Commit
•
3b118fe
1
Parent(s):
b16bd8b
End of training
Browse files- last-checkpoint/generation_config.json → generation_config.json +0 -0
- last-checkpoint/added_tokens.json +0 -6
- last-checkpoint/config.json +0 -32
- last-checkpoint/optimizer.pt +0 -3
- last-checkpoint/pytorch_model.bin +0 -3
- last-checkpoint/rng_state.pth +0 -3
- last-checkpoint/scheduler.pt +0 -3
- last-checkpoint/special_tokens_map.json +0 -11
- last-checkpoint/spiece.model +0 -3
- last-checkpoint/tokenizer_config.json +0 -12
- last-checkpoint/trainer_state.json +0 -2416
- last-checkpoint/training_args.bin +0 -3
- pytorch_model.bin +1 -1
- runs/Feb07_05-04-07_5214b674e698/events.out.tfevents.1675746342.5214b674e698.342.0 +2 -2
last-checkpoint/generation_config.json → generation_config.json
RENAMED
|
File without changes
|
last-checkpoint/added_tokens.json
DELETED
|
@@ -1,6 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"translate Arabic to English:": 250101,
|
| 3 |
-
"translate Arabic to Thai:": 250100,
|
| 4 |
-
"translate Thai to Arabic:": 250102,
|
| 5 |
-
"translate Thai to English:": 250103
|
| 6 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/config.json
DELETED
|
@@ -1,32 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "google/mt5-small",
|
| 3 |
-
"architectures": [
|
| 4 |
-
"MT5ForConditionalGeneration"
|
| 5 |
-
],
|
| 6 |
-
"d_ff": 1024,
|
| 7 |
-
"d_kv": 64,
|
| 8 |
-
"d_model": 512,
|
| 9 |
-
"decoder_start_token_id": 0,
|
| 10 |
-
"dense_act_fn": "gelu_new",
|
| 11 |
-
"dropout_rate": 0.1,
|
| 12 |
-
"eos_token_id": 1,
|
| 13 |
-
"feed_forward_proj": "gated-gelu",
|
| 14 |
-
"initializer_factor": 1.0,
|
| 15 |
-
"is_encoder_decoder": true,
|
| 16 |
-
"is_gated_act": true,
|
| 17 |
-
"layer_norm_epsilon": 1e-06,
|
| 18 |
-
"max_length": 300,
|
| 19 |
-
"model_type": "mt5",
|
| 20 |
-
"num_decoder_layers": 8,
|
| 21 |
-
"num_heads": 6,
|
| 22 |
-
"num_layers": 8,
|
| 23 |
-
"pad_token_id": 0,
|
| 24 |
-
"relative_attention_max_distance": 128,
|
| 25 |
-
"relative_attention_num_buckets": 32,
|
| 26 |
-
"tie_word_embeddings": false,
|
| 27 |
-
"tokenizer_class": "T5Tokenizer",
|
| 28 |
-
"torch_dtype": "float32",
|
| 29 |
-
"transformers_version": "4.26.0",
|
| 30 |
-
"use_cache": true,
|
| 31 |
-
"vocab_size": 250104
|
| 32 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/optimizer.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0cedde736261632f89acf33553d7f1b796fa4666797a76600bc7414afba87e83
|
| 3 |
-
size 2401461637
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/pytorch_model.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:3c830ba03e87ef89c167a65183788768bf905dc7f4c689fe476c351a7fc9b93b
|
| 3 |
-
size 1200739717
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/rng_state.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:dd594782425d26032039a6c2b147e5861095f1958bd2b0f4a2e42679340bff32
|
| 3 |
-
size 14575
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/scheduler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:e9febce275791009bb39a3242ecdd95ff1e50bc7a7886ff99da486505feed461
|
| 3 |
-
size 627
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/special_tokens_map.json
DELETED
|
@@ -1,11 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"additional_special_tokens": [
|
| 3 |
-
"translate Arabic to Thai:",
|
| 4 |
-
