Commit
·
d7c8652
0
Parent(s):
Checkpoint at 20000 steps
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- .gitattributes +4 -0
- checkpoint-100/config.json +31 -0
- checkpoint-100/generation_config.json +6 -0
- checkpoint-100/model.safetensors +3 -0
- checkpoint-100/optimizer.pt +3 -0
- checkpoint-100/rng_state.pth +3 -0
- checkpoint-100/scheduler.pt +3 -0
- checkpoint-100/special_tokens_map.json +24 -0
- checkpoint-100/tokenizer.json +0 -0
- checkpoint-100/tokenizer_config.json +215 -0
- checkpoint-100/trainer_state.json +48 -0
- checkpoint-100/training_args.bin +3 -0
- checkpoint-1000/config.json +31 -0
- checkpoint-1000/generation_config.json +6 -0
- checkpoint-1000/model.safetensors +3 -0
- checkpoint-1000/optimizer.pt +3 -0
- checkpoint-1000/rng_state.pth +3 -0
- checkpoint-1000/scheduler.pt +3 -0
- checkpoint-1000/special_tokens_map.json +24 -0
- checkpoint-1000/tokenizer.json +0 -0
- checkpoint-1000/tokenizer_config.json +215 -0
- checkpoint-1000/trainer_state.json +174 -0
- checkpoint-1000/training_args.bin +3 -0
- checkpoint-10000/config.json +31 -0
- checkpoint-10000/generation_config.json +6 -0
- checkpoint-10000/model.safetensors +3 -0
- checkpoint-10000/optimizer.pt +3 -0
- checkpoint-10000/rng_state.pth +3 -0
- checkpoint-10000/scheduler.pt +3 -0
- checkpoint-10000/special_tokens_map.json +24 -0
- checkpoint-10000/tokenizer.json +0 -0
- checkpoint-10000/tokenizer_config.json +215 -0
- checkpoint-10000/trainer_state.json +1434 -0
- checkpoint-10000/training_args.bin +3 -0
- checkpoint-10100/config.json +31 -0
- checkpoint-10100/generation_config.json +6 -0
- checkpoint-10100/model.safetensors +3 -0
- checkpoint-10100/optimizer.pt +3 -0
- checkpoint-10100/rng_state.pth +3 -0
- checkpoint-10100/scheduler.pt +3 -0
- checkpoint-10100/special_tokens_map.json +24 -0
- checkpoint-10100/tokenizer.json +0 -0
- checkpoint-10100/tokenizer_config.json +215 -0
- checkpoint-10100/trainer_state.json +1448 -0
- checkpoint-10100/training_args.bin +3 -0
- checkpoint-10200/config.json +31 -0
- checkpoint-10200/generation_config.json +6 -0
- checkpoint-10200/model.safetensors +3 -0
- checkpoint-10200/optimizer.pt +3 -0
- checkpoint-10200/rng_state.pth +3 -0
.gitattributes
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
2 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
3 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
4 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
checkpoint-100/config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"GPTNeoXForCausalLM"
|
4 |
+
],
|
5 |
+
"attention_bias": true,
|
6 |
+
"attention_dropout": 0.0,
|
7 |
+
"bos_token_id": 0,
|
8 |
+
"classifier_dropout": 0.1,
|
9 |
+
"eos_token_id": 0,
|
10 |
+
"hidden_act": "gelu",
|
11 |
+
"hidden_dropout": 0.0,
|
12 |
+
"hidden_size": 128,
|
13 |
+
"initializer_range": 0.02,
|
14 |
+
"intermediate_size": 512,
|
15 |
+
"layer_norm_eps": 1e-05,
|
16 |
+
"max_position_embeddings": 2048,
|
17 |
+
"model_type": "gpt_neox",
|
18 |
+
"num_attention_heads": 4,
|
19 |
+
"num_hidden_layers": 6,
|
20 |
+
"partial_rotary_factor": 0.25,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000,
|
23 |
+
"rotary_emb_base": 10000,
|
24 |
+
"rotary_pct": 0.25,
|
25 |
+
"tie_word_embeddings": false,
|
26 |
+
"torch_dtype": "bfloat16",
|
27 |
+
"transformers_version": "4.50.3",
|
28 |
+
"use_cache": true,
|
29 |
+
"use_parallel_residual": true,
|
30 |
+
"vocab_size": 50304
|
31 |
+
}
|
checkpoint-100/generation_config.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
+
"eos_token_id": 0,
|
5 |
+
"transformers_version": "4.50.3"
|
6 |
+
}
|
checkpoint-100/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e4c6d6fca71e5aeaf126c9a193343c9a7ca17a8060bf4fc59fbfad7019daca4c
|
3 |
+
size 28144000
|
checkpoint-100/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:944db6a6fa77288d4e8f52747811395d2ec0843513e118961282f08a1ce3398e
|
3 |
+
size 56334394
|
checkpoint-100/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
|
3 |
+
size 14244
|
checkpoint-100/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:467abc10c7d90be39dedcb07aeb5ab431826fa1c3718145ad8ea209f579afd08
|
3 |
+
size 1064
|
checkpoint-100/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<|endoftext|>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "<|endoftext|>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<|endoftext|>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<|endoftext|>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-100/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-100/tokenizer_config.json
ADDED
@@ -0,0 +1,215 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": false,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"add_prefix_space": false,
|
5 |
+
"added_tokens_decoder": {
|
6 |
+
"0": {
|
7 |
+
"content": "<|endoftext|>",
|
8 |
+
"lstrip": false,
|
9 |
+
"normalized": false,
|
10 |
+
"rstrip": false,
|
11 |
+
"single_word": false,
|
12 |
+
"special": true
|
13 |
+
},
|
14 |
+
"1": {
|
15 |
+
"content": "<|padding|>",
|
16 |
+
"lstrip": false,
|
17 |
+
"normalized": false,
|
18 |
+
"rstrip": false,
|
19 |
+
"single_word": false,
|
20 |
+
"special": true
|
21 |
+
},
|
22 |
+
"50254": {
|
23 |
+
"content": " ",
|
24 |
+
"lstrip": false,
|
25 |
+
"normalized": true,
|
26 |
+
"rstrip": false,
|
27 |
+
"single_word": false,
|
28 |
+
"special": false
|
29 |
+
},
|
30 |
+
"50255": {
|
31 |
+
"content": " ",
|
32 |
+
"lstrip": false,
|
33 |
+
"normalized": true,
|
34 |
+
"rstrip": false,
|
35 |
+
"single_word": false,
|
36 |
+
"special": false
|
37 |
+
},
|
38 |
+
"50256": {
|
39 |
+
"content": " ",
|
40 |
+
"lstrip": false,
|
41 |
+
"normalized": true,
|
42 |
+
"rstrip": false,
|
43 |
+
"single_word": false,
|
44 |
+
"special": false
|
45 |
+
},
|
46 |
+
"50257": {
|
47 |
+
"content": " ",
|
48 |
+
"lstrip": false,
|
49 |
+
"normalized": true,
|
50 |
+
"rstrip": false,
|
51 |
+
"single_word": false,
|
52 |
+
"special": false
|
53 |
+
},
|
54 |
+
"50258": {
|
55 |
+
"content": " ",
|
56 |
+
"lstrip": false,
|
57 |
+
"normalized": true,
|
58 |
+
"rstrip": false,
|
59 |
+
"single_word": false,
|
60 |
+
"special": false
|
61 |
+
},
|
62 |
+
"50259": {
|
63 |
+
"content": " ",
|
64 |
+
"lstrip": false,
|
65 |
+
"normalized": true,
|
66 |
+
"rstrip": false,
|
67 |
+
"single_word": false,
|
68 |
+
"special": false
|
69 |
+
},
|
70 |
+
"50260": {
|
71 |
+
"content": " ",
|
72 |
+
"lstrip": false,
|
73 |
+
"normalized": true,
|
74 |
+
"rstrip": false,
|
75 |
+
"single_word": false,
|
76 |
+
"special": false
|
77 |
+
},
|
78 |
+
"50261": {
|
79 |
+
"content": " ",
|
80 |
+
"lstrip": false,
|
81 |
+
"normalized": true,
|
82 |
+
"rstrip": false,
|
83 |
+
"single_word": false,
|
84 |
+
"special": false
|
85 |
+
},
|
86 |
+
"50262": {
|
87 |
+
"content": " ",
|
88 |
+
"lstrip": false,
|
89 |
+
"normalized": true,
|
90 |
+
"rstrip": false,
|
91 |
+
"single_word": false,
|
92 |
+
"special": false
|
93 |
+
},
|
94 |
+
"50263": {
|
95 |
+
"content": " ",
|
96 |
+
"lstrip": false,
|
97 |
+
"normalized": true,
|
98 |
+
"rstrip": false,
|
99 |
+
"single_word": false,
|
100 |
+
"special": false
|
101 |
+
},
|
102 |
+
"50264": {
|
103 |
+
"content": " ",
|
104 |
+
"lstrip": false,
|
105 |
+
"normalized": true,
|
106 |
+
"rstrip": false,
|
107 |
+
"single_word": false,
|
108 |
+
"special": false
|
109 |
+
},
|
110 |
+
"50265": {
|
111 |
+
"content": " ",
|
112 |
+
"lstrip": false,
|
113 |
+
"normalized": true,
|
114 |
+
"rstrip": false,
|
115 |
+
"single_word": false,
|
116 |
+
"special": false
|
117 |
+
},
|
118 |
+
"50266": {
|
119 |
+
"content": " ",
|
120 |
+
"lstrip": false,
|
121 |
+
"normalized": true,
|
122 |
+
"rstrip": false,
|
123 |
+
"single_word": false,
|
124 |
+
"special": false
|
125 |
+
},
|
126 |
+
"50267": {
|
127 |
+
"content": " ",
|
128 |
+
"lstrip": false,
|
129 |
+
"normalized": true,
|
130 |
+
"rstrip": false,
|
131 |
+
"single_word": false,
|
132 |
+
"special": false
|
133 |
+
},
|
134 |
+
"50268": {
|
135 |
+
"content": " ",
|
136 |
+
"lstrip": false,
|
137 |
+
"normalized": true,
|
138 |
+
"rstrip": false,
|
139 |
+
"single_word": false,
|
140 |
+
"special": false
|
141 |
+
},
|
142 |
+
"50269": {
|
143 |
+
"content": " ",
|
144 |
+
"lstrip": false,
|
145 |
+
"normalized": true,
|
146 |
+
"rstrip": false,
|
147 |
+
"single_word": false,
|
148 |
+
"special": false
|
149 |
+
},
|
150 |
+
"50270": {
|
151 |
+
"content": " ",
|
152 |
+
"lstrip": false,
|
153 |
+
"normalized": true,
|
154 |
+
"rstrip": false,
|
155 |
+
"single_word": false,
|
156 |
+
"special": false
|
157 |
+
},
|
158 |
+
"50271": {
|
159 |
+
"content": " ",
|
160 |
+
"lstrip": false,
|
161 |
+
"normalized": true,
|
162 |
+
"rstrip": false,
|
163 |
+
"single_word": false,
|
164 |
+
"special": false
|
165 |
+
},
|
166 |
+
"50272": {
|
167 |
+
"content": " ",
|
168 |
+
"lstrip": false,
|
169 |
+
"normalized": true,
|
170 |
+
"rstrip": false,
|
171 |
+
"single_word": false,
|
172 |
+
"special": false
|
173 |
+
},
|
174 |
+
"50273": {
|
175 |
+
"content": " ",
|
176 |
+
"lstrip": false,
|
177 |
+
"normalized": true,
|
178 |
+
"rstrip": false,
|
179 |
+
"single_word": false,
|
180 |
+
"special": false
|
181 |
+
},
|
182 |
+
"50274": {
|
183 |
+
"content": " ",
|
184 |
+
"lstrip": false,
|
185 |
+
"normalized": true,
|
186 |
+
"rstrip": false,
|
187 |
+
"single_word": false,
|
188 |
+
"special": false
|
189 |
+
},
|
190 |
+
"50275": {
|
191 |
+
"content": " ",
|
192 |
+
"lstrip": false,
|
193 |
+
"normalized": true,
|
194 |
+
"rstrip": false,
|
195 |
+
"single_word": false,
|
196 |
+
"special": false
|
197 |
+
},
|
198 |
+
"50276": {
|
199 |
+
"content": " ",
|
200 |
+
"lstrip": false,
|
201 |
+
"normalized": true,
|
202 |
+
"rstrip": false,
|
203 |
+
"single_word": false,
|
204 |
+
"special": false
|
205 |
+
}
|
206 |
+
},
|
207 |
+
"bos_token": "<|endoftext|>",
|
208 |
+
"clean_up_tokenization_spaces": true,
|
209 |
+
"eos_token": "<|endoftext|>",
|
210 |
+
"extra_special_tokens": {},
|
211 |
+
"model_max_length": 2048,
|
212 |
+
"pad_token": "<|endoftext|>",
|
213 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
214 |
+
"unk_token": "<|endoftext|>"
|
215 |
+
}
|
checkpoint-100/trainer_state.json
ADDED
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_global_step": null,
|
3 |
+
"best_metric": null,
|
4 |
+
"best_model_checkpoint": null,
|
5 |
+
"epoch": 0.005,
|
6 |
+
"eval_steps": 500,
|
7 |
+
"global_step": 100,
|
8 |
+
"is_hyper_param_search": false,
|
9 |
+
"is_local_process_zero": true,
|
10 |
+
"is_world_process_zero": true,
|
11 |
+
"log_history": [
|
12 |
+
{
|
13 |
+
"epoch": 0.0025,
|
14 |
+
"grad_norm": 1.0703125,
|
15 |
+
"learning_rate": 0.00025,
|
16 |
+
"loss": 10.4938,
|
17 |
+
"step": 50
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 0.005,
|
21 |
+
"grad_norm": 0.90625,
|
22 |
+
"learning_rate": 0.0005,
|
23 |
+
"loss": 9.1324,
|
24 |
+
"step": 100
|
25 |
+
}
|
26 |
+
],
|
27 |
+
"logging_steps": 50,
|
28 |
+
"max_steps": 20000,
|
29 |
+
"num_input_tokens_seen": 0,
|
30 |
+
"num_train_epochs": 9223372036854775807,
|
31 |
+
"save_steps": 100,
|
32 |
+
"stateful_callbacks": {
|
33 |
+
"TrainerControl": {
|
34 |
+
"args": {
|
35 |
+
"should_epoch_stop": false,
|
36 |
+
"should_evaluate": false,
|
37 |
+
"should_log": false,
|
38 |
+
"should_save": true,
|
39 |
+
"should_training_stop": false
|
40 |
+
},
|
41 |
+
"attributes": {}
|
42 |
+
}
|
43 |
+
},
|
44 |
+
"total_flos": 144915952435200.0,
|
45 |
+
"train_batch_size": 16,
|
46 |
+
"trial_name": null,
|
47 |
+
"trial_params": null
|
48 |
+
}
|
checkpoint-100/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:83031100336f2dd4b43e2539aae80864690fb3fc70607039ce5c50a2ff4a21bb
|
3 |
+
size 5368
|
checkpoint-1000/config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"GPTNeoXForCausalLM"
|
4 |
+
],
|
5 |
+
"attention_bias": true,
|
6 |
+
"attention_dropout": 0.0,
|
7 |
+
"bos_token_id": 0,
|
8 |
+
"classifier_dropout": 0.1,
|
9 |
+
"eos_token_id": 0,
|
10 |
+
"hidden_act": "gelu",
|
11 |
+
"hidden_dropout": 0.0,
|
12 |
+
"hidden_size": 128,
|
13 |
+
"initializer_range": 0.02,
|
14 |
+
"intermediate_size": 512,
|
15 |
+
"layer_norm_eps": 1e-05,
|
16 |
+
"max_position_embeddings": 2048,
|
17 |
+
"model_type": "gpt_neox",
|
18 |
+
"num_attention_heads": 4,
|
19 |
+
"num_hidden_layers": 6,
|
20 |
+
"partial_rotary_factor": 0.25,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000,
|
23 |
+
"rotary_emb_base": 10000,
|
24 |
+
"rotary_pct": 0.25,
|
25 |
+
"tie_word_embeddings": false,
|
26 |
+
"torch_dtype": "bfloat16",
|
27 |
+
"transformers_version": "4.50.3",
|
28 |
+
"use_cache": true,
|
29 |
+
"use_parallel_residual": true,
|
30 |
+
"vocab_size": 50304
|
31 |
+
}
|
checkpoint-1000/generation_config.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
+
"eos_token_id": 0,
|
5 |
+
"transformers_version": "4.50.3"
|
6 |
+
}
|
checkpoint-1000/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b939cb088df13e370faac88da48bf500e5065fe5490cbd37feec3c676de701f7
|
3 |
+
size 28144000
|
checkpoint-1000/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2a95ee6c363c2c4983c714b55a3db7b68842d40e0cac96b7b8a1b04422034404
|
3 |
+
size 56334394
|
checkpoint-1000/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
|
3 |
+
size 14244
|
checkpoint-1000/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:00e2d4a3153d331663068b0b25f4f389a2586c4670a05386af2b11671ba6ad7c
|
3 |
+
size 1064
|
checkpoint-1000/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<|endoftext|>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "<|endoftext|>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<|endoftext|>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<|endoftext|>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-1000/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-1000/tokenizer_config.json
ADDED
@@ -0,0 +1,215 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": false,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"add_prefix_space": false,
|
5 |
+
"added_tokens_decoder": {
|
6 |
+
"0": {
|
7 |
+
"content": "<|endoftext|>",
|
8 |
+
"lstrip": false,
|
9 |
+
"normalized": false,
|
10 |
+
"rstrip": false,
|
11 |
+
"single_word": false,
|
12 |
+
"special": true
|
13 |
+
},
|
14 |
+
"1": {
|
15 |
+
"content": "<|padding|>",
|
16 |
+
"lstrip": false,
|
17 |
+
"normalized": false,
|
18 |
+
"rstrip": false,
|
19 |
+
"single_word": false,
|
20 |
+
"special": true
|
21 |
+
},
|
22 |
+
"50254": {
|
23 |
+
"content": " ",
|
24 |
+
"lstrip": false,
|
25 |
+
"normalized": true,
|
26 |
+
"rstrip": false,
|
27 |
+
"single_word": false,
|
28 |
+
"special": false
|
29 |
+
},
|
30 |
+
"50255": {
|
31 |
+
"content": " ",
|
32 |
+
"lstrip": false,
|
33 |
+
"normalized": true,
|
34 |
+
"rstrip": false,
|
35 |
+
"single_word": false,
|
36 |
+
"special": false
|
37 |
+
},
|
38 |
+
"50256": {
|
