
+ return state_dict
472
+
473
+
474
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
475
+ """
476
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
477
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
478
+ via a model hub.
479
+
480
+ Args:
481
+ - ``checkpoint_dir``: path to the desired checkpoint folder
482
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
483
+ - ``exclude_frozen_parameters``: exclude frozen parameters
484
+
485
+ Returns:
486
+ - pytorch ``state_dict``
487
+
488
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
489
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
490
+ the checkpoint.
491
+
492
+ A typical usage might be ::
493
+
494
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
495
+ # do the training and checkpoint saving
496
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
497
+ model = model.cpu() # move to cpu
498
+ model.load_state_dict(state_dict)
499
+ # submit to model hub or save the model to share with others
500
+
501
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
502
+ application. i.e. you will need to re-initialize the deepspeed engine, since
503
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
504
+
505
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
506
+
507
+ """
508
+ if tag is None:
509
+ latest_path = os.path.join(checkpoint_dir, 'latest')
510
+ if os.path.isfile(latest_path):
511
+ with open(latest_path, 'r') as fd:
512
+ tag = fd.read().strip()
513
+ else:
514
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
515
+
516
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
517
+
518
+ if not os.path.isdir(ds_checkpoint_dir):
519
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
520
+
521
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
522
+
523
+
524
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
525
+ """
526
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
527
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
528
+
529
+ Args:
530
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
531
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
532
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
533
+ - ``exclude_frozen_parameters``: exclude frozen parameters
534
+ """
535
+
536
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
537
+ print(f"Saving fp32 state dict to {output_file}")
538
+ torch.save(state_dict, output_file)
539
+
540
+
541
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
542
+ """
543
+ 1. Put the provided model to cpu
544
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
545
+ 3. Load it into the provided model
546
+
547
+ Args:
548
+ - ``model``: the model object to update
549
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
550
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
551
+
552
+ Returns:
553
+ - ``model`: modified model
554
+
555
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
556
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
557
+ conveniently placed for you in the checkpoint folder.
558
+
559
+ A typical usage might be ::
560
+
561
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
562
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
563
+ # submit to model hub or save the model to share with others
564
+
565
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
566
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
567
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
568
+
569
+ """
570
+ logger.info(f"Extracting fp32 weights")
571
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
572
+
573
+ logger.info(f"Overwriting model with fp32 weights")
574
+ model = model.cpu()
575
+ model.load_state_dict(state_dict, strict=False)
576
+
577
+ return model
578
+
579
+
580
+ if __name__ == "__main__":
581
+
582
+ parser = argparse.ArgumentParser()
583
+ parser.add_argument("checkpoint_dir",
584
+ type=str,
585
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
586
+ parser.add_argument(
587
+ "output_file",
588
+ type=str,
589
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
590
+ parser.add_argument("-t",
591
+ "--tag",
592
+ type=str,
593
+ default=None,
594
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
595
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
596
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
597
+ args = parser.parse_args()
598
+
599
+ debug = args.debug
600
+
601
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
602
+ args.output_file,
603
+ tag=args.tag,
604
+ exclude_frozen_parameters=args.exclude_frozen_parameters)