39 |
+
"content": " ",
|
40 |
+
"lstrip": false,
|
41 |
+
"normalized": true,
|
42 |
+
"rstrip": false,
|
43 |
+
"single_word": false,
|
44 |
+
"special": false
|
45 |
+
},
|
46 |
+
"50257": {
|
47 |
+
"content": " ",
|
48 |
+
"lstrip": false,
|
49 |
+
"normalized": true,
|
50 |
+
"rstrip": false,
|
51 |
+
"single_word": false,
|
52 |
+
"special": false
|
53 |
+
},
|
54 |
+
"50258": {
|
55 |
+
"content": " ",
|
56 |
+
"lstrip": false,
|
57 |
+
"normalized": true,
|
58 |
+
"rstrip": false,
|
59 |
+
"single_word": false,
|
60 |
+
"special": false
|
61 |
+
},
|
62 |
+
"50259": {
|
63 |
+
"content": " ",
|
64 |
+
"lstrip": false,
|
65 |
+
"normalized": true,
|
66 |
+
"rstrip": false,
|
67 |
+
"single_word": false,
|
68 |
+
"special": false
|
69 |
+
},
|
70 |
+
"50260": {
|
71 |
+
"content": " ",
|
72 |
+
"lstrip": false,
|
73 |
+
"normalized": true,
|
74 |
+
"rstrip": false,
|
75 |
+
"single_word": false,
|
76 |
+
"special": false
|
77 |
+
},
|
78 |
+
"50261": {
|
79 |
+
"content": " ",
|
80 |
+
"lstrip": false,
|
81 |
+
"normalized": true,
|
82 |
+
"rstrip": false,
|
83 |
+
"single_word": false,
|
84 |
+
"special": false
|
85 |
+
},
|
86 |
+
"50262": {
|
87 |
+
"content": " ",
|
88 |
+
"lstrip": false,
|
89 |
+
"normalized": true,
|
90 |
+
"rstrip": false,
|
91 |
+
"single_word": false,
|
92 |
+
"special": false
|
93 |
+
},
|
94 |
+
"50263": {
|
95 |
+
"content": " ",
|
96 |
+
"lstrip": false,
|
97 |
+
"normalized": true,
|
98 |
+
"rstrip": false,
|
99 |
+
"single_word": false,
|
100 |
+
"special": false
|
101 |
+
},
|
102 |
+
"50264": {
|
103 |
+
"content": " ",
|
104 |
+
"lstrip": false,
|
105 |
+
"normalized": true,
|
106 |
+
"rstrip": false,
|
107 |
+
"single_word": false,
|
108 |
+
"special": false
|
109 |
+
},
|
110 |
+
"50265": {
|
111 |
+
"content": " ",
|
112 |
+
"lstrip": false,
|
113 |
+
"normalized": true,
|
114 |
+
"rstrip": false,
|
115 |
+
"single_word": false,
|
116 |
+
"special": false
|
117 |
+
},
|
118 |
+
"50266": {
|
119 |
+
"content": " ",
|
120 |
+
"lstrip": false,
|
121 |
+
"normalized": true,
|
122 |
+
"rstrip": false,
|
123 |
+
"single_word": false,
|
124 |
+
"special": false
|
125 |
+
},
|
126 |
+
"50267": {
|
127 |
+
"content": " ",
|
128 |
+
"lstrip": false,
|
129 |
+
"normalized": true,
|
130 |
+
"rstrip": false,
|
131 |
+
"single_word": false,
|
132 |
+
"special": false
|
133 |
+
},
|
134 |
+
"50268": {
|
135 |
+
"content": " ",
|
136 |
+
"lstrip": false,
|
137 |
+
"normalized": true,
|
138 |
+
"rstrip": false,
|
139 |
+
"single_word": false,
|
140 |
+
"special": false
|
141 |
+
},
|
142 |
+
"50269": {
|
143 |
+
"content": " ",
|
144 |
+
"lstrip": false,
|
145 |
+
"normalized": true,
|
146 |
+
"rstrip": false,
|
147 |
+
"single_word": false,
|
148 |
+
"special": false
|
149 |
+
},
|
150 |
+
"50270": {
|
151 |
+
"content": " ",
|
152 |
+
"lstrip": false,
|
153 |
+
"normalized": true,
|
154 |
+
"rstrip": false,
|
155 |
+
"single_word": false,
|
156 |
+
"special": false
|
157 |
+
},
|
158 |
+
"50271": {
|
159 |
+
"content": " ",
|
160 |
+
"lstrip": false,
|
161 |
+
"normalized": true,
|
162 |
+
"rstrip": false,
|
163 |
+
"single_word": false,
|
164 |
+
"special": false
|
165 |
+
},
|
166 |
+
"50272": {
|
167 |
+
"content": " ",
|
168 |
+
"lstrip": false,
|
169 |
+
"normalized": true,
|
170 |
+
"rstrip": false,
|
171 |
+
"single_word": false,
|
172 |
+
"special": false
|
173 |
+
},
|
174 |
+
"50273": {
|
175 |
+
"content": " ",
|
176 |
+
"lstrip": false,
|
177 |
+
"normalized": true,
|
178 |
+
"rstrip": false,
|
179 |
+
"single_word": false,
|
180 |
+
"special": false
|
181 |
+
},
|
182 |
+
"50274": {
|
183 |
+
"content": " ",
|
184 |
+
"lstrip": false,
|
185 |
+
"normalized": true,
|
186 |
+
"rstrip": false,
|
187 |
+
"single_word": false,
|
188 |
+
"special": false
|
189 |
+
},
|
190 |
+
"50275": {
|
191 |
+
"content": " ",
|
192 |
+
"lstrip": false,
|
193 |
+
"normalized": true,
|
194 |
+
"rstrip": false,
|
195 |
+
"single_word": false,
|
196 |
+
"special": false
|
197 |
+
},
|
198 |
+
"50276": {
|
199 |
+
"content": " ",
|
200 |
+
"lstrip": false,
|
201 |
+
"normalized": true,
|
202 |
+
"rstrip": false,
|
203 |
+
"single_word": false,
|
204 |
+
"special": false
|
205 |
+
}
|
206 |
+
},
|
207 |
+
"bos_token": "<|endoftext|>",
|
208 |
+
"clean_up_tokenization_spaces": true,
|
209 |
+
"eos_token": "<|endoftext|>",
|
210 |
+
"extra_special_tokens": {},
|
211 |
+
"model_max_length": 2048,
|
212 |
+
"pad_token": "<|endoftext|>",
|
213 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
214 |
+
"unk_token": "<|endoftext|>"
|
215 |
+
}
|
checkpoint-1000/trainer_state.json
ADDED
@@ -0,0 +1,174 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_global_step": null,
|
3 |
+
"best_metric": null,
|
4 |
+
"best_model_checkpoint": null,
|
5 |
+
"epoch": 0.05,
|
6 |
+
"eval_steps": 500,
|
7 |
+
"global_step": 1000,
|
8 |
+
"is_hyper_param_search": false,
|
9 |
+
"is_local_process_zero": true,
|
10 |
+
"is_world_process_zero": true,
|
11 |
+
"log_history": [
|
12 |
+
{
|
13 |
+
"epoch": 0.0025,
|
14 |
+
"grad_norm": 1.0703125,
|
15 |
+
"learning_rate": 0.00025,
|
16 |
+
"loss": 10.4938,
|
17 |
+
"step": 50
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 0.005,
|
21 |
+
"grad_norm": 0.90625,
|
22 |
+
"learning_rate": 0.0005,
|
23 |
+
"loss": 9.1324,
|
24 |
+
"step": 100
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.0075,
|
28 |
+
"grad_norm": 0.62890625,
|
29 |
+
"learning_rate": 0.00075,
|
30 |
+
"loss": 7.883,
|
31 |
+
"step": 150
|
32 |
+
},
|
33 |
+
{
|
34 |
+
"epoch": 0.01,
|
35 |
+
"grad_norm": 0.42578125,
|
36 |
+
"learning_rate": 0.001,
|
37 |
+
"loss": 7.3925,
|
38 |
+
"step": 200
|
39 |
+
},
|
40 |
+
{
|
41 |
+
"epoch": 0.0125,
|
42 |
+
"grad_norm": 0.6171875,
|
43 |
+
"learning_rate": 0.0009999842657116666,
|
44 |
+
"loss": 7.062,
|
45 |
+
"step": 250
|
46 |
+
},
|
47 |
+
{
|
48 |
+
"epoch": 0.015,
|
49 |
+
"grad_norm": 0.498046875,
|
50 |
+
"learning_rate": 0.0009999370638369377,
|
51 |
+
"loss": 6.8396,
|
52 |
+
"step": 300
|
53 |
+
},
|
54 |
+
{
|
55 |
+
"epoch": 0.0175,
|
56 |
+
"grad_norm": 0.4296875,
|
57 |
+
"learning_rate": 0.0009998583973465647,
|
58 |
+
"loss": 6.6879,
|
59 |
+
"step": 350
|
60 |
+
},
|
61 |
+
{
|
62 |
+
"epoch": 0.02,
|
63 |
+
"grad_norm": 0.435546875,
|
64 |
+
"learning_rate": 0.0009997482711915926,
|
65 |
+
"loss": 6.6086,
|
66 |
+
"step": 400
|
67 |
+
},
|
68 |
+
{
|
69 |
+
"epoch": 0.0225,
|
70 |
+
"grad_norm": 0.63671875,
|
71 |
+
"learning_rate": 0.0009996066923030483,
|
72 |
+
"loss": 6.5239,
|
73 |
+
"step": 450
|
74 |
+
},
|
75 |
+
{
|
76 |
+
"epoch": 0.025,
|
77 |
+
"grad_norm": 0.37109375,
|
78 |
+
"learning_rate": 0.000999433669591504,
|
79 |
+
"loss": 6.552,
|
80 |
+
"step": 500
|
81 |
+
},
|
82 |
+
{
|
83 |
+
"epoch": 0.0275,
|
84 |
+
"grad_norm": 0.5234375,
|
85 |
+
"learning_rate": 0.0009992292139465165,
|
86 |
+
"loss": 6.4607,
|
87 |
+
"step": 550
|
88 |
+
},
|
89 |
+
{
|
90 |
+
"epoch": 0.03,
|
91 |
+
"grad_norm": 0.61328125,
|
92 |
+
"learning_rate": 0.0009989933382359422,
|
93 |
+
"loss": 6.3974,
|
94 |
+
"step": 600
|
95 |
+
},
|
96 |
+
{
|
97 |
+
"epoch": 0.0325,
|
98 |
+
"grad_norm": 0.67578125,
|
99 |
+
"learning_rate": 0.0009987260573051267,
|
100 |
+
"loss": 6.3249,
|
101 |
+
"step": 650
|
102 |
+
},
|
103 |
+
{
|
104 |
+
"epoch": 0.035,
|
105 |
+
"grad_norm": 0.51953125,
|
106 |
+
"learning_rate": 0.0009984273879759713,
|
107 |
+
"loss": 6.3374,
|
108 |
+
"step": 700
|
109 |
+
},
|
110 |
+
{
|
111 |
+
"epoch": 0.0375,
|
112 |
+
"grad_norm": 0.482421875,
|
113 |
+
"learning_rate": 0.0009980973490458728,
|
114 |
+
"loss": 6.3157,
|
115 |
+
"step": 750
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 0.04,
|
119 |
+
"grad_norm": 0.55859375,
|
120 |
+
"learning_rate": 0.0009977359612865424,
|
121 |
+
"loss": 6.2525,
|
122 |
+
"step": 800
|
123 |
+
},
|
124 |
+
{
|
125 |
+
"epoch": 0.0425,
|
126 |
+
"grad_norm": 0.53125,
|
127 |
+
"learning_rate": 0.0009973432474426967,
|
128 |
+
"loss": 6.2116,
|
129 |
+
"step": 850
|
130 |
+
},
|
131 |
+
{
|
132 |
+
"epoch": 0.045,
|
133 |
+
"grad_norm": 0.439453125,
|
134 |
+
"learning_rate": 0.000996919232230627,
|
135 |
+
"loss": 6.1993,
|
136 |
+
"step": 900
|
137 |
+
},
|
138 |
+
{
|
139 |
+
"epoch": 0.0475,
|
140 |
+
"grad_norm": 0.384765625,
|
141 |
+
"learning_rate": 0.0009964639423366442,
|
142 |
+
"loss": 6.0863,
|
143 |
+
"step": 950
|
144 |
+
},
|
145 |
+
{
|
146 |
+
"epoch": 0.05,
|
147 |
+
"grad_norm": 0.349609375,
|
148 |
+
"learning_rate": 0.0009959774064153978,
|
149 |
+
"loss": 6.049,
|
150 |
+
"step": 1000
|
151 |
+
}
|
152 |
+
],
|
153 |
+
"logging_steps": 50,
|
154 |
+
"max_steps": 20000,
|
155 |
+
"num_input_tokens_seen": 0,
|
156 |
+
"num_train_epochs": 9223372036854775807,
|
157 |
+
"save_steps": 100,
|
158 |
+
"stateful_callbacks": {
|
159 |
+
"TrainerControl": {
|
160 |
+
"args": {
|
161 |
+
"should_epoch_stop": false,
|
162 |
+
"should_evaluate": false,
|
163 |
+
"should_log": false,
|
164 |
+
"should_save": true,
|
165 |
+
"should_training_stop": false
|
166 |
+
},
|
167 |
+
"attributes": {}
|
168 |
+
}
|
169 |
+
},
|
170 |
+
"total_flos": 1455999079219200.0,
|
171 |
+
"train_batch_size": 16,
|
172 |
+
"trial_name": null,
|
173 |
+
"trial_params": null
|
174 |
+
}
|
checkpoint-1000/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:83031100336f2dd4b43e2539aae80864690fb3fc70607039ce5c50a2ff4a21bb
|
3 |
+
size 5368
|
checkpoint-10000/config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"GPTNeoXForCausalLM"
|
4 |
+
],
|
5 |
+
"attention_bias": true,
|
6 |
+
"attention_dropout": 0.0,
|
7 |
+
"bos_token_id": 0,
|
8 |
+
"classifier_dropout": 0.1,
|
9 |
+
"eos_token_id": 0,
|
10 |
+
"hidden_act": "gelu",
|
11 |
+
"hidden_dropout": 0.0,
|
12 |
+
"hidden_size": 128,
|
13 |
+
"initializer_range": 0.02,
|
14 |
+
"intermediate_size": 512,
|
15 |
+
"layer_norm_eps": 1e-05,
|
16 |
+
"max_position_embeddings": 2048,
|
17 |
+
"model_type": "gpt_neox",
|
18 |
+
"num_attention_heads": 4,
|
19 |
+
"num_hidden_layers": 6,
|
20 |
+
"partial_rotary_factor": 0.25,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000,
|
23 |
+
"rotary_emb_base": 10000,
|
24 |
+
"rotary_pct": 0.25,
|
25 |
+
"tie_word_embeddings": false,
|
26 |
+
"torch_dtype": "bfloat16",
|
27 |
+
"transformers_version": "4.50.3",
|
28 |
+
"use_cache": true,
|
29 |
+
"use_parallel_residual": true,
|
30 |
+
"vocab_size": 50304
|
31 |
+
}
|
checkpoint-10000/generation_config.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
+
"eos_token_id": 0,
|
5 |
+
"transformers_version": "4.50.3"
|
6 |
+
}
|
checkpoint-10000/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bc52f122cb3e0f10148ccab9d01a0d326e57e5eb952d0e06a336430a493ecd10
|
3 |
+
size 28144000
|
checkpoint-10000/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ca55db4ef4829c80f74276df6dde2aa4463f2c8062a751640f3fffd2a762caae
|
3 |
+
size 56334394
|
checkpoint-10000/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
|
3 |
+
size 14244
|
checkpoint-10000/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e967a5665eef5c5d2352dcac242d642b03cca63f0fab25e2e1f8d133154ab98a
|
3 |
+
size 1064
|
checkpoint-10000/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<|endoftext|>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "<|endoftext|>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<|endoftext|>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<|endoftext|>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-10000/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-10000/tokenizer_config.json
ADDED
@@ -0,0 +1,215 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": false,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"add_prefix_space": false,
|
5 |
+
"added_tokens_decoder": {
|
6 |
+
"0": {
|
7 |
+
"content": "<|endoftext|>",
|
8 |
+
"lstrip": false,
|
9 |
+
"normalized": false,
|
10 |
+
"rstrip": false,
|
11 |
+
"single_word": false,
|
12 |
+
"special": true
|
13 |
+
},
|
14 |
+
"1": {
|
15 |
+
"content": "<|padding|>",
|
16 |
+
"lstrip": false,
|
17 |
+
"normalized": false,
|
18 |
+
"rstrip": false,
|
19 |
+
"single_word": false,
|
20 |
+
"special": true
|
21 |
+
},
|
22 |
+
"50254": {
|
23 |
+
"content": " ",
|
24 |
+
"lstrip": false,
|
25 |
+
"normalized": true,
|
26 |
+
"rstrip": false,
|
27 |
+
"single_word": false,
|
28 |
+
"special": false
|
29 |
+
},
|
30 |
+
"50255": {
|
31 |
+
"content": " ",
|
32 |
+
"lstrip": false,
|
33 |
+
"normalized": true,
|
34 |
+
"rstrip": false,
|
35 |
+
"single_word": false,
|
36 |
+
"special": false
|
37 |
+
},
|
38 |
+
"50256": {
|
39 |
+
"content": " ",
|
40 |
+
"lstrip": false,
|
41 |
+
"normalized": true,
|
42 |
+
"rstrip": false,
|
43 |
+
"single_word": false,
|
44 |
+
"special": false
|
45 |
+
},
|
46 |
+
"50257": {
|
47 |
+
"content": " ",
|
48 |
+
"lstrip": false,
|
49 |
+
"normalized": true,
|
50 |
+
"rstrip": false,
|
51 |
+
"single_word": false,
|
52 |
+
"special": false
|
53 |
+
},
|
54 |
+
"50258": {
|
55 |
+
"content": " ",
|
56 |
+
"lstrip": false,
|
57 |
+
"normalized": true,
|
58 |
+
"rstrip": false,
|
59 |
+
"single_word": false,
|
60 |
+
"special": false
|
61 |
+
},
|
62 |
+
"50259": {
|
63 |
+
"content": " ",
|
64 |
+
"lstrip": false,
|
65 |
+
"normalized": true,
|
66 |
+
"rstrip": false,
|
67 |
+
"single_word": false,
|
68 |
+
"special": false
|
69 |
+
},
|
70 |
+
"50260": {
|
71 |
+
"content": " ",
|
72 |
+
"lstrip": false,
|
73 |
+
"normalized": true,
|
74 |
+
"rstrip": false,
|
75 |
+
"single_word": false,
|
76 |
+
"special": false
|
77 |
+
},
|
78 |
+
"50261": {
|
79 |
+
"content": " ",
|
80 |
+
"lstrip": false,
|
81 |
+
"normalized": true,
|
82 |
+
"rstrip": false,
|
83 |
+
"single_word": false,
|
84 |
+
"special": false
|
85 |
+
},
|
86 |
+
"50262": {
|
87 |
+
"content": " ",
|
88 |
+
"lstrip": false,
|
89 |
+
"normalized": true,
|
90 |
+
"rstrip": false,
|
91 |
+
"single_word": false,
|
92 |
+
"special": false
|
93 |
+
},
|
94 |
+
"50263": {
|
95 |
+
"content": " ",
|
96 |
+
"lstrip": false,
|
97 |
+
"normalized": true,
|
98 |
+
"rstrip": false,
|
99 |
+
"single_word": false,
|
100 |
+
"special": false
|
101 |
+
},
|
102 |
+
"50264": {
|
103 |
+
"content": " ",
|
104 |
+
"lstrip": false,
|
105 |
+
"normalized": true,
|
106 |
+
"rstrip": false,
|
107 |
+
"single_word": false,
|
108 |
+
"special": false
|
109 |
+
},
|
110 |
+
"50265": {
|
111 |
+
"content": " ",
|
112 |
+
"lstrip": false,
|
113 |
+
"normalized": true,
|
114 |
+
"rstrip": false,
|
115 |
+
"single_word": false,
|
116 |
+
"special": false
|
117 |
+
},
|
118 |
+
"50266": {
|
119 |
+
"content": " ",
|
120 |
+
"lstrip": false,
|
121 |
+
"normalized": true,
|
122 |
+
"rstrip": false,
|
123 |
+
"single_word": false,
|
124 |
+
"special": false
|
125 |
+
},
|
126 |
+
"50267": {
|
127 |
+
"content": " ",
|
128 |
+
"lstrip": false,
|
129 |
+
"normalized": true,
|
130 |
+
"rstrip": false,
|
131 |
+
"single_word": false,
|
132 |
+
"special": false
|
133 |
+
},
|
134 |
+
"50268": {
|
135 |
+
"content": " ",
|
136 |
+
"lstrip": false,
|
137 |
+
"normalized": true,
|
138 |
+
"rstrip": false,
|
139 |
+
"single_word": false,
|
140 |
+
"special": false
|
141 |
+
},
|
142 |
+
"50269": {
|
143 |
+
"content": " ",
|
144 |
+
"lstrip": false,
|
145 |
+
"normalized": true,
|
146 |
+
"rstrip": false,
|
147 |
+
"single_word": false,
|
148 |
+
"special": false
|
149 |
+
},
|
150 |
+
"50270": {
|
151 |
+
"content": " ",
|
152 |
+
"lstrip": false,
|
153 |
+
"normalized": true,
|
154 |
+
"rstrip": false,
|
155 |
+
"single_word": false,
|
156 |
+
"special": false
|
157 |
+
},
|
158 |
+
"50271": {
|
159 |
+
"content": " ",
|
160 |
+
"lstrip": false,
|
161 |
+
"normalized": true,
|
162 |
+
"rstrip": false,
|
163 |
+
"single_word": false,
|
164 |
+
"special": false
|
165 |
+
},
|
166 |
+
"50272": {
|
167 |
+
"content": " ",
|
168 |
+
"lstrip": false,
|
169 |
+
"normalized": true,
|
170 |
+
"rstrip": false,
|
171 |
+
"single_word": false,
|
172 |
+
"special": false
|
173 |
+
},
|
174 |
+
"50273": {
|
175 |
+
"content": " ",
|
176 |
+
"lstrip": false,
|
177 |
+
"normalized": true,
|
178 |
+
"rstrip": false,
|
179 |
+
"single_word": false,
|
180 |
+
"special": false
|
181 |
+
},
|
182 |
+
"50274": {
|
183 |
+
"content": " ",
|
184 |
+
"lstrip": false,
|
185 |
+
"normalized": true,
|
186 |
+
"rstrip": false,
|
187 |
+
"single_word": false,
|
188 |
+
"special": false
|
189 |
+
},
|
190 |
+
"50275": {
|
191 |
+
"content": " ",
|
192 |
+
"lstrip": false,
|
193 |
+
"normalized": true,
|
194 |
+
"rstrip": false,
|
195 |
+
"single_word": false,
|
196 |
+
"special": false
|
197 |
+
},
|
198 |
+
"50276": {
|
199 |
+
"content": " ",
|
200 |
+
"lstrip": false,
|
201 |
+
"normalized": true,
|
202 |
+
"rstrip": false,
|
203 |
+
"single_word": false,
|
204 |
+
"special": false
|
205 |
+
}
|
206 |
+
},
|
207 |
+
"bos_token": "<|endoftext|>",
|
208 |
+
"clean_up_tokenization_spaces": true,
|
209 |
+
"eos_token": "<|endoftext|>",
|
210 |
+
"extra_special_tokens": {},
|
211 |
+
"model_max_length": 2048,
|
212 |
+
"pad_token": "<|endoftext|>",
|
213 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
214 |
+
"unk_token": "<|endoftext|>"
|
215 |
+
}
|
checkpoint-10000/trainer_state.json
ADDED
@@ -0,0 +1,1434 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_global_step": null,
|
3 |
+
"best_metric": null,
|
4 |
+
"best_model_checkpoint": null,
|
5 |
+
"epoch": 0.5,
|
6 |
+
"eval_steps": 500,
|
7 |
+
"global_step": 10000,
|
8 |
+
"is_hyper_param_search": false,
|
9 |
+
"is_local_process_zero": true,
|
10 |
+
"is_world_process_zero": true,
|
11 |
+
"log_history": [
|
12 |
+
{
|
13 |
+
"epoch": 0.0025,
|
14 |
+
"grad_norm": 1.0703125,
|
15 |
+
"learning_rate": 0.00025,
|
16 |
+
"loss": 10.4938,
|
17 |
+
"step": 50
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 0.005,
|
21 |
+
"grad_norm": 0.90625,
|
22 |
+
"learning_rate": 0.0005,
|
23 |
+
"loss": 9.1324,
|
24 |
+
"step": 100
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.0075,
|
28 |
+
"grad_norm": 0.62890625,
|
29 |
+
"learning_rate": 0.00075,
|
30 |
+
"loss": 7.883,
|
31 |
+
"step": 150
|
32 |
+
},
|
33 |
+
{
|
34 |
+
"epoch": 0.01,
|
35 |
+
"grad_norm": 0.42578125,
|
36 |
+
"learning_rate": 0.001,
|
37 |
+
"loss": 7.3925,
|
38 |
+
"step": 200
|
39 |
+
},
|
40 |
+
{
|
41 |
+
"epoch": 0.0125,
|
42 |
+
"grad_norm": 0.6171875,
|
43 |
+
"learning_rate": 0.0009999842657116666,
|
44 |
+
"loss": 7.062,
|
45 |
+
"step": 250
|
46 |
+
},
|
47 |
+
{
|
48 |
+
"epoch": 0.015,
|
49 |
+
"grad_norm": 0.498046875,
|
50 |
+
"learning_rate": 0.0009999370638369377,
|
51 |
+
"loss": 6.8396,
|
52 |
+
"step": 300
|
53 |
+
},
|
54 |
+
{
|
55 |
+
"epoch": 0.0175,
|
56 |
+
"grad_norm": 0.4296875,
|
57 |
+
"learning_rate": 0.0009998583973465647,
|
58 |
+
"loss": 6.6879,
|
59 |
+
"step": 350
|
60 |
+
},
|
61 |
+
{
|
62 |
+
"epoch": 0.02,
|
63 |
+
"grad_norm": 0.435546875,
|
64 |
+
"learning_rate": 0.0009997482711915926,
|
65 |
+
"loss": 6.6086,
|
66 |
+
"step": 400
|
67 |
+
},
|
68 |
+
{
|
69 |
+
"epoch": 0.0225,
|
70 |
+
"grad_norm": 0.63671875,
|
71 |
+
"learning_rate": 0.0009996066923030483,
|
72 |
+
"loss": 6.5239,
|
73 |
+
"step": 450
|
74 |
+
},
|
75 |
+
{
|
76 |
+
"epoch": 0.025,
|
77 |
+
"grad_norm": 0.37109375,
|
78 |
+
"learning_rate": 0.000999433669591504,
|
79 |
+
"loss": 6.552,
|
80 |
+
"step": 500
|
81 |
+
},
|
82 |
+
{
|
83 |
+
"epoch": 0.0275,
|
84 |
+
"grad_norm": 0.5234375,
|
85 |
+
"learning_rate": 0.0009992292139465165,
|
86 |
+
"loss": 6.4607,
|
87 |
+
"step": 550
|
88 |
+
},
|
89 |
+
{
|
90 |
+
"epoch": 0.03,
|
91 |
+
"grad_norm": 0.61328125,
|
92 |
+
"learning_rate": 0.0009989933382359422,
|
93 |
+
"loss": 6.3974,
|
94 |
+
"step": 600
|
95 |
+
},
|
96 |
+
{
|
97 |
+
"epoch": 0.0325,
|
98 |
+
"grad_norm": 0.67578125,
|
99 |
+
"learning_rate": 0.0009987260573051267,
|
100 |
+
"loss": 6.3249,
|
101 |
+
"step": 650
|
102 |
+
},
|
103 |
+
{
|
104 |
+
"epoch": 0.035,
|
105 |
+
"grad_norm": 0.51953125,
|
106 |
+
"learning_rate": 0.0009984273879759713,
|
107 |
+
"loss": 6.3374,
|
108 |
+
"step": 700
|
109 |
+
},
|
110 |
+
{
|
111 |
+
"epoch": 0.0375,
|
112 |
+
"grad_norm": 0.482421875,
|
113 |
+
"learning_rate": 0.0009980973490458728,
|
114 |
+
"loss": 6.3157,
|
115 |
+
"step": 750
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 0.04,
|
119 |
+
"grad_norm": 0.55859375,
|
120 |
+
"learning_rate": 0.0009977359612865424,
|
121 |
+
"loss": 6.2525,
|
122 |
+
"step": 800
|
123 |
+
},
|
124 |
+
{
|
125 |
+
"epoch": 0.0425,
|
126 |
+
"grad_norm": 0.53125,
|
127 |
+
"learning_rate": 0.0009973432474426967,
|
128 |
+
"loss": 6.2116,
|
129 |
+
"step": 850
|
130 |
+
},
|
131 |
+
{
|
132 |
+
"epoch": 0.045,
|
133 |
+
"grad_norm": 0.439453125,
|
134 |
+
"learning_rate": 0.000996919232230627,
|
135 |
+
"loss": 6.1993,
|
136 |
+
"step": 900
|
137 |
+
},
|
138 |
+
{
|
139 |
+
"epoch": 0.0475,
|
140 |
+
"grad_norm": 0.384765625,
|
141 |
+
"learning_rate": 0.0009964639423366442,
|
142 |
+
"loss": 6.0863,
|
143 |
+
"step": 950
|
144 |
+
},
|
145 |
+
{
|
146 |
+
"epoch": 0.05,
|
147 |
+
"grad_norm": 0.349609375,
|
148 |
+
"learning_rate": 0.0009959774064153978,
|
149 |
+
"loss": 6.049,
|
150 |
+
"step": 1000
|
151 |
+
},
|
152 |
+
{
|
153 |
+
"epoch": 0.0525,
|
154 |
+
"grad_norm": 0.44140625,
|
155 |
+
"learning_rate": 0.0009954596550880734,
|
156 |
+
"loss": 6.0222,
|
157 |
+
"step": 1050
|
158 |
+
},
|
159 |
+
{
|
160 |
+
"epoch": 0.055,
|
161 |
+
"grad_norm": 0.453125,
|
162 |
+
"learning_rate": 0.0009949107209404665,
|
163 |
+
"loss": 6.0078,
|
164 |
+
"step": 1100
|
165 |
+
},
|
166 |
+
{
|
167 |
+
"epoch": 0.0575,
|
168 |
+
"grad_norm": 0.494140625,
|
169 |
+
"learning_rate": 0.000994330638520929,
|
170 |
+
"loss": 6.0114,
|
171 |
+
"step": 1150
|
172 |
+
},
|
173 |
+
{
|
174 |
+
"epoch": 0.06,
|
175 |
+
"grad_norm": 0.46484375,
|
176 |
+
"learning_rate": 0.0009937194443381972,
|
177 |
+
"loss": 5.9764,
|
178 |
+
"step": 1200
|
179 |
+
},
|
180 |
+
{
|
181 |
+
"epoch": 0.0625,
|
182 |
+
"grad_norm": 0.53515625,
|
183 |
+
"learning_rate": 0.0009930771768590933,
|
184 |
+
"loss": 5.9146,
|
185 |
+
"step": 1250
|
186 |
+
},
|
187 |
+
{
|
188 |
+
"epoch": 0.065,
|
189 |
+
"grad_norm": 0.65625,
|
190 |
+
"learning_rate": 0.000992403876506104,
|
191 |
+
"loss": 5.9021,
|
192 |
+
"step": 1300
|
193 |
+
},
|
194 |
+
{
|
195 |
+
"epoch": 0.0675,
|
196 |
+
"grad_norm": 0.396484375,
|
197 |
+
"learning_rate": 0.0009916995856548369,
|
198 |
+
"loss": 5.9053,
|
199 |
+
"step": 1350
|
200 |
+
},
|
201 |
+
{
|
202 |
+
"epoch": 0.07,
|
203 |
+
"grad_norm": 0.49609375,
|
204 |
+
"learning_rate": 0.0009909643486313534,
|
205 |
+
"loss": 5.8727,
|
206 |
+
"step": 1400
|
207 |
+
},
|
208 |
+
{
|
209 |
+
"epoch": 0.0725,
|
210 |
+
"grad_norm": 0.484375,
|
211 |
+
"learning_rate": 0.0009901982117093786,
|
212 |
+
"loss": 5.8429,
|
213 |
+
"step": 1450
|
214 |
+
},
|
215 |
+
{
|
216 |
+
"epoch": 0.075,
|
217 |
+
"grad_norm": 0.6640625,
|
218 |
+
"learning_rate": 0.0009894012231073895,
|
219 |
+
"loss": 5.8162,
|
220 |
+
"step": 1500
|
221 |
+
},
|
222 |
+
{
|
223 |
+
"epoch": 0.0775,
|
224 |
+
"grad_norm": 0.48828125,
|
225 |
+
"learning_rate": 0.0009885734329855799,
|
226 |
+
"loss": 5.8881,
|
227 |
+
"step": 1550
|
228 |
+
},
|
229 |
+
{
|
230 |
+
"epoch": 0.08,
|
231 |
+
"grad_norm": 0.56640625,
|
232 |
+
"learning_rate": 0.0009877148934427035,
|
233 |
+
"loss": 5.7903,
|
234 |
+
"step": 1600
|
235 |
+
},
|
236 |
+
{
|
237 |
+
"epoch": 0.0825,
|
238 |
+
"grad_norm": 0.5390625,
|
239 |
+
"learning_rate": 0.0009868256585127955,
|
240 |
+
"loss": 5.7843,
|
241 |
+
"step": 1650
|
242 |
+
},
|
243 |
+
{
|
244 |
+
"epoch": 0.085,
|
245 |
+
"grad_norm": 0.52734375,
|
246 |
+
"learning_rate": 0.000985905784161771,
|
247 |
+
"loss": 5.7443,
|
248 |
+
"step": 1700
|
249 |
+
},
|
250 |
+
{
|
251 |
+
"epoch": 0.0875,
|
252 |
+
"grad_norm": 0.69921875,
|
253 |
+
"learning_rate": 0.0009849553282839025,
|
254 |
+
"loss": 5.749,
|
255 |
+
"step": 1750
|
256 |
+
},
|
257 |
+
{
|
258 |
+
"epoch": 0.09,
|
259 |
+
"grad_norm": 0.37109375,
|
260 |
+
"learning_rate": 0.0009839743506981783,
|
261 |
+
"loss": 5.8367,
|
262 |
+
"step": 1800
|
263 |
+
},
|
264 |
+
{
|
265 |
+
"epoch": 0.0925,
|
266 |
+
"grad_norm": 0.5703125,
|
267 |
+
"learning_rate": 0.0009829629131445341,
|
268 |
+
"loss": 5.7485,
|
269 |
+
"step": 1850
|
270 |
+
},
|
271 |
+
{
|
272 |
+
"epoch": 0.095,
|
273 |
+
"grad_norm": 0.5625,
|
274 |
+
"learning_rate": 0.000981921079279971,
|
275 |
+
"loss": 5.7068,
|
276 |
+
"step": 1900
|
277 |
+
},
|
278 |
+
{
|
279 |
+
"epoch": 0.0975,
|
280 |
+
"grad_norm": 0.474609375,
|
281 |
+
"learning_rate": 0.0009808489146745465,
|
282 |
+
"loss": 5.7167,
|
283 |
+
"step": 1950
|
284 |
+
},
|
285 |
+
{
|
286 |
+
"epoch": 0.1,
|
287 |
+
"grad_norm": 0.462890625,
|
288 |
+
"learning_rate": 0.0009797464868072487,
|
289 |
+
"loss": 5.7298,
|
290 |
+
"step": 2000
|
291 |
+
},
|
292 |
+
{
|
293 |
+
"epoch": 0.1025,
|
294 |
+
"grad_norm": 0.53515625,
|
295 |
+
"learning_rate": 0.0009786138650617494,
|
296 |
+
"loss": 5.7332,
|
297 |
+
"step": 2050
|
298 |
+
},
|
299 |
+
{
|
300 |
+
"epoch": 0.105,
|
301 |
+
"grad_norm": 0.50390625,
|
302 |
+
"learning_rate": 0.0009774511207220368,
|
303 |
+
"loss": 5.6218,
|
304 |
+
"step": 2100
|
305 |
+
},
|
306 |
+
{
|
307 |
+
"epoch": 0.1075,
|
308 |
+
"grad_norm": 0.55859375,
|
309 |
+
"learning_rate": 0.0009762583269679303,
|
310 |
+
"loss": 5.662,
|
311 |
+
"step": 2150
|
312 |
+
},
|
313 |
+
{
|
314 |
+
"epoch": 0.11,
|
315 |
+
"grad_norm": 0.435546875,
|
316 |
+
"learning_rate": 0.0009750355588704727,
|
317 |
+
"loss": 5.6434,
|
318 |
+
"step": 2200
|
319 |
+
},
|
320 |
+
{
|
321 |
+
"epoch": 0.1125,
|
322 |
+
"grad_norm": 0.49609375,
|
323 |
+
"learning_rate": 0.0009737828933872075,
|
324 |
+
"loss": 5.585,
|
325 |
+
"step": 2250
|
326 |
+
},
|
327 |
+
{
|
328 |
+
"epoch": 0.115,
|
329 |
+
"grad_norm": 0.46484375,
|
330 |
+
"learning_rate": 0.0009725004093573342,
|
331 |
+
"loss": 5.537,
|
332 |
+
"step": 2300
|
333 |
+
},
|
334 |
+
{
|
335 |
+
"epoch": 0.1175,
|
336 |
+
"grad_norm": 0.51953125,
|
337 |
+
"learning_rate": 0.000971188187496747,
|
338 |
+
"loss": 5.605,
|
339 |
+
"step": 2350
|
340 |
+
},
|
341 |
+
{
|
342 |
+
"epoch": 0.12,
|
343 |
+
"grad_norm": 0.431640625,
|
344 |
+
"learning_rate": 0.0009698463103929542,
|
345 |
+
"loss": 5.5706,
|
346 |
+
"step": 2400
|
347 |
+
},
|
348 |
+
{
|
349 |
+
"epoch": 0.1225,
|
350 |
+
"grad_norm": 0.58984375,
|
351 |
+
"learning_rate": 0.000968474862499881,
|
352 |
+
"loss": 5.5877,
|
353 |
+
"step": 2450
|
354 |
+
},
|
355 |
+
{
|
356 |
+
"epoch": 0.125,
|
357 |
+
"grad_norm": 0.60546875,
|
358 |
+
"learning_rate": 0.0009670739301325534,
|
359 |
+
"loss": 5.61,
|
360 |
+
"step": 2500
|
361 |
+
},
|
362 |
+
{
|
363 |
+
"epoch": 0.1275,
|
364 |
+
"grad_norm": 0.70703125,
|
365 |
+
"learning_rate": 0.000965643601461667,
|
366 |
+
"loss": 5.6129,
|
367 |
+
"step": 2550
|
368 |
+
},
|
369 |
+
{
|
370 |
+
"epoch": 0.13,
|
371 |
+
"grad_norm": 0.4375,
|
372 |
+
"learning_rate": 0.0009641839665080363,
|
373 |
+
"loss": 5.5661,
|
374 |
+
"step": 2600
|
375 |
+
},
|
376 |
+
{
|
377 |
+
"epoch": 0.1325,
|
378 |
+
"grad_norm": 0.396484375,
|
379 |
+
"learning_rate": 0.0009626951171369304,
|
380 |
+
"loss": 5.6271,
|
381 |
+
"step": 2650
|
382 |
+
},
|
383 |
+
{
|
384 |
+
"epoch": 0.135,
|
385 |
+
"grad_norm": 0.51171875,
|
386 |
+
"learning_rate": 0.0009611771470522907,
|
387 |
+
"loss": 5.5214,
|
388 |
+
"step": 2700
|
389 |
+
},
|
390 |
+
{
|
391 |
+
"epoch": 0.1375,
|
392 |
+
"grad_norm": 0.5234375,
|
393 |
+
"learning_rate": 0.0009596301517908328,
|
394 |
+
"loss": 5.625,
|
395 |
+
"step": 2750
|
396 |
+
},
|
397 |
+
{
|
398 |
+
"epoch": 0.14,
|
399 |
+
"grad_norm": 0.453125,
|
400 |
+
"learning_rate": 0.0009580542287160348,
|
401 |
+
"loss": 5.5264,
|
402 |
+
"step": 2800
|
403 |
+
},
|
404 |
+
{
|
405 |
+
"epoch": 0.1425,
|
406 |
+
"grad_norm": 0.453125,
|
407 |
+
"learning_rate": 0.0009564494770120089,
|
408 |
+
"loss": 5.5348,
|
409 |
+
"step": 2850
|
410 |
+
},
|
411 |
+
{
|
412 |
+
"epoch": 0.145,
|
413 |
+
"grad_norm": 0.462890625,
|
414 |
+
"learning_rate": 0.0009548159976772592,
|
415 |
+
"loss": 5.4634,
|
416 |
+
"step": 2900
|
417 |
+
},
|
418 |
+
{
|
419 |
+
"epoch": 0.1475,
|
420 |
+
"grad_norm": 0.48828125,
|
421 |
+
"learning_rate": 0.0009531538935183251,
|
422 |
+
"loss": 5.4421,
|
423 |
+
"step": 2950
|
424 |
+
},
|
425 |
+
{
|
426 |
+
"epoch": 0.15,
|
427 |
+
"grad_norm": 0.439453125,
|
428 |
+
"learning_rate": 0.0009514632691433108,
|
429 |
+
"loss": 5.47,
|
430 |
+
"step": 3000
|
431 |
+
},
|
432 |
+
{
|
433 |
+
"epoch": 0.1525,
|
434 |
+
"grad_norm": 0.84375,
|
435 |
+
"learning_rate": 0.0009497442309553016,
|
436 |
+
"loss": 5.4528,
|
437 |
+
"step": 3050
|
438 |
+
},
|
439 |
+
{
|
440 |
+
"epoch": 0.155,
|
441 |
+
"grad_norm": 0.61328125,
|
442 |
+
"learning_rate": 0.0009479968871456679,
|
443 |
+
"loss": 5.3831,
|
444 |
+
"step": 3100
|
445 |
+
},
|
446 |
+
{
|
447 |
+
"epoch": 0.1575,
|
448 |
+
"grad_norm": 0.439453125,
|
449 |
+
"learning_rate": 0.000946221347687255,
|
450 |
+
"loss": 5.3801,
|
451 |
+
"step": 3150
|
452 |
+
},
|
453 |
+
{
|
454 |
+
"epoch": 0.16,
|
455 |
+
"grad_norm": 0.474609375,
|
456 |
+
"learning_rate": 0.0009444177243274617,
|
457 |
+
"loss": 5.394,
|
458 |
+
"step": 3200
|
459 |
+
},
|
460 |
+
{
|
461 |
+
"epoch": 0.1625,
|
462 |
+
"grad_norm": 0.90234375,
|
463 |
+
"learning_rate": 0.0009425861305812082,
|
464 |
+
"loss": 5.4232,
|
465 |
+
"step": 3250
|
466 |
+
},
|
467 |
+
{
|
468 |
+
"epoch": 0.165,
|
469 |
+
"grad_norm": 0.5,
|
470 |
+
"learning_rate": 0.000940726681723791,
|
471 |
+
"loss": 5.4551,
|
472 |
+
"step": 3300
|
473 |
+
},
|
474 |
+
{
|
475 |
+
"epoch": 0.1675,
|
476 |
+
"grad_norm": 0.5625,
|
477 |
+
"learning_rate": 0.0009388394947836278,
|
478 |
+
"loss": 5.4354,
|
479 |
+
"step": 3350
|
480 |
+
},
|
481 |
+
{
|
482 |
+
"epoch": 0.17,
|
483 |
+
"grad_norm": 0.455078125,
|
484 |
+
"learning_rate": 0.0009369246885348925,
|
485 |
+
"loss": 5.4606,
|
486 |
+
"step": 3400
|
487 |
+
},
|
488 |
+
{
|
489 |
+
"epoch": 0.1725,
|
490 |
+
"grad_norm": 0.474609375,
|
491 |
+
"learning_rate": 0.0009349823834900395,
|
492 |
+
"loss": 5.3785,
|
493 |
+
"step": 3450
|
494 |
+
},
|
495 |
+
{
|
496 |
+
"epoch": 0.175,
|
497 |
+
"grad_norm": 0.6328125,
|
498 |
+
"learning_rate": 0.0009330127018922195,
|
499 |
+
"loss": 5.3472,
|
500 |
+
"step": 3500
|
501 |
+
},
|
502 |
+
{
|
503 |
+
"epoch": 0.1775,
|
504 |
+
"grad_norm": 0.40625,
|
505 |
+
"learning_rate": 0.0009310157677075847,
|
506 |
+
"loss": 5.4156,
|
507 |
+
"step": 3550
|
508 |
+
},
|
509 |
+
{
|
510 |
+
"epoch": 0.18,
|
511 |
+
"grad_norm": 0.423828125,
|
512 |
+
"learning_rate": 0.0009289917066174886,
|
513 |
+
"loss": 5.2857,
|
514 |
+
"step": 3600
|
515 |
+
},
|
516 |
+
{
|
517 |
+
"epoch": 0.1825,
|
518 |
+
"grad_norm": 0.470703125,
|
519 |
+
"learning_rate": 0.000926940646010574,
|
520 |
+
"loss": 5.3815,
|
521 |
+
"step": 3650
|
522 |
+
},
|
523 |
+
{
|
524 |
+
"epoch": 0.185,
|
525 |
+
"grad_norm": 0.46484375,
|
526 |
+
"learning_rate": 0.0009248627149747573,
|
527 |
+
"loss": 5.3596,
|
528 |
+
"step": 3700
|
529 |
+
},
|
530 |
+
{
|
531 |
+
"epoch": 0.1875,
|
532 |
+
"grad_norm": 0.455078125,
|
533 |
+
"learning_rate": 0.0009227580442891022,
|
534 |
+
"loss": 5.3197,
|
535 |
+
"step": 3750
|
536 |
+
},
|
537 |
+
{
|
538 |
+
"epoch": 0.19,
|
539 |
+
"grad_norm": 0.49609375,
|
540 |
+
"learning_rate": 0.0009206267664155906,
|
541 |
+
"loss": 5.2859,
|
542 |
+
"step": 3800
|
543 |
+
},
|
544 |
+
{
|
545 |
+
"epoch": 0.1925,
|
546 |
+
"grad_norm": 0.66015625,
|
547 |
+
"learning_rate": 0.0009184690154907849,
|
548 |
+
"loss": 5.3134,
|
549 |
+
"step": 3850
|
550 |
+
},
|
551 |
+
{
|
552 |
+
"epoch": 0.195,
|
553 |
+
"grad_norm": 0.625,
|
554 |
+
"learning_rate": 0.0009162849273173857,
|
555 |
+
"loss": 5.218,
|
556 |
+
"step": 3900
|
557 |
+
},
|
558 |
+
{
|
559 |
+
"epoch": 0.1975,
|
560 |
+
"grad_norm": 0.9296875,
|
561 |
+
"learning_rate": 0.0009140746393556853,
|
562 |
+
"loss": 5.3191,
|
563 |
+
"step": 3950
|
564 |
+
},
|
565 |
+
{
|
566 |
+
"epoch": 0.2,
|
567 |
+
"grad_norm": 0.416015625,
|
568 |
+
"learning_rate": 0.0009118382907149164,
|
569 |
+
"loss": 5.2908,
|
570 |
+
"step": 4000
|
571 |
+
},
|
572 |
+
{
|
573 |
+
"epoch": 0.2025,
|
574 |
+
"grad_norm": 0.498046875,
|
575 |
+
"learning_rate": 0.0009095760221444959,
|
576 |
+
"loss": 5.1909,
|
577 |
+
"step": 4050
|
578 |
+
},
|
579 |
+
{
|
580 |
+
"epoch": 0.205,
|
581 |
+
"grad_norm": 1.0234375,
|
582 |
+
"learning_rate": 0.0009072879760251679,
|
583 |
+
"loss": 5.2562,
|
584 |
+
"step": 4100
|
585 |
+
},
|
586 |
+
{
|
587 |
+
"epoch": 0.2075,
|
588 |
+
"grad_norm": 0.5859375,
|
589 |
+
"learning_rate": 0.0009049742963600418,
|
590 |
+
"loss": 5.2542,
|
591 |
+
"step": 4150
|
592 |
+
},
|
593 |
+
{
|
594 |
+
"epoch": 0.21,
|
595 |
+
"grad_norm": 0.43359375,
|
596 |
+
"learning_rate": 0.0009026351287655293,
|
597 |
+
"loss": 5.1161,
|
598 |
+
"step": 4200
|
599 |
+
},
|
600 |
+
{
|
601 |
+
"epoch": 0.2125,
|
602 |
+
"grad_norm": 0.51953125,
|
603 |
+
"learning_rate": 0.0009002706204621802,
|
604 |
+
"loss": 5.1941,
|
605 |
+
"step": 4250
|
606 |
+
},
|
607 |
+
{
|
608 |
+
"epoch": 0.215,
|
609 |
+
"grad_norm": 0.6015625,
|
610 |
+
"learning_rate": 0.0008978809202654162,
|
611 |
+
"loss": 5.2478,
|
612 |
+
"step": 4300
|
613 |
+
},
|
614 |
+
{
|
615 |
+
"epoch": 0.2175,
|
616 |
+
"grad_norm": 0.46875,
|
617 |
+
"learning_rate": 0.0008954661785761646,
|
618 |
+
"loss": 5.2331,
|
619 |
+
"step": 4350
|
620 |
+
},
|
621 |
+
{
|
622 |
+
"epoch": 0.22,
|
623 |
+
"grad_norm": 0.5078125,
|
624 |
+
"learning_rate": 0.0008930265473713938,
|
625 |
+
"loss": 5.1296,
|
626 |
+
"step": 4400
|
627 |
+
},
|
628 |
+
{
|
629 |
+
"epoch": 0.2225,
|
630 |
+
"grad_norm": 0.4765625,
|
631 |
+
"learning_rate": 0.0008905621801945467,
|
632 |
+
"loss": 5.194,
|
633 |
+
"step": 4450
|
634 |
+
},
|
635 |
+
{
|
636 |
+
"epoch": 0.225,
|
637 |
+
"grad_norm": 0.48828125,
|
638 |
+
"learning_rate": 0.0008880732321458784,
|
639 |
+
"loss": 5.1855,
|
640 |
+
"step": 4500
|
641 |
+
},
|
642 |
+
{
|
643 |
+
"epoch": 0.2275,
|
644 |
+
"grad_norm": 0.42578125,
|
645 |
+
"learning_rate": 0.0008855598598726938,
|
646 |
+
"loss": 5.1613,
|
647 |
+
"step": 4550
|
648 |
+
},
|
649 |
+
{
|
650 |
+
"epoch": 0.23,
|
651 |
+
"grad_norm": 0.578125,
|
652 |
+
"learning_rate": 0.000883022221559489,
|
653 |
+
"loss": 5.1545,
|
654 |
+
"step": 4600
|
655 |
+
},
|
656 |
+
{
|
657 |
+
"epoch": 0.2325,
|
658 |
+
"grad_norm": 0.41015625,
|
659 |
+
"learning_rate": 0.0008804604769179958,
|
660 |
+
"loss": 5.1522,
|
661 |
+
"step": 4650
|
662 |
+
},
|
663 |
+
{
|
664 |
+
"epoch": 0.235,
|
665 |
+
"grad_norm": 0.44140625,
|
666 |
+
"learning_rate": 0.0008778747871771292,
|
667 |
+
"loss": 5.1246,
|
668 |
+
"step": 4700
|
669 |
+
},
|
670 |
+
{
|
671 |
+
"epoch": 0.2375,
|
672 |
+
"grad_norm": 0.625,
|
673 |
+
"learning_rate": 0.0008752653150728412,
|
674 |
+
"loss": 5.1371,
|
675 |
+
"step": 4750
|
676 |
+
},
|
677 |
+
{
|
678 |
+
"epoch": 0.24,
|
679 |
+
"grad_norm": 0.50390625,
|
680 |
+
"learning_rate": 0.0008726322248378774,
|
681 |
+
"loss": 5.0847,
|
682 |
+
"step": 4800
|
683 |
+
},
|
684 |
+
{
|
685 |
+
"epoch": 0.2425,
|
686 |
+
"grad_norm": 0.44140625,
|
687 |
+
"learning_rate": 0.0008699756821914419,
|
688 |
+
"loss": 5.0916,
|
689 |
+
"step": 4850
|
690 |
+
},
|
691 |
+
{
|
692 |
+
"epoch": 0.245,
|
693 |
+
"grad_norm": 0.59765625,
|
694 |
+
"learning_rate": 0.0008672958543287666,
|
695 |
+
"loss": 5.0772,
|
696 |
+
"step": 4900
|
697 |
+
},
|
698 |
+
{
|
699 |
+
"epoch": 0.2475,
|
700 |
+
"grad_norm": 0.51953125,
|
701 |
+
"learning_rate": 0.0008645929099105886,
|
702 |
+
"loss": 5.1363,
|
703 |
+
"step": 4950
|
704 |
+
},
|
705 |
+
{
|
706 |
+
"epoch": 0.25,
|
707 |
+
"grad_norm": 0.451171875,
|
708 |
+
"learning_rate": 0.000861867019052535,
|
709 |
+
"loss": 5.075,
|
710 |
+
"step": 5000
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"epoch": 0.2525,
|
714 |
+
"grad_norm": 0.51171875,
|
715 |
+
"learning_rate": 0.0008591183533144171,
|
716 |
+
"loss": 5.1377,
|
717 |
+
"step": 5050
|
718 |
+
},
|
719 |
+
{
|
720 |
+
"epoch": 0.255,
|
721 |
+
"grad_norm": 0.470703125,
|
722 |
+
"learning_rate": 0.0008563470856894315,
|
723 |
+
"loss": 5.0829,
|
724 |
+
"step": 5100
|
725 |
+
},
|
726 |
+
{
|
727 |
+
"epoch": 0.2575,
|
728 |
+
"grad_norm": 0.59375,
|
729 |
+
"learning_rate": 0.0008535533905932737,
|
730 |
+
"loss": 5.0808,
|
731 |
+
"step": 5150
|
732 |
+
},
|
733 |
+
{
|
734 |
+
"epoch": 0.26,
|
735 |
+
"grad_norm": 0.49609375,
|
736 |
+
"learning_rate": 0.0008507374438531607,
|
737 |
+
"loss": 5.0014,
|
738 |
+
"step": 5200
|
739 |
+
},
|
740 |
+
{
|
741 |
+
"epoch": 0.2625,
|
742 |
+
"grad_norm": 0.5859375,
|
743 |
+
"learning_rate": 0.0008478994226967638,
|
744 |
+
"loss": 5.0243,
|
745 |
+
"step": 5250
|
746 |
+
},
|
747 |
+
{
|
748 |
+
"epoch": 0.265,
|
749 |
+
"grad_norm": 0.4140625,
|
750 |
+
"learning_rate": 0.000845039505741056,
|
751 |
+
"loss": 5.128,
|
752 |
+
"step": 5300
|
753 |
+
},
|
754 |
+
{
|
755 |
+
"epoch": 0.2675,
|
756 |
+
"grad_norm": 0.453125,
|
757 |
+
"learning_rate": 0.0008421578729810691,
|
758 |
+
"loss": 5.0646,
|
759 |
+
"step": 5350
|
760 |
+
},
|
761 |
+
{
|
762 |
+
"epoch": 0.27,
|
763 |
+
"grad_norm": 0.50390625,
|
764 |
+
"learning_rate": 0.0008392547057785661,
|
765 |
+
"loss": 5.0641,
|
766 |
+
"step": 5400
|
767 |
+
},
|
768 |
+
{
|
769 |
+
"epoch": 0.2725,
|
770 |
+
"grad_norm": 0.48046875,
|
771 |
+
"learning_rate": 0.0008363301868506264,
|
772 |
+
"loss": 5.0466,
|
773 |
+
"step": 5450
|
774 |
+
},
|
775 |
+
{
|
776 |
+
"epoch": 0.275,
|
777 |
+
"grad_norm": 0.51171875,
|
778 |
+
"learning_rate": 0.0008333845002581458,
|
779 |
+
"loss": 5.0462,
|
780 |
+
"step": 5500
|
781 |
+
},
|
782 |
+
{
|
783 |
+
"epoch": 0.2775,
|
784 |
+
"grad_norm": 0.5390625,
|
785 |
+
"learning_rate": 0.0008304178313942535,
|
786 |
+
"loss": 5.1451,
|
787 |
+
"step": 5550
|
788 |
+
},
|
789 |
+
{
|
790 |
+
"epoch": 0.28,
|
791 |
+
"grad_norm": 0.5859375,
|
792 |
+
"learning_rate": 0.0008274303669726426,
|
793 |
+
"loss": 5.1001,
|
794 |
+
"step": 5600
|
795 |
+
},
|
796 |
+
{
|
797 |
+
"epoch": 0.2825,
|
798 |
+
"grad_norm": 0.4921875,
|
799 |
+
"learning_rate": 0.0008244222950158193,
|
800 |
+
"loss": 4.9816,
|
801 |
+
"step": 5650
|
802 |
+
},
|
803 |
+
{
|
804 |
+
"epoch": 0.285,
|
805 |
+
"grad_norm": 0.466796875,
|
806 |
+
"learning_rate": 0.0008213938048432696,
|
807 |
+
"loss": 5.0636,
|
808 |
+
"step": 5700
|
809 |
+
},
|
810 |
+
{
|
811 |
+
"epoch": 0.2875,
|
812 |
+
"grad_norm": 0.60546875,
|
813 |
+
"learning_rate": 0.0008183450870595441,
|
814 |
+
"loss": 4.977,
|
815 |
+
"step": 5750
|
816 |
+
},
|
817 |
+
{
|
818 |
+
"epoch": 0.29,
|
819 |
+
"grad_norm": 0.52734375,
|
820 |
+
"learning_rate": 0.0008152763335422613,
|
821 |
+
"loss": 4.9954,
|
822 |
+
"step": 5800
|
823 |
+
},
|
824 |
+
{
|
825 |
+
"epoch": 0.2925,
|
826 |
+
"grad_norm": 0.5546875,
|
827 |
+
"learning_rate": 0.0008121877374300317,
|
828 |
+
"loss": 4.9429,
|
829 |
+
"step": 5850
|
830 |
+
},
|
831 |
+
{
|
832 |
+
"epoch": 0.295,
|
833 |
+
"grad_norm": 0.51171875,
|
834 |
+
"learning_rate": 0.0008090794931103026,
|
835 |
+
"loss": 4.9819,
|
836 |
+
"step": 5900
|
837 |
+
},
|
838 |
+
{
|
839 |
+
"epoch": 0.2975,
|
840 |
+
"grad_norm": 0.57421875,
|
841 |
+
"learning_rate": 0.0008059517962071233,
|
842 |
+
"loss": 5.0432,
|
843 |
+
"step": 5950
|
844 |
+
},
|
845 |
+
{
|
846 |
+
"epoch": 0.3,
|
847 |
+
"grad_norm": 0.50390625,
|
848 |
+
"learning_rate": 0.0008028048435688333,
|
849 |
+
"loss": 4.8642,
|
850 |
+
"step": 6000
|
851 |
+
},
|
852 |
+
{
|
853 |
+
"epoch": 0.3025,
|
854 |
+
"grad_norm": 0.7421875,
|
855 |
+
"learning_rate": 0.0007996388332556734,
|
856 |
+
"loss": 5.0357,
|
857 |
+
"step": 6050
|
858 |
+
},
|
859 |
+
{
|
860 |
+
"epoch": 0.305,
|
861 |
+
"grad_norm": 0.46484375,
|
862 |
+
"learning_rate": 0.0007964539645273203,
|
863 |
+
"loss": 4.9587,
|
864 |
+
"step": 6100
|
865 |
+
},
|
866 |
+
{
|
867 |
+
"epoch": 0.3075,
|
868 |
+
"grad_norm": 0.470703125,
|
869 |
+
"learning_rate": 0.0007932504378303451,
|
870 |
+
"loss": 4.9245,
|
871 |
+
"step": 6150
|
872 |
+
},
|
873 |
+
{
|
874 |
+
"epoch": 0.31,
|
875 |
+
"grad_norm": 0.625,
|
876 |
+
"learning_rate": 0.0007900284547855992,
|
877 |
+
"loss": 4.9813,
|
878 |
+
"step": 6200
|
879 |
+
},
|
880 |
+
{
|
881 |
+
"epoch": 0.3125,
|
882 |
+
"grad_norm": 0.51171875,
|
883 |
+
"learning_rate": 0.0007867882181755231,
|
884 |
+
"loss": 5.0188,
|
885 |
+
"step": 6250
|
886 |
+
},
|
887 |
+
{
|
888 |
+
"epoch": 0.315,
|
889 |
+
"grad_norm": 0.6953125,
|
890 |
+
"learning_rate": 0.0007835299319313853,
|
891 |
+
"loss": 4.9338,
|
892 |
+
"step": 6300
|
893 |
+
},
|
894 |
+
{
|
895 |
+
"epoch": 0.3175,
|
896 |
+
"grad_norm": 0.59375,
|
897 |
+
"learning_rate": 0.000780253801120447,
|
898 |
+
"loss": 4.8996,
|
899 |
+
"step": 6350
|
900 |
+
},
|
901 |
+
{
|
902 |
+
"epoch": 0.32,
|
903 |
+
"grad_norm": 0.671875,
|
904 |
+
"learning_rate": 0.0007769600319330552,
|
905 |
+
"loss": 4.9161,
|
906 |
+
"step": 6400
|
907 |
+
},
|
908 |
+
{
|
909 |
+
"epoch": 0.3225,
|
910 |
+
"grad_norm": 0.5078125,
|
911 |
+
"learning_rate": 0.0007736488316696662,
|
912 |
+
"loss": 4.9648,
|
913 |
+
"step": 6450
|
914 |
+
},
|
915 |
+
{
|
916 |
+
"epoch": 0.325,
|
917 |
+
"grad_norm": 0.625,
|
918 |
+
"learning_rate": 0.0007703204087277988,
|
919 |
+
"loss": 4.908,
|
920 |
+
"step": 6500
|
921 |
+
},
|
922 |
+
{
|
923 |
+
"epoch": 0.3275,
|
924 |
+
"grad_norm": 0.50390625,
|
925 |
+
"learning_rate": 0.0007669749725889182,
|
926 |
+
"loss": 4.9536,
|
927 |
+
"step": 6550
|
928 |
+
},
|
929 |
+
{
|
930 |
+
"epoch": 0.33,
|
931 |
+
"grad_norm": 0.87890625,
|
932 |
+
"learning_rate": 0.0007636127338052513,
|
933 |
+
"loss": 4.8755,
|
934 |
+
"step": 6600
|
935 |
+
},
|
936 |
+
{
|
937 |
+
"epoch": 0.3325,
|
938 |
+
"grad_norm": 0.5546875,
|
939 |
+
"learning_rate": 0.0007602339039865362,
|
940 |
+
"loss": 4.8819,
|
941 |
+
"step": 6650
|
942 |
+
},
|
943 |
+
{
|
944 |
+
"epoch": 0.335,
|
945 |
+
"grad_norm": 0.53125,
|
946 |
+
"learning_rate": 0.0007568386957867032,
|
947 |
+
"loss": 4.8511,
|
948 |
+
"step": 6700
|
949 |
+
},
|
950 |
+
{
|
951 |
+
"epoch": 0.3375,
|
952 |
+
"grad_norm": 0.515625,
|
953 |
+
"learning_rate": 0.0007534273228904916,
|
954 |
+
"loss": 4.9324,
|
955 |
+
"step": 6750
|
956 |
+
},
|
957 |
+
{
|
958 |
+
"epoch": 0.34,
|
959 |
+
"grad_norm": 0.58203125,
|
960 |
+
"learning_rate": 0.00075,
|
961 |
+
"loss": 4.8898,
|
962 |
+
"step": 6800
|
963 |
+
},
|
964 |
+
{
|
965 |
+
"epoch": 0.3425,
|
966 |
+
"grad_norm": 0.63671875,
|
967 |
+
"learning_rate": 0.0007465569428211752,
|
968 |
+
"loss": 4.9447,
|
969 |
+
"step": 6850
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 0.345,
|
973 |
+
"grad_norm": 0.58203125,
|
974 |
+
"learning_rate": 0.0007430983680502344,
|
975 |
+
"loss": 4.8954,
|
976 |
+
"step": 6900
|
977 |
+
},
|
978 |
+
{
|
979 |
+
"epoch": 0.3475,
|
980 |
+
"grad_norm": 0.435546875,
|
981 |
+
"learning_rate": 0.0007396244933600284,
|
982 |
+
"loss": 4.8977,
|
983 |
+
"step": 6950
|
984 |
+
},
|
985 |
+
{
|
986 |
+
"epoch": 0.35,
|
987 |
+
"grad_norm": 0.7734375,
|
988 |
+
"learning_rate": 0.0007361355373863414,
|
989 |
+
"loss": 4.7837,
|
990 |
+
"step": 7000
|
991 |
+
},
|
992 |
+
{
|
993 |
+
"epoch": 0.3525,
|
994 |
+
"grad_norm": 0.458984375,
|
995 |
+
"learning_rate": 0.0007326317197141304,
|
996 |
+
"loss": 4.9241,
|
997 |
+
"step": 7050
|
998 |
+
},
|
999 |
+
{
|
1000 |
+
"epoch": 0.355,
|
1001 |
+
"grad_norm": 0.58203125,
|
1002 |
+
"learning_rate": 0.0007291132608637052,
|
1003 |
+
"loss": 4.9113,
|
1004 |
+
"step": 7100
|
1005 |
+
},
|
1006 |
+
{
|
1007 |
+
"epoch": 0.3575,
|
1008 |
+
"grad_norm": 0.65234375,
|
1009 |
+
"learning_rate": 0.0007255803822768504,
|
1010 |
+
"loss": 4.9187,
|
1011 |
+
"step": 7150
|
1012 |
+
},
|
1013 |
+
{
|
1014 |
+
"epoch": 0.36,
|
1015 |
+
"grad_norm": 0.60546875,
|
1016 |
+
"learning_rate": 0.0007220333063028871,
|
1017 |
+
"loss": 4.8632,
|
1018 |
+
"step": 7200
|
1019 |
+
},
|
1020 |
+
{
|
1021 |
+
"epoch": 0.3625,
|
1022 |
+
"grad_norm": 0.58203125,
|
1023 |
+
"learning_rate": 0.0007184722561846798,
|
1024 |
+
"loss": 5.0078,
|
1025 |
+
"step": 7250
|
1026 |
+
},
|
1027 |
+
{
|
1028 |
+
"epoch": 0.365,
|
1029 |
+
"grad_norm": 0.50390625,
|
1030 |
+
"learning_rate": 0.0007148974560445859,
|
1031 |
+
"loss": 4.7923,
|
1032 |
+
"step": 7300
|
1033 |
+
},
|
1034 |
+
{
|
1035 |
+
"epoch": 0.3675,
|
1036 |
+
"grad_norm": 0.52734375,
|
1037 |
+
"learning_rate": 0.0007113091308703497,
|
1038 |
+
"loss": 4.8948,
|
1039 |
+
"step": 7350
|
1040 |
+
},
|
1041 |
+
{
|
1042 |
+
"epoch": 0.37,
|
1043 |
+
"grad_norm": 0.466796875,
|
1044 |
+
"learning_rate": 0.0007077075065009433,
|
1045 |
+
"loss": 4.8259,
|
1046 |
+
"step": 7400
|
1047 |
+
},
|
1048 |
+
{
|
1049 |
+
"epoch": 0.3725,
|
1050 |
+
"grad_norm": 0.55859375,
|
1051 |
+
"learning_rate": 0.0007040928096123516,
|
1052 |
+
"loss": 4.8009,
|
1053 |
+
"step": 7450
|
1054 |
+
},
|
1055 |
+
{
|
1056 |
+
"epoch": 0.375,
|
1057 |
+
"grad_norm": 0.4375,
|
1058 |
+
"learning_rate": 0.0007004652677033068,
|
1059 |
+
"loss": 4.847,
|
1060 |
+
"step": 7500
|
1061 |
+
},
|
1062 |
+
{
|
1063 |
+
"epoch": 0.3775,
|
1064 |
+
"grad_norm": 0.486328125,
|
1065 |
+
"learning_rate": 0.0006968251090809707,
|
1066 |
+
"loss": 4.835,
|
1067 |
+
"step": 7550
|
1068 |
+
},
|
1069 |
+
{
|
1070 |
+
"epoch": 0.38,
|
1071 |
+
"grad_norm": 0.455078125,
|
1072 |
+
"learning_rate": 0.0006931725628465643,
|
1073 |
+
"loss": 4.8838,
|
1074 |
+
"step": 7600
|
1075 |
+
},
|
1076 |
+
{
|
1077 |
+
"epoch": 0.3825,
|
1078 |
+
"grad_norm": 0.68359375,
|
1079 |
+
"learning_rate": 0.0006895078588809502,
|
1080 |
+
"loss": 4.8705,
|
1081 |
+
"step": 7650
|
1082 |
+
},
|
1083 |
+
{
|
1084 |
+
"epoch": 0.385,
|
1085 |
+
"grad_norm": 0.46875,
|
1086 |
+
"learning_rate": 0.0006858312278301637,
|
1087 |
+
"loss": 4.763,
|
1088 |
+
"step": 7700
|
1089 |
+
},
|
1090 |
+
{
|
1091 |
+
"epoch": 0.3875,
|
1092 |
+
"grad_norm": 0.53125,
|
1093 |
+
"learning_rate": 0.0006821429010908972,
|
1094 |
+
"loss": 4.8961,
|
1095 |
+
"step": 7750
|
1096 |
+
},
|
1097 |
+
{
|
1098 |
+
"epoch": 0.39,
|
1099 |
+
"grad_norm": 0.53125,
|
1100 |
+
"learning_rate": 0.0006784431107959359,
|
1101 |
+
"loss": 4.8141,
|
1102 |
+
"step": 7800
|
1103 |
+
},
|
1104 |
+
{
|
1105 |
+
"epoch": 0.3925,
|
1106 |
+
"grad_norm": 0.5078125,
|
1107 |
+
"learning_rate": 0.0006747320897995492,
|
1108 |
+
"loss": 4.8718,
|
1109 |
+
"step": 7850
|
1110 |
+
},
|
1111 |
+
{
|
1112 |
+
"epoch": 0.395,
|
1113 |
+
"grad_norm": 0.5390625,
|
1114 |
+
"learning_rate": 0.0006710100716628344,
|
1115 |
+
"loss": 4.8394,
|
1116 |
+
"step": 7900
|
1117 |
+
},
|
1118 |
+
{
|
1119 |
+
"epoch": 0.3975,
|
1120 |
+
"grad_norm": 0.65234375,
|
1121 |
+
"learning_rate": 0.0006672772906390176,
|
1122 |
+
"loss": 4.8264,
|
1123 |
+
"step": 7950
|
1124 |
+
},
|
1125 |
+
{
|
1126 |
+
"epoch": 0.4,
|
1127 |
+
"grad_norm": 0.80078125,
|
1128 |
+
"learning_rate": 0.0006635339816587109,
|
1129 |
+
"loss": 4.769,
|
1130 |
+
"step": 8000
|
1131 |
+
},
|
1132 |
+
{
|
1133 |
+
"epoch": 0.4025,
|
1134 |
+
"grad_norm": 0.61328125,
|
1135 |
+
"learning_rate": 0.000659780380315125,
|
1136 |
+
"loss": 4.8428,
|
1137 |
+
"step": 8050
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"epoch": 0.405,
|
1141 |
+
"grad_norm": 0.515625,
|
1142 |
+
"learning_rate": 0.0006560167228492435,
|
1143 |
+
"loss": 4.8591,
|
1144 |
+
"step": 8100
|
1145 |
+
},
|
1146 |
+
{
|
1147 |
+
"epoch": 0.4075,
|
1148 |
+
"grad_norm": 0.80078125,
|
1149 |
+
"learning_rate": 0.0006522432461349536,
|
1150 |
+
"loss": 4.9224,
|
1151 |
+
"step": 8150
|
1152 |
+
},
|
1153 |
+
{
|
1154 |
+
"epoch": 0.41,
|
1155 |
+
"grad_norm": 0.640625,
|
1156 |
+
"learning_rate": 0.0006484601876641375,
|
1157 |
+
"loss": 4.7773,
|
1158 |
+
"step": 8200
|
1159 |
+
},
|
1160 |
+
{
|
1161 |
+
"epoch": 0.4125,
|
1162 |
+
"grad_norm": 0.5703125,
|
1163 |
+
"learning_rate": 0.0006446677855317265,
|
1164 |
+
"loss": 4.8107,
|
1165 |
+
"step": 8250
|
1166 |
+
},
|
1167 |
+
{
|
1168 |
+
"epoch": 0.415,
|
1169 |
+
"grad_norm": 0.53515625,
|
1170 |
+
"learning_rate": 0.0006408662784207149,
|
1171 |
+
"loss": 4.8252,
|
1172 |
+
"step": 8300
|
1173 |
+
},
|
1174 |
+
{
|
1175 |
+
"epoch": 0.4175,
|
1176 |
+
"grad_norm": 0.546875,
|
1177 |
+
"learning_rate": 0.0006370559055871389,
|
1178 |
+
"loss": 4.8562,
|
1179 |
+
"step": 8350
|
1180 |
+
},
|
1181 |
+
{
|
1182 |
+
"epoch": 0.42,
|
1183 |
+
"grad_norm": 0.5625,
|
1184 |
+
"learning_rate": 0.0006332369068450174,
|
1185 |
+
"loss": 4.7781,
|
1186 |
+
"step": 8400
|
1187 |
+
},
|
1188 |
+
{
|
1189 |
+
"epoch": 0.4225,
|
1190 |
+
"grad_norm": 0.462890625,
|
1191 |
+
"learning_rate": 0.0006294095225512603,
|
1192 |
+
"loss": 4.8325,
|
1193 |
+
"step": 8450
|
1194 |
+
},
|
1195 |
+
{
|
1196 |
+
"epoch": 0.425,
|
1197 |
+
"grad_norm": 0.45703125,
|
1198 |
+
"learning_rate": 0.0006255739935905395,
|
1199 |
+
"loss": 4.8363,
|
1200 |
+
"step": 8500
|
1201 |
+
},
|
1202 |
+
{
|
1203 |
+
"epoch": 0.4275,
|
1204 |
+
"grad_norm": 0.6015625,
|
1205 |
+
"learning_rate": 0.0006217305613601295,
|
1206 |
+
"loss": 4.8252,
|
1207 |
+
"step": 8550
|
1208 |
+
},
|
1209 |
+
{
|
1210 |
+
"epoch": 0.43,
|
1211 |
+
"grad_norm": 0.609375,
|
1212 |
+
"learning_rate": 0.0006178794677547138,
|
1213 |
+
"loss": 4.8079,
|
1214 |
+
"step": 8600
|
1215 |
+
},
|
1216 |
+
{
|
1217 |
+
"epoch": 0.4325,
|
1218 |
+
"grad_norm": 0.470703125,
|
1219 |
+
"learning_rate": 0.0006140209551511608,
|
1220 |
+
"loss": 4.8508,
|
1221 |
+
"step": 8650
|
1222 |
+
},
|
1223 |
+
{
|
1224 |
+
"epoch": 0.435,
|
1225 |
+
"grad_norm": 0.53515625,
|
1226 |
+
"learning_rate": 0.0006101552663932703,
|
1227 |
+
"loss": 4.8389,
|
1228 |
+
"step": 8700
|
1229 |
+
},
|
1230 |
+
{
|
1231 |
+
"epoch": 0.4375,
|
1232 |
+
"grad_norm": 0.66796875,
|
1233 |
+
"learning_rate": 0.0006062826447764884,
|
1234 |
+
"loss": 4.7695,
|
1235 |
+
"step": 8750
|
1236 |
+
},
|
1237 |
+
{
|
1238 |
+
"epoch": 0.44,
|
1239 |
+
"grad_norm": 0.455078125,
|
1240 |
+
"learning_rate": 0.0006024033340325954,
|
1241 |
+
"loss": 4.7953,
|
1242 |
+
"step": 8800
|
1243 |
+
},
|
1244 |
+
{
|
1245 |
+
"epoch": 0.4425,
|
1246 |
+
"grad_norm": 0.56640625,
|
1247 |
+
"learning_rate": 0.0005985175783143666,
|
1248 |
+
"loss": 4.8144,
|
1249 |
+
"step": 8850
|
1250 |
+
},
|
1251 |
+
{
|
1252 |
+
"epoch": 0.445,
|
1253 |
+
"grad_norm": 0.48046875,
|
1254 |
+
"learning_rate": 0.0005946256221802051,
|
1255 |
+
"loss": 4.8123,
|
1256 |
+
"step": 8900
|
1257 |
+
},
|
1258 |
+
{
|
1259 |
+
"epoch": 0.4475,
|
1260 |
+
"grad_norm": 0.52734375,
|
1261 |
+
"learning_rate": 0.0005907277105787513,
|
1262 |
+
"loss": 4.7778,
|
1263 |
+
"step": 8950
|
1264 |
+
},
|
1265 |
+
{
|
1266 |
+
"epoch": 0.45,
|
1267 |
+
"grad_norm": 0.53515625,
|
1268 |
+
"learning_rate": 0.0005868240888334653,
|
1269 |
+
"loss": 4.7237,
|
1270 |
+
"step": 9000
|
1271 |
+
},
|
1272 |
+
{
|
1273 |
+
"epoch": 0.4525,
|
1274 |
+
"grad_norm": 0.58984375,
|
1275 |
+
"learning_rate": 0.0005829150026271871,
|
1276 |
+
"loss": 4.7938,
|
1277 |
+
"step": 9050
|
1278 |
+
},
|
1279 |
+
{
|
1280 |
+
"epoch": 0.455,
|
1281 |
+
"grad_norm": 0.56640625,
|
1282 |
+
"learning_rate": 0.000579000697986675,
|
1283 |
+
"loss": 4.7879,
|
1284 |
+
"step": 9100
|
1285 |
+
},
|
1286 |
+
{
|
1287 |
+
"epoch": 0.4575,
|
1288 |
+
"grad_norm": 0.578125,
|
1289 |
+
"learning_rate": 0.0005750814212671201,
|
1290 |
+
"loss": 4.7317,
|
1291 |
+
"step": 9150
|
1292 |
+
},
|
1293 |
+
{
|
1294 |
+
"epoch": 0.46,
|
1295 |
+
"grad_norm": 0.578125,
|
1296 |
+
"learning_rate": 0.0005711574191366427,
|
1297 |
+
"loss": 4.7819,
|
1298 |
+
"step": 9200
|
1299 |
+
},
|
1300 |
+
{
|
1301 |
+
"epoch": 0.4625,
|
1302 |
+
"grad_norm": 0.61328125,
|
1303 |
+
"learning_rate": 0.0005672289385607659,
|
1304 |
+
"loss": 4.8393,
|
1305 |
+
"step": 9250
|
1306 |
+
},
|
1307 |
+
{
|
1308 |
+
"epoch": 0.465,
|
1309 |
+
"grad_norm": 0.69921875,
|
1310 |
+
"learning_rate": 0.0005632962267868747,
|
1311 |
+
"loss": 4.7797,
|
1312 |
+
"step": 9300
|
1313 |
+
},
|
1314 |
+
{
|
1315 |
+
"epoch": 0.4675,
|
1316 |
+
"grad_norm": 0.71484375,
|
1317 |
+
"learning_rate": 0.0005593595313286526,
|
1318 |
+
"loss": 4.764,
|
1319 |
+
"step": 9350
|
1320 |
+
},
|
1321 |
+
{
|
1322 |
+
"epoch": 0.47,
|
1323 |
+
"grad_norm": 0.4765625,
|
1324 |
+
"learning_rate": 0.0005554190999505056,
|
1325 |
+
"loss": 4.7639,
|
1326 |
+
"step": 9400
|
1327 |
+
},
|
1328 |
+
{
|
1329 |
+
"epoch": 0.4725,
|
1330 |
+
"grad_norm": 0.53125,
|
1331 |
+
"learning_rate": 0.0005514751806519673,
|
1332 |
+
"loss": 4.8052,
|
1333 |
+
"step": 9450
|
1334 |
+
},
|
1335 |
+
{
|
1336 |
+
"epoch": 0.475,
|
1337 |
+
"grad_norm": 0.6328125,
|
1338 |
+
"learning_rate": 0.0005475280216520913,
|
1339 |
+
"loss": 4.8682,
|
1340 |
+
"step": 9500
|
1341 |
+
},
|
1342 |
+
{
|
1343 |
+
"epoch": 0.4775,
|
1344 |
+
"grad_norm": 0.640625,
|
1345 |
+
"learning_rate": 0.0005435778713738292,
|
1346 |
+
"loss": 4.748,
|
1347 |
+
"step": 9550
|
1348 |
+
},
|
1349 |
+
{
|
1350 |
+
"epoch": 0.48,
|
1351 |
+
"grad_norm": 0.66015625,
|
1352 |
+
"learning_rate": 0.0005396249784283942,
|
1353 |
+
"loss": 4.7625,
|
1354 |
+
"step": 9600
|
1355 |
+
},
|
1356 |
+
{
|
1357 |
+
"epoch": 0.4825,
|
1358 |
+
"grad_norm": 0.51171875,
|
1359 |
+
"learning_rate": 0.0005356695915996161,
|
1360 |
+
"loss": 4.824,
|
1361 |
+
"step": 9650
|
1362 |
+
},
|
1363 |
+
{
|
1364 |
+
"epoch": 0.485,
|
1365 |
+
"grad_norm": 0.54296875,
|
1366 |
+
"learning_rate": 0.0005317119598282822,
|
1367 |
+
"loss": 4.7946,
|
1368 |
+
"step": 9700
|
1369 |
+
},
|
1370 |
+
{
|
1371 |
+
"epoch": 0.4875,
|
1372 |
+
"grad_norm": 0.671875,
|
1373 |
+
"learning_rate": 0.0005277523321964701,
|
1374 |
+
"loss": 4.8476,
|
1375 |
+
"step": 9750
|
1376 |
+
},
|
1377 |
+
{
|
1378 |
+
"epoch": 0.49,
|
1379 |
+
"grad_norm": 0.55078125,
|
1380 |
+
"learning_rate": 0.0005237909579118712,
|
1381 |
+
"loss": 4.735,
|
1382 |
+
"step": 9800
|
1383 |
+
},
|
1384 |
+
{
|
1385 |
+
"epoch": 0.4925,
|
1386 |
+
"grad_norm": 0.47265625,
|
1387 |
+
"learning_rate": 0.0005198280862921062,
|
1388 |
+
"loss": 4.8119,
|
1389 |
+
"step": 9850
|
1390 |
+
},
|
1391 |
+
{
|
1392 |
+
"epoch": 0.495,
|
1393 |
+
"grad_norm": 0.49609375,
|
1394 |
+
"learning_rate": 0.0005158639667490339,
|
1395 |
+
"loss": 4.794,
|
1396 |
+
"step": 9900
|
1397 |
+
},
|
1398 |
+
{
|
1399 |
+
"epoch": 0.4975,
|
1400 |
+
"grad_norm": 1.0546875,
|
1401 |
+
"learning_rate": 0.0005118988487730537,
|
1402 |
+
"loss": 4.7226,
|
1403 |
+
"step": 9950
|
1404 |
+
},
|
1405 |
+
{
|
1406 |
+
"epoch": 0.5,
|
1407 |
+
"grad_norm": 0.56640625,
|
1408 |
+
"learning_rate": 0.000507932981917404,
|
1409 |
+
"loss": 4.9084,
|
1410 |
+
"step": 10000
|
1411 |
+
}
|
1412 |
+
],
|
1413 |
+
"logging_steps": 50,
|
1414 |
+
"max_steps": 20000,
|
1415 |
+
"num_input_tokens_seen": 0,
|
1416 |
+
"num_train_epochs": 9223372036854775807,
|
1417 |
+
"save_steps": 100,
|
1418 |
+
"stateful_callbacks": {
|
1419 |
+
"TrainerControl": {
|
1420 |
+
"args": {
|
1421 |
+
"should_epoch_stop": false,
|
1422 |
+
"should_evaluate": false,
|
1423 |
+
"should_log": false,
|
1424 |
+
"should_save": true,
|
1425 |
+
"should_training_stop": false
|
1426 |
+
},
|
1427 |
+
"attributes": {}
|
1428 |
+
}
|
1429 |
+
},
|
1430 |
+
"total_flos": 1.45593184813056e+16,
|
1431 |
+
"train_batch_size": 16,
|
1432 |
+
"trial_name": null,
|
1433 |
+
"trial_params": null
|
1434 |
+
}
|
checkpoint-10000/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:83031100336f2dd4b43e2539aae80864690fb3fc70607039ce5c50a2ff4a21bb
|
3 |
+
size 5368
|
checkpoint-10100/config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"GPTNeoXForCausalLM"
|
4 |
+
],
|
5 |
+
"attention_bias": true,
|
6 |
+
"attention_dropout": 0.0,
|
7 |
+
"bos_token_id": 0,
|
8 |
+
"classifier_dropout": 0.1,
|
9 |
+
"eos_token_id": 0,
|
10 |
+
"hidden_act": "gelu",
|
11 |
+
"hidden_dropout": 0.0,
|
12 |
+
"hidden_size": 128,
|
13 |
+
"initializer_range": 0.02,
|
14 |
+
"intermediate_size": 512,
|
15 |
+
"layer_norm_eps": 1e-05,
|
16 |
+
"max_position_embeddings": 2048,
|
17 |
+
"model_type": "gpt_neox",
|
18 |
+
"num_attention_heads": 4,
|
19 |
+
"num_hidden_layers": 6,
|
20 |
+
"partial_rotary_factor": 0.25,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000,
|
23 |
+
"rotary_emb_base": 10000,
|
24 |
+
"rotary_pct": 0.25,
|
25 |
+
"tie_word_embeddings": false,
|
26 |
+
"torch_dtype": "bfloat16",
|
27 |
+
"transformers_version": "4.50.3",
|
28 |
+
"use_cache": true,
|
29 |
+
"use_parallel_residual": true,
|
30 |
+
"vocab_size": 50304
|
31 |
+
}
|
checkpoint-10100/generation_config.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
+
"eos_token_id": 0,
|
5 |
+
"transformers_version": "4.50.3"
|
6 |
+
}
|
checkpoint-10100/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:30c0592afc6bf18be87bd500aff0fae7eaf874b7bb1c580d8ba15932873b9603
|
3 |
+
size 28144000
|
checkpoint-10100/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d1082cb3c282658cac5c5fdbca128ee6328ba109a388633c68ed7ee76617ed8a
|
3 |
+
size 56334394
|
checkpoint-10100/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
|
3 |
+
size 14244
|
checkpoint-10100/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dbc46c838934201946a568f5a0c975b25e76c650028bdd46811a9509faf91595
|
3 |
+
size 1064
|
checkpoint-10100/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<|endoftext|>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "<|endoftext|>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<|endoftext|>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<|endoftext|>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-10100/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-10100/tokenizer_config.json
ADDED
@@ -0,0 +1,215 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": false,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"add_prefix_space": false,
|
5 |
+
"added_tokens_decoder": {
|
6 |
+
"0": {
|
7 |
+
"content": "<|endoftext|>",
|
8 |
+
"lstrip": false,
|
9 |
+
"normalized": false,
|
10 |
+
"rstrip": false,
|
11 |
+
"single_word": false,
|
12 |
+
"special": true
|
13 |
+
},
|
14 |
+
"1": {
|
15 |
+
"content": "<|padding|>",
|
16 |
+
"lstrip": false,
|
17 |
+
"normalized": false,
|
18 |
+
"rstrip": false,
|
19 |
+
"single_word": false,
|
20 |
+
"special": true
|
21 |
+
},
|
22 |
+
"50254": {
|
23 |
+
"content": " ",
|
24 |
+
"lstrip": false,
|
25 |
+
"normalized": true,
|
26 |
+
"rstrip": false,
|
27 |
+
"single_word": false,
|
28 |
+
"special": false
|
29 |
+
},
|
30 |
+
"50255": {
|
31 |
+
"content": " ",
|
32 |
+
"lstrip": false,
|
33 |
+
"normalized": true,
|
34 |
+
"rstrip": false,
|
35 |
+
"single_word": false,
|
36 |
+
"special": false
|
37 |
+
},
|
38 |
+
"50256": {
|
39 |
+
"content": " ",
|
40 |
+
"lstrip": false,
|
41 |
+
"normalized": true,
|
42 |
+
"rstrip": false,
|
43 |
+
"single_word": false,
|
44 |
+
"special": false
|
45 |
+
},
|
46 |
+
"50257": {
|
47 |
+
"content": " ",
|
48 |
+
"lstrip": false,
|
49 |
+
"normalized": true,
|
50 |
+
"rstrip": false,
|
51 |
+
"single_word": false,
|
52 |
+
"special": false
|
53 |
+
},
|
54 |
+
"50258": {
|
55 |
+
"content": " ",
|
56 |
+
"lstrip": false,
|
57 |
+
"normalized": true,
|
58 |
+
"rstrip": false,
|
59 |
+
"single_word": false,
|
60 |
+
"special": false
|
61 |
+
},
|
62 |
+
"50259": {
|
63 |
+
"content": " ",
|
64 |
+
"lstrip": false,
|
65 |
+
"normalized": true,
|
66 |
+
"rstrip": false,
|
67 |
+
"single_word": false,
|
68 |
+
"special": false
|
69 |
+
},
|
70 |
+
"50260": {
|
71 |
+
"content": " ",
|
72 |
+
"lstrip": false,
|
73 |
+
"normalized": true,
|
74 |
+
"rstrip": false,
|
75 |
+
"single_word": false,
|
76 |
+
"special": false
|
77 |
+
},
|
78 |
+
"50261": {
|
79 |
+
"content": " ",
|
80 |
+
"lstrip": false,
|
81 |
+
"normalized": true,
|
82 |
+
"rstrip": false,
|
83 |
+
"single_word": false,
|
84 |
+
"special": false
|
85 |
+
},
|
86 |
+
"50262": {
|
87 |
+
"content": " ",
|
88 |
+
"lstrip": false,
|
89 |
+
"normalized": true,
|
90 |
+
"rstrip": false,
|
91 |
+
"single_word": false,
|
92 |
+
"special": false
|
93 |
+
},
|
94 |
+
"50263": {
|
95 |
+
"content": " ",
|
96 |
+
"lstrip": false,
|
97 |
+
"normalized": true,
|
98 |
+
"rstrip": false,
|
99 |
+
"single_word": false,
|
100 |
+
"special": false
|
101 |
+
},
|
102 |
+
"50264": {
|
103 |
+
"content": " ",
|
104 |
+
"lstrip": false,
|
105 |
+
"normalized": true,
|
106 |
+
"rstrip": false,
|
107 |
+
"single_word": false,
|
108 |
+
"special": false
|
109 |
+
},
|
110 |
+
"50265": {
|
111 |
+
"content": " ",
|
112 |
+
"lstrip": false,
|
113 |
+
"normalized": true,
|
114 |
+
"rstrip": false,
|
115 |
+
"single_word": false,
|
116 |
+
"special": false
|
117 |
+
},
|
118 |
+
"50266": {
|
119 |
+
"content": " ",
|
120 |
+
"lstrip": false,
|
121 |
+
"normalized": true,
|
122 |
+
"rstrip": false,
|
123 |
+
"single_word": false,
|
124 |
+
"special": false
|
125 |
+
},
|
126 |
+
"50267": {
|
127 |
+
"content": " ",
|
128 |
+
"lstrip": false,
|
129 |
+
"normalized": true,
|
130 |
+
"rstrip": false,
|
131 |
+
"single_word": false,
|
132 |
+
"special": false
|
133 |
+
},
|
134 |
+
"50268": {
|
135 |
+
"content": " ",
|
136 |
+
"lstrip": false,
|
137 |
+
"normalized": true,
|
138 |
+
"rstrip": false,
|
139 |
+
"single_word": false,
|
140 |
+
"special": false
|
141 |
+
},
|
142 |
+
"50269": {
|
143 |
+
"content": " ",
|
144 |
+
"lstrip": false,
|
145 |
+
"normalized": true,
|
146 |
+
"rstrip": false,
|
147 |
+
"single_word": false,
|
148 |
+
"special": false
|
149 |
+
},
|
150 |
+
"50270": {
|
151 |
+
"content": " ",
|
152 |
+
"lstrip": false,
|
153 |
+
"normalized": true,
|
154 |
+
"rstrip": false,
|
155 |
+
"single_word": false,
|
156 |
+
"special": false
|
157 |
+
},
|
158 |
+
"50271": {
|
159 |
+
"content": " ",
|
160 |
+
"lstrip": false,
|
161 |
+
"normalized": true,
|
162 |
+
"rstrip": false,
|
163 |
+
"single_word": false,
|
164 |
+
"special": false
|
165 |
+
},
|
166 |
+
"50272": {
|
167 |
+
"content": " ",
|
168 |
+
"lstrip": false,
|
169 |
+
"normalized": true,
|
170 |
+
"rstrip": false,
|
171 |
+
"single_word": false,
|
172 |
+
"special": false
|
173 |
+
},
|
174 |
+
"50273": {
|
175 |
+
"content": " ",
|
176 |
+
"lstrip": false,
|
177 |
+
"normalized": true,
|
178 |
+
"rstrip": false,
|
179 |
+
"single_word": false,
|
180 |
+
"special": false
|
181 |
+
},
|
182 |
+
"50274": {
|
183 |
+
"content": " ",
|
184 |
+
"lstrip": false,
|
185 |
+
"normalized": true,
|
186 |
+
"rstrip": false,
|
187 |
+
"single_word": false,
|
188 |
+
"special": false
|
189 |
+
},
|
190 |
+
"50275": {
|
191 |
+
"content": " ",
|
192 |
+
"lstrip": false,
|
193 |
+
"normalized": true,
|
194 |
+
"rstrip": false,
|
195 |
+
"single_word": false,
|
196 |
+
"special": false
|
197 |
+
},
|
198 |
+
"50276": {
|
199 |
+
"content": " ",
|
200 |
+
"lstrip": false,
|
201 |
+
"normalized": true,
|
202 |
+
"rstrip": false,
|
203 |
+
"single_word": false,
|
204 |
+
"special": false
|
205 |
+
}
|
206 |
+
},
|
207 |
+
"bos_token": "<|endoftext|>",
|
208 |
+
"clean_up_tokenization_spaces": true,
|
209 |
+
"eos_token": "<|endoftext|>",
|
210 |
+
"extra_special_tokens": {},
|
211 |
+
"model_max_length": 2048,
|
212 |
+
"pad_token": "<|endoftext|>",
|
213 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
214 |
+
"unk_token": "<|endoftext|>"
|
215 |
+
}
|
checkpoint-10100/trainer_state.json
ADDED
@@ -0,0 +1,1448 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_global_step": null,
|
3 |
+
"best_metric": null,
|
4 |
+
"best_model_checkpoint": null,
|
5 |
+
"epoch": 0.505,
|
6 |
+
"eval_steps": 500,
|
7 |
+
"global_step": 10100,
|
8 |
+
"is_hyper_param_search": false,
|
9 |
+
"is_local_process_zero": true,
|
10 |
+
"is_world_process_zero": true,
|
11 |
+
"log_history": [
|
12 |
+
{
|
13 |
+
"epoch": 0.0025,
|
14 |
+
"grad_norm": 1.0703125,
|
15 |
+
"learning_rate": 0.00025,
|
16 |
+
"loss": 10.4938,
|
17 |
+
"step": 50
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 0.005,
|
21 |
+
"grad_norm": 0.90625,
|
22 |
+
"learning_rate": 0.0005,
|
23 |
+
"loss": 9.1324,
|
24 |
+
"step": 100
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.0075,
|
28 |
+
"grad_norm": 0.62890625,
|
29 |
+
"learning_rate": 0.00075,
|
30 |
+
"loss": 7.883,
|
31 |
+
"step": 150
|
32 |
+
},
|
33 |
+
{
|
34 |
+
"epoch": 0.01,
|
35 |
+
"grad_norm": 0.42578125,
|
36 |
+
"learning_rate": 0.001,
|
37 |
+
"loss": 7.3925,
|
38 |
+
"step": 200
|
39 |
+
},
|
40 |
+
{
|
41 |
+
"epoch": 0.0125,
|
42 |
+
"grad_norm": 0.6171875,
|
43 |
+
"learning_rate": 0.0009999842657116666,
|
44 |
+
"loss": 7.062,
|
45 |
+
"step": 250
|
46 |
+
},
|
47 |
+
{
|
48 |
+
"epoch": 0.015,
|
49 |
+
"grad_norm": 0.498046875,
|
50 |
+
"learning_rate": 0.0009999370638369377,
|
51 |
+
"loss": 6.8396,
|
52 |
+
"step": 300
|
53 |
+
},
|
54 |
+
{
|
55 |
+
"epoch": 0.0175,
|
56 |
+
"grad_norm": 0.4296875,
|
57 |
+
"learning_rate": 0.0009998583973465647,
|
58 |
+
"loss": 6.6879,
|
59 |
+
"step": 350
|
60 |
+
},
|
61 |
+
{
|
62 |
+
"epoch": 0.02,
|
63 |
+
"grad_norm": 0.435546875,
|
64 |
+
"learning_rate": 0.0009997482711915926,
|
65 |
+
"loss": 6.6086,
|
66 |
+
"step": 400
|
67 |
+
},
|
68 |
+
{
|
69 |
+
"epoch": 0.0225,
|
70 |
+
"grad_norm": 0.63671875,
|
71 |
+
"learning_rate": 0.0009996066923030483,
|
72 |
+
"loss": 6.5239,
|
73 |
+
"step": 450
|
74 |
+
},
|
75 |
+
{
|
76 |
+
"epoch": 0.025,
|
77 |
+
"grad_norm": 0.37109375,
|
78 |
+
"learning_rate": 0.000999433669591504,
|
79 |
+
"loss": 6.552,
|
80 |
+
"step": 500
|
81 |
+
},
|
82 |
+
{
|
83 |
+
"epoch": 0.0275,
|
84 |
+
"grad_norm": 0.5234375,
|
85 |
+
"learning_rate": 0.0009992292139465165,
|
86 |
+
"loss": 6.4607,
|
87 |
+
"step": 550
|
88 |
+
},
|
89 |
+
{
|
90 |
+
"epoch": 0.03,
|
91 |
+
"grad_norm": 0.61328125,
|
92 |
+
"learning_rate": 0.0009989933382359422,
|
93 |
+
"loss": 6.3974,
|
94 |
+
"step": 600
|
95 |
+
},
|
96 |
+
{
|
97 |
+
"epoch": 0.0325,
|
98 |
+
"grad_norm": 0.67578125,
|
99 |
+
"learning_rate": 0.0009987260573051267,
|
100 |
+
"loss": 6.3249,
|
101 |
+
"step": 650
|
102 |
+
},
|
103 |
+
{
|
104 |
+
"epoch": 0.035,
|
105 |
+
"grad_norm": 0.51953125,
|
106 |
+
"learning_rate": 0.0009984273879759713,
|
107 |
+
"loss": 6.3374,
|
108 |
+
"step": 700
|
109 |
+
},
|
110 |
+
{
|
111 |
+
"epoch": 0.0375,
|
112 |
+
"grad_norm": 0.482421875,
|
113 |
+
"learning_rate": 0.0009980973490458728,
|
114 |
+
"loss": 6.3157,
|
115 |
+
"step": 750
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 0.04,
|
119 |
+
"grad_norm": 0.55859375,
|
120 |
+
"learning_rate": 0.0009977359612865424,
|
121 |
+
"loss": 6.2525,
|
122 |
+
"step": 800
|
123 |
+
},
|
124 |
+
{
|
125 |
+
"epoch": 0.0425,
|
126 |
+
"grad_norm": 0.53125,
|
127 |
+
"learning_rate": 0.0009973432474426967,
|
128 |
+
"loss": 6.2116,
|
129 |
+
"step": 850
|
130 |
+
},
|
131 |
+
{
|
132 |
+
"epoch": 0.045,
|
133 |
+
"grad_norm": 0.439453125,
|
134 |
+
"learning_rate": 0.000996919232230627,
|
135 |
+
"loss": 6.1993,
|
136 |
+
"step": 900
|
137 |
+
},
|
138 |
+
{
|
139 |
+
"epoch": 0.0475,
|
140 |
+
"grad_norm": 0.384765625,
|
141 |
+
"learning_rate": 0.0009964639423366442,
|
142 |
+
"loss": 6.0863,
|
143 |
+
"step": 950
|
144 |
+
},
|
145 |
+
{
|
146 |
+
"epoch": 0.05,
|
147 |
+
"grad_norm": 0.349609375,
|
148 |
+
"learning_rate": 0.0009959774064153978,
|
149 |
+
"loss": 6.049,
|
150 |
+
"step": 1000
|
151 |
+
},
|
152 |
+
{
|
153 |
+
"epoch": 0.0525,
|
154 |
+
"grad_norm": 0.44140625,
|
155 |
+
"learning_rate": 0.0009954596550880734,
|
156 |
+
"loss": 6.0222,
|
157 |
+
"step": 1050
|
158 |
+
},
|
159 |
+
{
|
160 |
+
"epoch": 0.055,
|
161 |
+
"grad_norm": 0.453125,
|
162 |
+
"learning_rate": 0.0009949107209404665,
|
163 |
+
"loss": 6.0078,
|
164 |
+
"step": 1100
|
165 |
+
},
|
166 |
+
{
|
167 |
+
"epoch": 0.0575,
|
168 |
+
"grad_norm": 0.494140625,
|
169 |
+
"learning_rate": 0.000994330638520929,
|
170 |
+
"loss": 6.0114,
|
171 |
+
"step": 1150
|
172 |
+
},
|
173 |
+
{
|
174 |
+
"epoch": 0.06,
|
175 |
+
"grad_norm": 0.46484375,
|
176 |
+
"learning_rate": 0.0009937194443381972,
|
177 |
+
"loss": 5.9764,
|
178 |
+
"step": 1200
|
179 |
+
},
|
180 |
+
{
|
181 |
+
"epoch": 0.0625,
|
182 |
+
"grad_norm": 0.53515625,
|
183 |
+
"learning_rate": 0.0009930771768590933,
|
184 |
+
"loss": 5.9146,
|
185 |
+
"step": 1250
|
186 |
+
},
|
187 |
+
{
|
188 |
+
"epoch": 0.065,
|
189 |
+
"grad_norm": 0.65625,
|
190 |
+
"learning_rate": 0.000992403876506104,
|
191 |
+
"loss": 5.9021,
|
192 |
+
"step": 1300
|
193 |
+
},
|
194 |
+
{
|
195 |
+
"epoch": 0.0675,
|
196 |
+
"grad_norm": 0.396484375,
|
197 |
+
"learning_rate": 0.0009916995856548369,
|
198 |
+
"loss": 5.9053,
|
199 |
+
"step": 1350
|
200 |
+
},
|
201 |
+
{
|
202 |
+
"epoch": 0.07,
|
203 |
+
"grad_norm": 0.49609375,
|
204 |
+
"learning_rate": 0.0009909643486313534,
|
205 |
+
"loss": 5.8727,
|
206 |
+
"step": 1400
|
207 |
+
},
|
208 |
+
{
|
209 |
+
"epoch": 0.0725,
|
210 |
+
"grad_norm": 0.484375,
|
211 |
+
"learning_rate": 0.0009901982117093786,
|
212 |
+
"loss": 5.8429,
|
213 |
+
"step": 1450
|
214 |
+
},
|
215 |
+
{
|
216 |
+
"epoch": 0.075,
|
217 |
+
"grad_norm": 0.6640625,
|
218 |
+
"learning_rate": 0.0009894012231073895,
|
219 |
+
"loss": 5.8162,
|
220 |
+
"step": 1500
|
221 |
+
},
|
222 |
+
{
|
223 |
+
"epoch": 0.0775,
|
224 |
+
"grad_norm": 0.48828125,
|
225 |
+
"learning_rate": 0.0009885734329855799,
|
226 |
+
"loss": 5.8881,
|
227 |
+
"step": 1550
|
228 |
+
},
|
229 |
+
{
|
230 |
+
"epoch": 0.08,
|
231 |
+
"grad_norm": 0.56640625,
|
232 |
+
"learning_rate": 0.0009877148934427035,
|
233 |
+
"loss": 5.7903,
|
234 |
+
"step": 1600
|
235 |
+
},
|
236 |
+
{
|
237 |
+
"epoch": 0.0825,
|
238 |
+
"grad_norm": 0.5390625,
|
239 |
+
"learning_rate": 0.0009868256585127955,
|
240 |
+
"loss": 5.7843,
|
241 |
+
"step": 1650
|
242 |
+
},
|
243 |
+
{
|
244 |
+
"epoch": 0.085,
|
245 |
+
"grad_norm": 0.52734375,
|
246 |
+
"learning_rate": 0.000985905784161771,
|
247 |
+
"loss": 5.7443,
|
248 |
+
"step": 1700
|
249 |
+
},
|
250 |
+
{
|
251 |
+
"epoch": 0.0875,
|
252 |
+
"grad_norm": 0.69921875,
|
253 |
+
"learning_rate": 0.0009849553282839025,
|
254 |
+
"loss": 5.749,
|
255 |
+
"step": 1750
|
256 |
+
},
|
257 |
+
{
|
258 |
+
"epoch": 0.09,
|
259 |
+
"grad_norm": 0.37109375,
|
260 |
+
"learning_rate": 0.0009839743506981783,
|
261 |
+
"loss": 5.8367,
|
262 |
+
"step": 1800
|
263 |
+
},
|
264 |
+
{
|
265 |
+
"epoch": 0.0925,
|
266 |
+
"grad_norm": 0.5703125,
|
267 |
+
"learning_rate": 0.0009829629131445341,
|
268 |
+
"loss": 5.7485,
|
269 |
+
"step": 1850
|
270 |
+
},
|
271 |
+
{
|
272 |
+
"epoch": 0.095,
|
273 |
+
"grad_norm": 0.5625,
|
274 |
+
"learning_rate": 0.000981921079279971,
|
275 |
+
"loss": 5.7068,
|
276 |
+
"step": 1900
|
277 |
+
},
|
278 |
+
{
|
279 |
+
"epoch": 0.0975,
|
280 |
+
"grad_norm": 0.474609375,
|
281 |
+
"learning_rate": 0.0009808489146745465,
|
282 |
+
"loss": 5.7167,
|
283 |
+
"step": 1950
|
284 |
+
},
|
285 |
+
{
|
286 |
+
"epoch": 0.1,
|
287 |
+
"grad_norm": 0.462890625,
|
288 |
+
"learning_rate": 0.0009797464868072487,
|
289 |
+
"loss": 5.7298,
|
290 |
+
"step": 2000
|
291 |
+
},
|
292 |
+
{
|
293 |
+
"epoch": 0.1025,
|
294 |
+
"grad_norm": 0.53515625,
|
295 |
+
"learning_rate": 0.0009786138650617494,
|
296 |
+
"loss": 5.7332,
|
297 |
+
"step": 2050
|
298 |
+
},
|
299 |
+
{
|
300 |
+
"epoch": 0.105,
|
301 |
+
"grad_norm": 0.50390625,
|
302 |
+
"learning_rate": 0.0009774511207220368,
|
303 |
+
"loss": 5.6218,
|
304 |
+
"step": 2100
|
305 |
+
},
|
306 |
+
{
|
307 |
+
"epoch": 0.1075,
|
308 |
+
"grad_norm": 0.55859375,
|
309 |
+
"learning_rate": 0.0009762583269679303,
|
310 |
+
"loss": 5.662,
|
311 |
+
"step": 2150
|
312 |
+
},
|
313 |
+
{
|
314 |
+
"epoch": 0.11,
|
315 |
+
"grad_norm": 0.435546875,
|
316 |
+
"learning_rate": 0.0009750355588704727,
|
317 |
+
"loss": 5.6434,
|
318 |
+
"step": 2200
|
319 |
+
},
|
320 |
+
{
|
321 |
+
"epoch": 0.1125,
|
322 |
+
"grad_norm": 0.49609375,
|
323 |
+
"learning_rate": 0.0009737828933872075,
|
324 |
+
"loss": 5.585,
|
325 |
+
"step": 2250
|
326 |
+
},
|
327 |
+
{
|
328 |
+
"epoch": 0.115,
|
329 |
+
"grad_norm": 0.46484375,
|
330 |
+
"learning_rate": 0.0009725004093573342,
|
331 |
+
"loss": 5.537,
|
332 |
+
"step": 2300
|
333 |
+
},
|
334 |
+
{
|
335 |
+
"epoch": 0.1175,
|
336 |
+
"grad_norm": 0.51953125,
|
337 |
+
"learning_rate": 0.000971188187496747,
|
338 |
+
"loss": 5.605,
|
339 |
+
"step": 2350
|
340 |
+
},
|
341 |
+
{
|
342 |
+
"epoch": 0.12,
|
343 |
+
"grad_norm": 0.431640625,
|
344 |
+
"learning_rate": 0.0009698463103929542,
|
345 |
+
"loss": 5.5706,
|
346 |
+
"step": 2400
|
347 |
+
},
|
348 |
+
{
|
349 |
+
"epoch": 0.1225,
|
350 |
+
"grad_norm": 0.58984375,
|
351 |
+
"learning_rate": 0.000968474862499881,
|
352 |
+
"loss": 5.5877,
|
353 |
+
"step": 2450
|
354 |
+
},
|
355 |
+
{
|
356 |
+
"epoch": 0.125,
|
357 |
+
"grad_norm": 0.60546875,
|
358 |
+
"learning_rate": 0.0009670739301325534,
|
359 |
+
"loss": 5.61,
|
360 |
+
"step": 2500
|
361 |
+
},
|
362 |
+
{
|
363 |
+
"epoch": 0.1275,
|
364 |
+
"grad_norm": 0.70703125,
|
365 |
+
"learning_rate": 0.000965643601461667,
|
366 |
+
"loss": 5.6129,
|
367 |
+
"step": 2550
|
368 |
+
},
|
369 |
+
{
|
370 |
+
"epoch": 0.13,
|
371 |
+
"grad_norm": 0.4375,
|
372 |
+
"learning_rate": 0.0009641839665080363,
|
373 |
+
"loss": 5.5661,
|
374 |
+
"step": 2600
|
375 |
+
},
|
376 |
+
{
|
377 |
+
"epoch": 0.1325,
|
378 |
+
"grad_norm": 0.396484375,
|
379 |
+
"learning_rate": 0.0009626951171369304,
|
380 |
+
"loss": 5.6271,
|
381 |
+
"step": 2650
|
382 |
+
},
|
383 |
+
{
|
384 |
+
"epoch": 0.135,
|
385 |
+
"grad_norm": 0.51171875,
|
386 |
+
"learning_rate": 0.0009611771470522907,
|
387 |
+
"loss": 5.5214,
|
388 |
+
"step": 2700
|
389 |
+
},
|
390 |
+
{
|
391 |
+
"epoch": 0.1375,
|
392 |
+
"grad_norm": 0.5234375,
|
393 |
+
"learning_rate": 0.0009596301517908328,
|
394 |
+
"loss": 5.625,
|
395 |
+
"step": 2750
|
396 |
+
},
|
397 |
+
{
|
398 |
+
"epoch": 0.14,
|
399 |
+
"grad_norm": 0.453125,
|
400 |
+
"learning_rate": 0.0009580542287160348,
|
401 |
+
"loss": 5.5264,
|
402 |
+
"step": 2800
|
403 |
+
},
|
404 |
+
{
|
405 |
+
"epoch": 0.1425,
|
406 |
+
"grad_norm": 0.453125,
|
407 |
+
"learning_rate": 0.0009564494770120089,
|
408 |
+
"loss": 5.5348,
|
409 |
+
"step": 2850
|
410 |
+
},
|
411 |
+
{
|
412 |
+
"epoch": 0.145,
|
413 |
+
"grad_norm": 0.462890625,
|
414 |
+
"learning_rate": 0.0009548159976772592,
|
415 |
+
"loss": 5.4634,
|
416 |
+
"step": 2900
|
417 |
+
},
|
418 |
+
{
|
419 |
+
"epoch": 0.1475,
|
420 |
+
"grad_norm": 0.48828125,
|
421 |
+
"learning_rate": 0.0009531538935183251,
|
422 |
+
"loss": 5.4421,
|
423 |
+
"step": 2950
|
424 |
+
},
|
425 |
+
{
|
426 |
+
"epoch": 0.15,
|
427 |
+
"grad_norm": 0.439453125,
|
428 |
+
"learning_rate": 0.0009514632691433108,
|
429 |
+
"loss": 5.47,
|
430 |
+
"step": 3000
|
431 |
+
},
|
432 |
+
{
|
433 |
+
"epoch": 0.1525,
|
434 |
+
"grad_norm": 0.84375,
|
435 |
+
"learning_rate": 0.0009497442309553016,
|
436 |
+
"loss": 5.4528,
|
437 |
+
"step": 3050
|
438 |
+
},
|
439 |
+
{
|
440 |
+
"epoch": 0.155,
|
441 |
+
"grad_norm": 0.61328125,
|
442 |
+
"learning_rate": 0.0009479968871456679,
|
443 |
+
"loss": 5.3831,
|
444 |
+
"step": 3100
|
445 |
+
},
|
446 |
+
{
|
447 |
+
"epoch": 0.1575,
|
448 |
+
"grad_norm": 0.439453125,
|
449 |
+
"learning_rate": 0.000946221347687255,
|
450 |
+
"loss": 5.3801,
|
451 |
+
"step": 3150
|
452 |
+
},
|
453 |
+
{
|
454 |
+
"epoch": 0.16,
|
455 |
+
"grad_norm": 0.474609375,
|
456 |
+
"learning_rate": 0.0009444177243274617,
|
457 |
+
"loss": 5.394,
|
458 |
+
"step": 3200
|
459 |
+
},
|
460 |
+
{
|
461 |
+
"epoch": 0.1625,
|
462 |
+
"grad_norm": 0.90234375,
|
463 |
+
"learning_rate": 0.0009425861305812082,
|
464 |
+
"loss": 5.4232,
|
465 |
+
"step": 3250
|
466 |
+
},
|
467 |
+
{
|
468 |
+
"epoch": 0.165,
|
469 |
+
"grad_norm": 0.5,
|
470 |
+
"learning_rate": 0.000940726681723791,
|
471 |
+
"loss": 5.4551,
|
472 |
+
"step": 3300
|
473 |
+
},
|
474 |
+
{
|
475 |
+
"epoch": 0.1675,
|
476 |
+
"grad_norm": 0.5625,
|
477 |
+
"learning_rate": 0.0009388394947836278,
|
478 |
+
"loss": 5.4354,
|
479 |
+
"step": 3350
|
480 |
+
},
|
481 |
+
{
|
482 |
+
"epoch": 0.17,
|
483 |
+
"grad_norm": 0.455078125,
|
484 |
+
"learning_rate": 0.0009369246885348925,
|
485 |
+
"loss": 5.4606,
|
486 |
+
"step": 3400
|
487 |
+
},
|
488 |
+
{
|
489 |
+
"epoch": 0.1725,
|
490 |
+
"grad_norm": 0.474609375,
|
491 |
+
"learning_rate": 0.0009349823834900395,
|
492 |
+
"loss": 5.3785,
|
493 |
+
"step": 3450
|
494 |
+
},
|
495 |
+
{
|
496 |
+
"epoch": 0.175,
|
497 |
+
"grad_norm": 0.6328125,
|
498 |
+
"learning_rate": 0.0009330127018922195,
|
499 |
+
"loss": 5.3472,
|
500 |
+
"step": 3500
|
501 |
+
},
|
502 |
+
{
|
503 |
+
"epoch": 0.1775,
|
504 |
+
"grad_norm": 0.40625,
|
505 |
+
"learning_rate": 0.0009310157677075847,
|
506 |
+
"loss": 5.4156,
|
507 |
+
"step": 3550
|
508 |
+
},
|
509 |
+
{
|
510 |
+
"epoch": 0.18,
|
511 |
+
"grad_norm": 0.423828125,
|
512 |
+
"learning_rate": 0.0009289917066174886,
|
513 |
+
"loss": 5.2857,
|
514 |
+
"step": 3600
|
515 |
+
},
|
516 |
+
{
|
517 |
+
"epoch": 0.1825,
|
518 |
+
"grad_norm": 0.470703125,
|
519 |
+
"learning_rate": 0.000926940646010574,
|
520 |
+
"loss": 5.3815,
|
521 |
+
"step": 3650
|
522 |
+
},
|
523 |
+
{
|
524 |
+
"epoch": 0.185,
|
525 |
+
"grad_norm": 0.46484375,
|
526 |
+
"learning_rate": 0.0009248627149747573,
|
527 |
+
"loss": 5.3596,
|
528 |
+
"step": 3700
|
529 |
+
},
|
530 |
+
{
|
531 |
+
"epoch": 0.1875,
|
532 |
+
"grad_norm": 0.455078125,
|
533 |
+
"learning_rate": 0.0009227580442891022,
|
534 |
+
"loss": 5.3197,
|
535 |
+
"step": 3750
|
536 |
+
},
|
537 |
+
{
|
538 |
+
"epoch": 0.19,
|
539 |
+
"grad_norm": 0.49609375,
|
540 |
+
"learning_rate": 0.0009206267664155906,
|
541 |
+
"loss": 5.2859,
|
542 |
+
"step": 3800
|
543 |
+
},
|
544 |
+
{
|
545 |
+
"epoch": 0.1925,
|
546 |
+
"grad_norm": 0.66015625,
|
547 |
+
"learning_rate": 0.0009184690154907849,
|
548 |
+
"loss": 5.3134,
|
549 |
+
"step": 3850
|
550 |
+
},
|
551 |
+
{
|
552 |
+
"epoch": 0.195,
|
553 |
+
"grad_norm": 0.625,
|
554 |
+
"learning_rate": 0.0009162849273173857,
|
555 |
+
"loss": 5.218,
|
556 |
+
"step": 3900
|
557 |
+
},
|
558 |
+
{
|
559 |
+
"epoch": 0.1975,
|
560 |
+
"grad_norm": 0.9296875,
|
561 |
+
"learning_rate": 0.0009140746393556853,
|
562 |
+
"loss": 5.3191,
|
563 |
+
"step": 3950
|
564 |
+
},
|
565 |
+
{
|
566 |
+
"epoch": 0.2,
|
567 |
+
"grad_norm": 0.416015625,
|
568 |
+
"learning_rate": 0.0009118382907149164,
|
569 |
+
"loss": 5.2908,
|
570 |
+
"step": 4000
|
571 |
+
},
|
572 |
+
{
|
573 |
+
"epoch": 0.2025,
|
574 |
+
"grad_norm": 0.498046875,
|
575 |
+
"learning_rate": 0.0009095760221444959,
|
576 |
+
"loss": 5.1909,
|
577 |
+
"step": 4050
|
578 |
+
},
|
579 |
+
{
|
580 |
+
"epoch": 0.205,
|
581 |
+
"grad_norm": 1.0234375,
|
582 |
+
"learning_rate": 0.0009072879760251679,
|
583 |
+
"loss": 5.2562,
|
584 |
+
"step": 4100
|
585 |
+
},
|
586 |
+
{
|
587 |
+
"epoch": 0.2075,
|
588 |
+
"grad_norm": 0.5859375,
|
589 |
+
"learning_rate": 0.0009049742963600418,
|
590 |
+
"loss": 5.2542,
|
591 |
+
"step": 4150
|
592 |
+
},
|
593 |
+
{
|
594 |
+
"epoch": 0.21,
|
595 |
+
"grad_norm": 0.43359375,
|
596 |
+
"learning_rate": 0.0009026351287655293,
|
597 |
+
"loss": 5.1161,
|
598 |
+
"step": 4200
|
599 |
+
},
|
600 |
+
{
|
601 |
+
"epoch": 0.2125,
|
602 |
+
"grad_norm": 0.51953125,
|
603 |
+
"learning_rate": 0.0009002706204621802,
|
604 |
+
"loss": 5.1941,
|
605 |
+
"step": 4250
|
606 |
+
},
|
607 |
+
{
|
608 |
+
"epoch": 0.215,
|
609 |
+
"grad_norm": 0.6015625,
|
610 |
+
"learning_rate": 0.0008978809202654162,
|
611 |
+
"loss": 5.2478,
|
612 |
+
"step": 4300
|
613 |
+
},
|
614 |
+
{
|
615 |
+
"epoch": 0.2175,
|
616 |
+
"grad_norm": 0.46875,
|
617 |
+
"learning_rate": 0.0008954661785761646,
|
618 |
+
"loss": 5.2331,
|
619 |
+
"step": 4350
|
620 |
+
},
|
621 |
+
{
|
622 |
+
"epoch": 0.22,
|
623 |
+
"grad_norm": 0.5078125,
|
624 |
+
"learning_rate": 0.0008930265473713938,
|
625 |
+
"loss": 5.1296,
|
626 |
+
"step": 4400
|
627 |
+
},
|
628 |
+
{
|
629 |
+
"epoch": 0.2225,
|
630 |
+
"grad_norm": 0.4765625,
|
631 |
+
"learning_rate": 0.0008905621801945467,
|
632 |
+
"loss": 5.194,
|
633 |
+
"step": 4450
|
634 |
+
},
|
635 |
+
{
|
636 |
+
"epoch": 0.225,
|
637 |
+
"grad_norm": 0.48828125,
|
638 |
+
"learning_rate": 0.0008880732321458784,
|
639 |
+
"loss": 5.1855,
|
640 |
+
"step": 4500
|
641 |
+
},
|
642 |
+
{
|
643 |
+
"epoch": 0.2275,
|
644 |
+
"grad_norm": 0.42578125,
|
645 |
+
"learning_rate": 0.0008855598598726938,
|
646 |
+
"loss": 5.1613,
|
647 |
+
"step": 4550
|
648 |
+
},
|
649 |
+
{
|
650 |
+
"epoch": 0.23,
|
651 |
+
"grad_norm": 0.578125,
|
652 |
+
"learning_rate": 0.000883022221559489,
|
653 |
+
"loss": 5.1545,
|
654 |
+
"step": 4600
|
655 |
+
},
|
656 |
+
{
|
657 |
+
"epoch": 0.2325,
|
658 |
+
"grad_norm": 0.41015625,
|
659 |
+
"learning_rate": 0.0008804604769179958,
|
660 |
+
"loss": 5.1522,
|
661 |
+
"step": 4650
|
662 |
+
},
|
663 |
+
{
|
664 |
+
"epoch": 0.235,
|
665 |
+
"grad_norm": 0.44140625,
|
666 |
+
"learning_rate": 0.0008778747871771292,
|
667 |
+
"loss": 5.1246,
|
668 |
+
"step": 4700
|
669 |
+
},
|
670 |
+
{
|
671 |
+
"epoch": 0.2375,
|
672 |
+
"grad_norm": 0.625,
|
673 |
+
"learning_rate": 0.0008752653150728412,
|
674 |
+
"loss": 5.1371,
|
675 |
+
"step": 4750
|
676 |
+
},
|
677 |
+
{
|
678 |
+
"epoch": 0.24,
|
679 |
+
"grad_norm": 0.50390625,
|
680 |
+
"learning_rate": 0.0008726322248378774,
|
681 |
+
"loss": 5.0847,
|
682 |
+
"step": 4800
|
683 |
+
},
|
684 |
+
{
|
685 |
+
"epoch": 0.2425,
|
686 |
+
"grad_norm": 0.44140625,
|
687 |
+
"learning_rate": 0.0008699756821914419,
|
688 |
+
"loss": 5.0916,
|
689 |
+
"step": 4850
|
690 |
+
},
|
691 |
+
{
|
692 |
+
"epoch": 0.245,
|
693 |
+
"grad_norm": 0.59765625,
|
694 |
+
"learning_rate": 0.0008672958543287666,
|
695 |
+
"loss": 5.0772,
|
696 |
+
"step": 4900
|
697 |
+
},
|
698 |
+
{
|
699 |
+
"epoch": 0.2475,
|
700 |
+
"grad_norm": 0.51953125,
|
701 |
+
"learning_rate": 0.0008645929099105886,
|
702 |
+
"loss": 5.1363,
|
703 |
+
"step": 4950
|
704 |
+
},
|
705 |
+
{
|
706 |
+
"epoch": 0.25,
|
707 |
+
"grad_norm": 0.451171875,
|
708 |
+
"learning_rate": 0.000861867019052535,
|
709 |
+
"loss": 5.075,
|
710 |
+
"step": 5000
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"epoch": 0.2525,
|
714 |
+
"grad_norm": 0.51171875,
|
715 |
+
"learning_rate": 0.0008591183533144171,
|
716 |
+
"loss": 5.1377,
|
717 |
+
"step": 5050
|
718 |
+
},
|
719 |
+
{
|
720 |
+
"epoch": 0.255,
|
721 |
+
"grad_norm": 0.470703125,
|
722 |
+
"learning_rate": 0.0008563470856894315,
|
723 |
+
"loss": 5.0829,
|
724 |
+
"step": 5100
|
725 |
+
},
|
726 |
+
{
|
727 |
+
"epoch": 0.2575,
|
728 |
+
"grad_norm": 0.59375,
|
729 |
+
"learning_rate": 0.0008535533905932737,
|
730 |
+
"loss": 5.0808,
|
731 |
+
"step": 5150
|
732 |
+
},
|
733 |
+
{
|
734 |
+
"epoch": 0.26,
|
735 |
+
"grad_norm": 0.49609375,
|
736 |
+
"learning_rate": 0.0008507374438531607,
|
737 |
+
"loss": 5.0014,
|
738 |
+
"step": 5200
|
739 |
+
},
|
740 |
+
{
|
741 |
+
"epoch": 0.2625,
|
742 |
+
"grad_norm": 0.5859375,
|
743 |
+
"learning_rate": 0.0008478994226967638,
|
744 |
+
"loss": 5.0243,
|
745 |
+
"step": 5250
|
746 |
+
},
|
747 |
+
{
|
748 |
+
"epoch": 0.265,
|
749 |
+
"grad_norm": 0.4140625,
|
750 |
+
"learning_rate": 0.000845039505741056,
|
751 |
+
"loss": 5.128,
|
752 |
+
"step": 5300
|
753 |
+
},
|
754 |
+
{
|
755 |
+
"epoch": 0.2675,
|
756 |
+
"grad_norm": 0.453125,
|
757 |
+
"learning_rate": 0.0008421578729810691,
|
758 |
+
"loss": 5.0646,
|
759 |
+
"step": 5350
|
760 |
+
},
|
761 |
+
{
|
762 |
+
"epoch": 0.27,
|
763 |
+
"grad_norm": 0.50390625,
|
764 |
+
"learning_rate": 0.0008392547057785661,
|
765 |
+
"loss": 5.0641,
|
766 |
+
"step": 5400
|
767 |
+
},
|
768 |
+
{
|
769 |
+
"epoch": 0.2725,
|
770 |
+
"grad_norm": 0.48046875,
|
771 |
+
"learning_rate": 0.0008363301868506264,
|
772 |
+
"loss": 5.0466,
|
773 |
+
"step": 5450
|
774 |
+
},
|
775 |
+
{
|
776 |
+
"epoch": 0.275,
|
777 |
+
"grad_norm": 0.51171875,
|
778 |
+
"learning_rate": 0.0008333845002581458,
|
779 |
+
"loss": 5.0462,
|
780 |
+
"step": 5500
|
781 |
+
},
|
782 |
+
{
|
783 |
+
"epoch": 0.2775,
|
784 |
+
"grad_norm": 0.5390625,
|
785 |
+
"learning_rate": 0.0008304178313942535,
|
786 |
+
"loss": 5.1451,
|
787 |
+
"step": 5550
|
788 |
+
},
|
789 |
+
{
|
790 |
+
"epoch": 0.28,
|
791 |
+
"grad_norm": 0.5859375,
|
792 |
+
"learning_rate": 0.0008274303669726426,
|
793 |
+
"loss": 5.1001,
|
794 |
+
"step": 5600
|
795 |
+
},
|
796 |
+
{
|
797 |
+
"epoch": 0.2825,
|
798 |
+
"grad_norm": 0.4921875,
|
799 |
+
"learning_rate": 0.0008244222950158193,
|
800 |
+
"loss": 4.9816,
|
801 |
+
"step": 5650
|
802 |
+
},
|
803 |
+
{
|
804 |
+
"epoch": 0.285,
|
805 |
+
"grad_norm": 0.466796875,
|
806 |
+
"learning_rate": 0.0008213938048432696,
|
807 |
+
"loss": 5.0636,
|
808 |
+
"step": 5700
|
809 |
+
},
|
810 |
+
{
|
811 |
+
"epoch": 0.2875,
|
812 |
+
"grad_norm": 0.60546875,
|
813 |
+
"learning_rate": 0.0008183450870595441,
|
814 |
+
"loss": 4.977,
|
815 |
+
"step": 5750
|
816 |
+
},
|
817 |
+
{
|
818 |
+
"epoch": 0.29,
|
819 |
+
"grad_norm": 0.52734375,
|
820 |
+
"learning_rate": 0.0008152763335422613,
|
821 |
+
"loss": 4.9954,
|
822 |
+
"step": 5800
|
823 |
+
},
|
824 |
+
{
|
825 |
+
"epoch": 0.2925,
|
826 |
+
"grad_norm": 0.5546875,
|
827 |
+
"learning_rate": 0.0008121877374300317,
|
828 |
+
"loss": 4.9429,
|
829 |
+
"step": 5850
|
830 |
+
},
|
831 |
+
{
|
832 |
+
"epoch": 0.295,
|
833 |
+
"grad_norm": 0.51171875,
|
834 |
+
"learning_rate": 0.0008090794931103026,
|
835 |
+
"loss": 4.9819,
|
836 |
+
"step": 5900
|
837 |
+
},
|
838 |
+
{
|
839 |
+
"epoch": 0.2975,
|
840 |
+
"grad_norm": 0.57421875,
|
841 |
+
"learning_rate": 0.0008059517962071233,
|
842 |
+
"loss": 5.0432,
|
843 |
+
"step": 5950
|
844 |
+
},
|
845 |
+
{
|
846 |
+
"epoch": 0.3,
|
847 |
+
"grad_norm": 0.50390625,
|
848 |
+
"learning_rate": 0.0008028048435688333,
|
849 |
+
"loss": 4.8642,
|
850 |
+
"step": 6000
|
851 |
+
},
|
852 |
+
{
|
853 |
+
"epoch": 0.3025,
|
854 |
+
"grad_norm": 0.7421875,
|
855 |
+
"learning_rate": 0.0007996388332556734,
|
856 |
+
"loss": 5.0357,
|
857 |
+
"step": 6050
|
858 |
+
},
|
859 |
+
{
|
860 |
+
"epoch": 0.305,
|
861 |
+
"grad_norm": 0.46484375,
|
862 |
+
"learning_rate": 0.0007964539645273203,
|
863 |
+
"loss": 4.9587,
|
864 |
+
"step": 6100
|
865 |
+
},
|
866 |
+
{
|
867 |
+
"epoch": 0.3075,
|
868 |
+
"grad_norm": 0.470703125,
|
869 |
+
"learning_rate": 0.0007932504378303451,
|
870 |
+
"loss": 4.9245,
|
871 |
+
"step": 6150
|
872 |
+
},
|
873 |
+
{
|
874 |
+
"epoch": 0.31,
|
875 |
+
"grad_norm": 0.625,
|
876 |
+
"learning_rate": 0.0007900284547855992,
|
877 |
+
"loss": 4.9813,
|
878 |
+
"step": 6200
|
879 |
+
},
|
880 |
+
{
|
881 |
+
"epoch": 0.3125,
|
882 |
+
"grad_norm": 0.51171875,
|
883 |
+
"learning_rate": 0.0007867882181755231,
|
884 |
+
"loss": 5.0188,
|
885 |
+
"step": 6250
|
886 |
+
},
|
887 |
+
{
|
888 |
+
"epoch": 0.315,
|
889 |
+
"grad_norm": 0.6953125,
|
890 |
+
"learning_rate": 0.0007835299319313853,
|
891 |
+
"loss": 4.9338,
|
892 |
+
"step": 6300
|
893 |
+
},
|
894 |
+
{
|
895 |
+
"epoch": 0.3175,
|
896 |
+
"grad_norm": 0.59375,
|
897 |
+
"learning_rate": 0.000780253801120447,
|
898 |
+
"loss": 4.8996,
|
899 |
+
"step": 6350
|
900 |
+
},
|
901 |
+
{
|
902 |
+
"epoch": 0.32,
|
903 |
+
"grad_norm": 0.671875,
|
904 |
+
"learning_rate": 0.0007769600319330552,
|
905 |
+
"loss": 4.9161,
|
906 |
+
"step": 6400
|
907 |
+
},
|
908 |
+
{
|
909 |
+
"epoch": 0.3225,
|
910 |
+
"grad_norm": 0.5078125,
|
911 |
+
"learning_rate": 0.0007736488316696662,
|
912 |
+
"loss": 4.9648,
|
913 |
+
"step": 6450
|
914 |
+
},
|
915 |
+
{
|
916 |
+
"epoch": 0.325,
|
917 |
+
"grad_norm": 0.625,
|
918 |
+
"learning_rate": 0.0007703204087277988,
|
919 |
+
"loss": 4.908,
|
920 |
+
"step": 6500
|
921 |
+
},
|
922 |
+
{
|
923 |
+
"epoch": 0.3275,
|
924 |
+
"grad_norm": 0.50390625,
|
925 |
+
"learning_rate": 0.0007669749725889182,
|
926 |
+
"loss": 4.9536,
|
927 |
+
"step": 6550
|
928 |
+
},
|
929 |
+
{
|
930 |
+
"epoch": 0.33,
|
931 |
+
"grad_norm": 0.87890625,
|
932 |
+
"learning_rate": 0.0007636127338052513,
|
933 |
+
"loss": 4.8755,
|
934 |
+
"step": 6600
|
935 |
+
},
|
936 |
+
{
|
937 |
+
"epoch": 0.3325,
|
938 |
+
"grad_norm": 0.5546875,
|
939 |
+
"learning_rate": 0.0007602339039865362,
|
940 |
+
"loss": 4.8819,
|
941 |
+
"step": 6650
|
942 |
+
},
|
943 |
+
{
|
944 |
+
"epoch": 0.335,
|
945 |
+
"grad_norm": 0.53125,
|
946 |
+
"learning_rate": 0.0007568386957867032,
|
947 |
+
"loss": 4.8511,
|
948 |
+
"step": 6700
|
949 |
+
},
|
950 |
+
{
|
951 |
+
"epoch": 0.3375,
|
952 |
+
"grad_norm": 0.515625,
|
953 |
+
"learning_rate": 0.0007534273228904916,
|
954 |
+
"loss": 4.9324,
|
955 |
+
"step": 6750
|
956 |
+
},
|
957 |
+
{
|
958 |
+
"epoch": 0.34,
|
959 |
+
"grad_norm": 0.58203125,
|
960 |
+
"learning_rate": 0.00075,
|
961 |
+
"loss": 4.8898,
|
962 |
+
"step": 6800
|
963 |
+
},
|
964 |
+
{
|
965 |
+
"epoch": 0.3425,
|
966 |
+
"grad_norm": 0.63671875,
|
967 |
+
"learning_rate": 0.0007465569428211752,
|
968 |
+
"loss": 4.9447,
|
969 |
+
"step": 6850
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 0.345,
|
973 |
+
"grad_norm": 0.58203125,
|
974 |
+
"learning_rate": 0.0007430983680502344,
|
975 |
+
"loss": 4.8954,
|
976 |
+
"step": 6900
|
977 |
+
},
|
978 |
+
{
|
979 |
+
"epoch": 0.3475,
|
980 |
+
"grad_norm": 0.435546875,
|
981 |
+
"learning_rate": 0.0007396244933600284,
|
982 |
+
"loss": 4.8977,
|
983 |
+
"step": 6950
|
984 |
+
},
|
985 |
+
{
|
986 |
+
"epoch": 0.35,
|
987 |
+
"grad_norm": 0.7734375,
|
988 |
+
"learning_rate": 0.0007361355373863414,
|
989 |
+
"loss": 4.7837,
|
990 |
+
"step": 7000
|
991 |
+
},
|
992 |
+
{
|
993 |
+
"epoch": 0.3525,
|
994 |
+
"grad_norm": 0.458984375,
|
995 |
+
"learning_rate": 0.0007326317197141304,
|
996 |
+
"loss": 4.9241,
|
997 |
+
"step": 7050
|
998 |
+
},
|
999 |
+
{
|
1000 |
+
"epoch": 0.355,
|
1001 |
+
"grad_norm": 0.58203125,
|
1002 |
+
"learning_rate": 0.0007291132608637052,
|
1003 |
+
"loss": 4.9113,
|
1004 |
+
"step": 7100
|
1005 |
+
},
|
1006 |
+
{
|
1007 |
+
"epoch": 0.3575,
|
1008 |
+
"grad_norm": 0.65234375,
|
1009 |
+
"learning_rate": 0.0007255803822768504,
|
1010 |
+
"loss": 4.9187,
|
1011 |
+
"step": 7150
|
1012 |
+
},
|
1013 |
+
{
|
1014 |
+
"epoch": 0.36,
|
1015 |
+
"grad_norm": 0.60546875,
|
1016 |
+
"learning_rate": 0.0007220333063028871,
|
1017 |
+
"loss": 4.8632,
|
1018 |
+
"step": 7200
|
1019 |
+
},
|
1020 |
+
{
|
1021 |
+
"epoch": 0.3625,
|
1022 |
+
"grad_norm": 0.58203125,
|
1023 |
+
"learning_rate": 0.0007184722561846798,
|
1024 |
+
"loss": 5.0078,
|
1025 |
+
"step": 7250
|
1026 |
+
},
|
1027 |
+
{
|
1028 |
+
"epoch": 0.365,
|
1029 |
+
"grad_norm": 0.50390625,
|
1030 |
+
"learning_rate": 0.0007148974560445859,
|
1031 |
+
"loss": 4.7923,
|
1032 |
+
"step": 7300
|
1033 |
+
},
|
1034 |
+
{
|
1035 |
+
"epoch": 0.3675,
|
1036 |
+
"grad_norm": 0.52734375,
|
1037 |
+
"learning_rate": 0.0007113091308703497,
|
1038 |
+
"loss": 4.8948,
|
1039 |
+
"step": 7350
|
1040 |
+
},
|
1041 |
+
{
|
1042 |
+
"epoch": 0.37,
|
1043 |
+
"grad_norm": 0.466796875,
|
1044 |
+
"learning_rate": 0.0007077075065009433,
|
1045 |
+
"loss": 4.8259,
|
1046 |
+
"step": 7400
|
1047 |
+
},
|
1048 |
+
{
|
1049 |
+
"epoch": 0.3725,
|
1050 |
+
"grad_norm": 0.55859375,
|
1051 |
+
"learning_rate": 0.0007040928096123516,
|
1052 |
+
"loss": 4.8009,
|
1053 |
+
"step": 7450
|
1054 |
+
},
|
1055 |
+
{
|
1056 |
+
"epoch": 0.375,
|
1057 |
+
"grad_norm": 0.4375,
|
1058 |
+
"learning_rate": 0.0007004652677033068,
|
1059 |
+
"loss": 4.847,
|
1060 |
+
"step": 7500
|
1061 |
+
},
|
1062 |
+
{
|
1063 |
+
"epoch": 0.3775,
|
1064 |
+
"grad_norm": 0.486328125,
|
1065 |
+
"learning_rate": 0.0006968251090809707,
|
1066 |
+
"loss": 4.835,
|
1067 |
+
"step": 7550
|
1068 |
+
},
|
1069 |
+
{
|
1070 |
+
"epoch": 0.38,
|
1071 |
+
"grad_norm": 0.455078125,
|
1072 |
+
"learning_rate": 0.0006931725628465643,
|
1073 |
+
"loss": 4.8838,
|
1074 |
+
"step": 7600
|
1075 |
+
},
|
1076 |
+
{
|
1077 |
+
"epoch": 0.3825,
|
1078 |
+
"grad_norm": 0.68359375,
|
1079 |
+
"learning_rate": 0.0006895078588809502,
|
1080 |
+
"loss": 4.8705,
|
1081 |
+
"step": 7650
|
1082 |
+
},
|
1083 |
+
{
|
1084 |
+
"epoch": 0.385,
|
1085 |
+
"grad_norm": 0.46875,
|
1086 |
+
"learning_rate": 0.0006858312278301637,
|
1087 |
+
"loss": 4.763,
|
1088 |
+
"step": 7700
|
1089 |
+
},
|
1090 |
+
{
|
1091 |
+
"epoch": 0.3875,
|
1092 |
+
"grad_norm": 0.53125,
|
1093 |
+
"learning_rate": 0.0006821429010908972,
|
1094 |
+
"loss": 4.8961,
|
1095 |
+
"step": 7750
|
1096 |
+
},
|
1097 |
+
{
|
1098 |
+
"epoch": 0.39,
|
1099 |
+
"grad_norm": 0.53125,
|
1100 |
+
"learning_rate": 0.0006784431107959359,
|
1101 |
+
"loss": 4.8141,
|
1102 |
+
"step": 7800
|
1103 |
+
},
|
1104 |
+
{
|
1105 |
+
"epoch": 0.3925,
|
1106 |
+
"grad_norm": 0.5078125,
|
1107 |
+
"learning_rate": 0.0006747320897995492,
|
1108 |
+
"loss": 4.8718,
|
1109 |
+
"step": 7850
|
1110 |
+
},
|
1111 |
+
{
|
1112 |
+
"epoch": 0.395,
|
1113 |
+
"grad_norm": 0.5390625,
|
1114 |
+
"learning_rate": 0.0006710100716628344,
|
1115 |
+
"loss": 4.8394,
|
1116 |
+
"step": 7900
|
1117 |
+
},
|
1118 |
+
{
|
1119 |
+
"epoch": 0.3975,
|
1120 |
+
"grad_norm": 0.65234375,
|
1121 |
+
"learning_rate": 0.0006672772906390176,
|
1122 |
+
"loss": 4.8264,
|
1123 |
+
"step": 7950
|
1124 |
+
},
|
1125 |
+
{
|
1126 |
+
"epoch": 0.4,
|
1127 |
+
"grad_norm": 0.80078125,
|
1128 |
+
"learning_rate": 0.0006635339816587109,
|
1129 |
+
"loss": 4.769,
|
1130 |
+
"step": 8000
|
1131 |
+
},
|
1132 |
+
{
|
1133 |
+
"epoch": 0.4025,
|
1134 |
+
"grad_norm": 0.61328125,
|
1135 |
+
"learning_rate": 0.000659780380315125,
|
1136 |
+
"loss": 4.8428,
|
1137 |
+
"step": 8050
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"epoch": 0.405,
|
1141 |
+
"grad_norm": 0.515625,
|
1142 |
+
"learning_rate": 0.0006560167228492435,
|
1143 |
+
"loss": 4.8591,
|
1144 |
+
"step": 8100
|
1145 |
+
},
|
1146 |
+
{
|
1147 |
+
"epoch": 0.4075,
|
1148 |
+
"grad_norm": 0.80078125,
|
1149 |
+
"learning_rate": 0.0006522432461349536,
|
1150 |
+
"loss": 4.9224,
|
1151 |
+
"step": 8150
|
1152 |
+
},
|
1153 |
+
{
|
1154 |
+
"epoch": 0.41,
|
1155 |
+
"grad_norm": 0.640625,
|
1156 |
+
"learning_rate": 0.0006484601876641375,
|
1157 |
+
"loss": 4.7773,
|
1158 |
+
"step": 8200
|
1159 |
+
},
|
1160 |
+
{
|
1161 |
+
"epoch": 0.4125,
|
1162 |
+
"grad_norm": 0.5703125,
|
1163 |
+
"learning_rate": 0.0006446677855317265,
|
1164 |
+
"loss": 4.8107,
|
1165 |
+
"step": 8250
|
1166 |
+
},
|
1167 |
+
{
|
1168 |
+
"epoch": 0.415,
|
1169 |
+
"grad_norm": 0.53515625,
|
1170 |
+
"learning_rate": 0.0006408662784207149,
|
1171 |
+
"loss": 4.8252,
|
1172 |
+
"step": 8300
|
1173 |
+
},
|
1174 |
+
{
|
1175 |
+
"epoch": 0.4175,
|
1176 |
+
"grad_norm": 0.546875,
|
1177 |
+
"learning_rate": 0.0006370559055871389,
|
1178 |
+
"loss": 4.8562,
|
1179 |
+
"step": 8350
|
1180 |
+
},
|
1181 |
+
{
|
1182 |
+
"epoch": 0.42,
|
1183 |
+
"grad_norm": 0.5625,
|
1184 |
+
"learning_rate": 0.0006332369068450174,
|
1185 |
+
"loss": 4.7781,
|
1186 |
+
"step": 8400
|
1187 |
+
},
|
1188 |
+
{
|
1189 |
+
"epoch": 0.4225,
|
1190 |
+
"grad_norm": 0.462890625,
|
1191 |
+
"learning_rate": 0.0006294095225512603,
|
1192 |
+
"loss": 4.8325,
|
1193 |
+
"step": 8450
|
1194 |
+
},
|
1195 |
+
{
|
1196 |
+
"epoch": 0.425,
|
1197 |
+
"grad_norm": 0.45703125,
|
1198 |
+
"learning_rate": 0.0006255739935905395,
|
1199 |
+
"loss": 4.8363,
|
1200 |
+
"step": 8500
|
1201 |
+
},
|
1202 |
+
{
|
1203 |
+
"epoch": 0.4275,
|
1204 |
+
"grad_norm": 0.6015625,
|
1205 |
+
"learning_rate": 0.0006217305613601295,
|
1206 |
+
"loss": 4.8252,
|
1207 |
+
"step": 8550
|
1208 |
+
},
|
1209 |
+
{
|
1210 |
+
"epoch": 0.43,
|
1211 |
+
"grad_norm": 0.609375,
|
1212 |
+
"learning_rate": 0.0006178794677547138,
|
1213 |
+
"loss": 4.8079,
|
1214 |
+
"step": 8600
|
1215 |
+
},
|
1216 |
+
{
|
1217 |
+
"epoch": 0.4325,
|
1218 |
+
"grad_norm": 0.470703125,
|
1219 |
+
"learning_rate": 0.0006140209551511608,
|
1220 |
+
"loss": 4.8508,
|
1221 |
+
"step": 8650
|
1222 |
+
},
|
1223 |
+
{
|
1224 |
+
"epoch": 0.435,
|
1225 |
+
"grad_norm": 0.53515625,
|
1226 |
+
"learning_rate": 0.0006101552663932703,
|
1227 |
+
"loss": 4.8389,
|
1228 |
+
"step": 8700
|
1229 |
+
},
|
1230 |
+
{
|
1231 |
+
"epoch": 0.4375,
|
1232 |
+
"grad_norm": 0.66796875,
|
1233 |
+
"learning_rate": 0.0006062826447764884,
|
1234 |
+
"loss": 4.7695,
|
1235 |
+
"step": 8750
|
1236 |
+
},
|
1237 |
+
{
|
1238 |
+
"epoch": 0.44,
|
1239 |
+
"grad_norm": 0.455078125,
|
1240 |
+
"learning_rate": 0.0006024033340325954,
|
1241 |
+
"loss": 4.7953,
|
1242 |
+
"step": 8800
|
1243 |
+
},
|
1244 |
+
{
|
1245 |
+
"epoch": 0.4425,
|
1246 |
+
"grad_norm": 0.56640625,
|
1247 |
+
"learning_rate": 0.0005985175783143666,
|
1248 |
+
"loss": 4.8144,
|
1249 |
+
"step": 8850
|
1250 |
+
},
|
1251 |
+
{
|
1252 |
+
"epoch": 0.445,
|
1253 |
+
"grad_norm": 0.48046875,
|
1254 |
+
"learning_rate": 0.0005946256221802051,
|
1255 |
+
"loss": 4.8123,
|
1256 |
+
"step": 8900
|
1257 |
+
},
|
1258 |
+
{
|
1259 |
+
"epoch": 0.4475,
|
1260 |
+
"grad_norm": 0.52734375,
|
1261 |
+
"learning_rate": 0.0005907277105787513,
|
1262 |
+
"loss": 4.7778,
|
1263 |
+
"step": 8950
|
1264 |
+
},
|
1265 |
+
{
|
1266 |
+
"epoch": 0.45,
|
1267 |
+
"grad_norm": 0.53515625,
|
1268 |
+
"learning_rate": 0.0005868240888334653,
|
1269 |
+
"loss": 4.7237,
|
1270 |
+
"step": 9000
|
1271 |
+
},
|
1272 |
+
{
|
1273 |
+
"epoch": 0.4525,
|
1274 |
+
"grad_norm": 0.58984375,
|
1275 |
+
"learning_rate": 0.0005829150026271871,
|
1276 |
+
"loss": 4.7938,
|
1277 |
+
"step": 9050
|
1278 |
+
},
|
1279 |
+
{
|
1280 |
+
"epoch": 0.455,
|
1281 |
+
"grad_norm": 0.56640625,
|
1282 |
+
"learning_rate": 0.000579000697986675,
|
1283 |
+
"loss": 4.7879,
|
1284 |
+
"step": 9100
|
1285 |
+
},
|
1286 |
+
{
|
1287 |
+
"epoch": 0.4575,
|
1288 |
+
"grad_norm": 0.578125,
|
1289 |
+
"learning_rate": 0.0005750814212671201,
|
1290 |
+
"loss": 4.7317,
|
1291 |
+
"step": 9150
|
1292 |
+
},
|
1293 |
+
{
|
1294 |
+
"epoch": 0.46,
|
1295 |
+
"grad_norm": 0.578125,
|
1296 |
+
"learning_rate": 0.0005711574191366427,
|
1297 |
+
"loss": 4.7819,
|
1298 |
+
"step": 9200
|
1299 |
+
},
|
1300 |
+
{
|
1301 |
+
"epoch": 0.4625,
|
1302 |
+
"grad_norm": 0.61328125,
|
1303 |
+
"learning_rate": 0.0005672289385607659,
|
1304 |
+
"loss": 4.8393,
|
1305 |
+
"step": 9250
|
1306 |
+
},
|
1307 |
+
{
|
1308 |
+
"epoch": 0.465,
|
1309 |
+
"grad_norm": 0.69921875,
|
1310 |
+
"learning_rate": 0.0005632962267868747,
|
1311 |
+
"loss": 4.7797,
|
1312 |
+
"step": 9300
|
1313 |
+
},
|
1314 |
+
{
|
1315 |
+
"epoch": 0.4675,
|
1316 |
+
"grad_norm": 0.71484375,
|
1317 |
+
"learning_rate": 0.0005593595313286526,
|
1318 |
+
"loss": 4.764,
|
1319 |
+
"step": 9350
|
1320 |
+
},
|
1321 |
+
{
|
1322 |
+
"epoch": 0.47,
|
1323 |
+
"grad_norm": 0.4765625,
|
1324 |
+
"learning_rate": 0.0005554190999505056,
|
1325 |
+
"loss": 4.7639,
|
1326 |
+
"step": 9400
|
1327 |
+
},
|
1328 |
+
{
|
1329 |
+
"epoch": 0.4725,
|
1330 |
+
"grad_norm": 0.53125,
|
1331 |
+
"learning_rate": 0.0005514751806519673,
|
1332 |
+
"loss": 4.8052,
|
1333 |
+
"step": 9450
|
1334 |
+
},
|
1335 |
+
{
|
1336 |
+
"epoch": 0.475,
|
1337 |
+
"grad_norm": 0.6328125,
|
1338 |
+
"learning_rate": 0.0005475280216520913,
|
1339 |
+
"loss": 4.8682,
|
1340 |
+
"step": 9500
|
1341 |
+
},
|
1342 |
+
{
|
1343 |
+
"epoch": 0.4775,
|
1344 |
+
"grad_norm": 0.640625,
|
1345 |
+
"learning_rate": 0.0005435778713738292,
|
1346 |
+
"loss": 4.748,
|
1347 |
+
"step": 9550
|
1348 |
+
},
|
1349 |
+
{
|
1350 |
+
"epoch": 0.48,
|
1351 |
+
"grad_norm": 0.66015625,
|
1352 |
+
"learning_rate": 0.0005396249784283942,
|
1353 |
+
"loss": 4.7625,
|
1354 |
+
"step": 9600
|
1355 |
+
},
|
1356 |
+
{
|
1357 |
+
"epoch": 0.4825,
|
1358 |
+
"grad_norm": 0.51171875,
|
1359 |
+
"learning_rate": 0.0005356695915996161,
|
1360 |
+
"loss": 4.824,
|
1361 |
+
"step": 9650
|
1362 |
+
},
|
1363 |
+
{
|
1364 |
+
"epoch": 0.485,
|
1365 |
+
"grad_norm": 0.54296875,
|
1366 |
+
"learning_rate": 0.0005317119598282822,
|
1367 |
+
"loss": 4.7946,
|
1368 |
+
"step": 9700
|
1369 |
+
},
|
1370 |
+
{
|
1371 |
+
"epoch": 0.4875,
|
1372 |
+
"grad_norm": 0.671875,
|
1373 |
+
"learning_rate": 0.0005277523321964701,
|
1374 |
+
"loss": 4.8476,
|
1375 |
+
"step": 9750
|
1376 |
+
},
|
1377 |
+
{
|
1378 |
+
"epoch": 0.49,
|
1379 |
+
"grad_norm": 0.55078125,
|
1380 |
+
"learning_rate": 0.0005237909579118712,
|
1381 |
+
"loss": 4.735,
|
1382 |
+
"step": 9800
|
1383 |
+
},
|
1384 |
+
{
|
1385 |
+
"epoch": 0.4925,
|
1386 |
+
"grad_norm": 0.47265625,
|
1387 |
+
"learning_rate": 0.0005198280862921062,
|
1388 |
+
"loss": 4.8119,
|
1389 |
+
"step": 9850
|
1390 |
+
},
|
1391 |
+
{
|
1392 |
+
"epoch": 0.495,
|
1393 |
+
"grad_norm": 0.49609375,
|
1394 |
+
"learning_rate": 0.0005158639667490339,
|
1395 |
+
"loss": 4.794,
|
1396 |
+
"step": 9900
|
1397 |
+
},
|
1398 |
+
{
|
1399 |
+
"epoch": 0.4975,
|
1400 |
+
"grad_norm": 1.0546875,
|
1401 |
+
"learning_rate": 0.0005118988487730537,
|
1402 |
+
"loss": 4.7226,
|
1403 |
+
"step": 9950
|
1404 |
+
},
|
1405 |
+
{
|
1406 |
+
"epoch": 0.5,
|
1407 |
+
"grad_norm": 0.56640625,
|
1408 |
+
"learning_rate": 0.000507932981917404,
|
1409 |
+
"loss": 4.9084,
|
1410 |
+
"step": 10000
|
1411 |
+
},
|
1412 |
+
{
|
1413 |
+
"epoch": 0.5025,
|
1414 |
+
"grad_norm": 0.54296875,
|
1415 |
+
"learning_rate": 0.0005039666157824549,
|
1416 |
+
"loss": 4.6737,
|
1417 |
+
"step": 10050
|
1418 |
+
},
|
1419 |
+
{
|
1420 |
+
"epoch": 0.505,
|
1421 |
+
"grad_norm": 0.59765625,
|
1422 |
+
"learning_rate": 0.0005,
|
1423 |
+
"loss": 4.7808,
|
1424 |
+
"step": 10100
|
1425 |
+
}
|
1426 |
+
],
|
1427 |
+
"logging_steps": 50,
|
1428 |
+
"max_steps": 20000,
|
1429 |
+
"num_input_tokens_seen": 0,
|
1430 |
+
"num_train_epochs": 9223372036854775807,
|
1431 |
+
"save_steps": 100,
|
1432 |
+
"stateful_callbacks": {
|
1433 |
+
"TrainerControl": {
|
1434 |
+
"args": {
|
1435 |
+
"should_epoch_stop": false,
|
1436 |
+
"should_evaluate": false,
|
1437 |
+
"should_log": false,
|
1438 |
+
"should_save": true,
|
1439 |
+
"should_training_stop": false
|
1440 |
+
},
|
1441 |
+
"attributes": {}
|
1442 |
+
}
|
1443 |
+
},
|
1444 |
+
"total_flos": 1.47044614668288e+16,
|
1445 |
+
"train_batch_size": 16,
|
1446 |
+
"trial_name": null,
|
1447 |
+
"trial_params": null
|
1448 |
+
}
|
checkpoint-10100/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:83031100336f2dd4b43e2539aae80864690fb3fc70607039ce5c50a2ff4a21bb
|
3 |
+
size 5368
|
checkpoint-10200/config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"GPTNeoXForCausalLM"
|
4 |
+
],
|
5 |
+
"attention_bias": true,
|
6 |
+
"attention_dropout": 0.0,
|
7 |
+
"bos_token_id": 0,
|
8 |
+
"classifier_dropout": 0.1,
|
9 |
+
"eos_token_id": 0,
|
10 |
+
"hidden_act": "gelu",
|
11 |
+
"hidden_dropout": 0.0,
|
12 |
+
"hidden_size": 128,
|
13 |
+
"initializer_range": 0.02,
|
14 |
+
"intermediate_size": 512,
|
15 |
+
"layer_norm_eps": 1e-05,
|
16 |
+
"max_position_embeddings": 2048,
|
17 |
+
"model_type": "gpt_neox",
|
18 |
+
"num_attention_heads": 4,
|
19 |
+
"num_hidden_layers": 6,
|
20 |
+
"partial_rotary_factor": 0.25,
|
21 |
+
"rope_scaling": null,
|
22 |
+
"rope_theta": 10000,
|
23 |
+
"rotary_emb_base": 10000,
|
24 |
+
"rotary_pct": 0.25,
|
25 |
+
"tie_word_embeddings": false,
|
26 |
+
"torch_dtype": "bfloat16",
|
27 |
+
"transformers_version": "4.50.3",
|
28 |
+
"use_cache": true,
|
29 |
+
"use_parallel_residual": true,
|
30 |
+
"vocab_size": 50304
|
31 |
+
}
|
checkpoint-10200/generation_config.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
+
"eos_token_id": 0,
|
5 |
+
"transformers_version": "4.50.3"
|
6 |
+
}
|
checkpoint-10200/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1df07a325f20501a191f97965770735aefaf24d0fa8e1a88f416f2b9284b1e64
|
3 |
+
size 28144000
|
checkpoint-10200/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:91991637ad3c7e37057533f6361addfbe29a97057885569ee5ec22209af45da9
|
3 |
+
size 56334394
|
checkpoint-10200/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
|
3 |
+
size 14244
|