"translate Arabic to English:",
|
| 5 |
-
"translate Thai to Arabic:",
|
| 6 |
-
"translate Thai to English:"
|
| 7 |
-
],
|
| 8 |
-
"eos_token": "</s>",
|
| 9 |
-
"pad_token": "<pad>",
|
| 10 |
-
"unk_token": "<unk>"
|
| 11 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/spiece.model
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:ef78f86560d809067d12bac6c09f19a462cb3af3f54d2b8acbba26e1433125d6
|
| 3 |
-
size 4309802
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/tokenizer_config.json
DELETED
|
@@ -1,12 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"additional_special_tokens": null,
|
| 3 |
-
"eos_token": "</s>",
|
| 4 |
-
"extra_ids": 0,
|
| 5 |
-
"model_max_length": 1000000000000000019884624838656,
|
| 6 |
-
"name_or_path": "google/mt5-small",
|
| 7 |
-
"pad_token": "<pad>",
|
| 8 |
-
"sp_model_kwargs": {},
|
| 9 |
-
"special_tokens_map_file": "/root/.cache/huggingface/hub/models--google--mt5-small/snapshots/38f23af8ec210eb6c376d40e9c56bd25a80f195d/special_tokens_map.json",
|
| 10 |
-
"tokenizer_class": "T5Tokenizer",
|
| 11 |
-
"unk_token": "<unk>"
|
| 12 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/trainer_state.json
DELETED
|
@@ -1,2416 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": null,
|
| 3 |
-
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 2.9583172175775787,
|
| 5 |
-
"global_step": 198864,
|
| 6 |
-
"is_hyper_param_search": false,
|
| 7 |
-
"is_local_process_zero": true,
|
| 8 |
-
"is_world_process_zero": true,
|
| 9 |
-
"log_history": [
|
| 10 |
-
{
|
| 11 |
-
"epoch": 0.01,
|
| 12 |
-
"learning_rate": 1.9950413059216727e-05,
|
| 13 |
-
"loss": 14.0627,
|
| 14 |
-
"step": 500
|
| 15 |
-
},
|
| 16 |
-
{
|
| 17 |
-
"epoch": 0.01,
|
| 18 |
-
"learning_rate": 1.9900826118433453e-05,
|
| 19 |
-
"loss": 6.3799,
|
| 20 |
-
"step": 1000
|
| 21 |
-
},
|
| 22 |
-
{
|
| 23 |
-
"epoch": 0.02,
|
| 24 |
-
"learning_rate": 1.9851239177650176e-05,
|
| 25 |
-
"loss": 5.31,
|
| 26 |
-
"step": 1500
|
| 27 |
-
},
|
| 28 |
-
{
|
| 29 |
-
"epoch": 0.03,
|
| 30 |
-
"learning_rate": 1.9801652236866898e-05,
|
| 31 |
-
"loss": 4.9781,
|
| 32 |
-
"step": 2000
|
| 33 |
-
},
|
| 34 |
-
{
|
| 35 |
-
"epoch": 0.04,
|
| 36 |
-
"learning_rate": 1.9752065296083624e-05,
|
| 37 |
-
"loss": 4.7392,
|
| 38 |
-
"step": 2500
|
| 39 |
-
},
|
| 40 |
-
{
|
| 41 |
-
"epoch": 0.04,
|
| 42 |
-
"learning_rate": 1.970247835530035e-05,
|
| 43 |
-
"loss": 4.5779,
|
| 44 |
-
"step": 3000
|
| 45 |
-
},
|
| 46 |
-
{
|
| 47 |
-
"epoch": 0.05,
|
| 48 |
-
"learning_rate": 1.9652891414517075e-05,
|
| 49 |
-
"loss": 4.4691,
|
| 50 |
-
"step": 3500
|
| 51 |
-
},
|
| 52 |
-
{
|
| 53 |
-
"epoch": 0.06,
|
| 54 |
-
"learning_rate": 1.96033044737338e-05,
|
| 55 |
-
"loss": 4.3745,
|
| 56 |
-
"step": 4000
|
| 57 |
-
},
|
| 58 |
-
{
|
| 59 |
-
"epoch": 0.07,
|
| 60 |
-
"learning_rate": 1.9553717532950524e-05,
|
| 61 |
-
"loss": 4.2883,
|
| 62 |
-
"step": 4500
|
| 63 |
-
},
|
| 64 |
-
{
|
| 65 |
-
"epoch": 0.07,
|
| 66 |
-
"learning_rate": 1.9504130592167246e-05,
|
| 67 |
-
"loss": 4.2342,
|
| 68 |
-
"step": 5000
|
| 69 |
-
},
|
| 70 |
-
{
|
| 71 |
-
"epoch": 0.08,
|
| 72 |
-
"learning_rate": 1.9454543651383972e-05,
|
| 73 |
-
"loss": 4.1614,
|
| 74 |
-
"step": 5500
|
| 75 |
-
},
|
| 76 |
-
{
|
| 77 |
-
"epoch": 0.09,
|
| 78 |
-
"learning_rate": 1.9404956710600698e-05,
|
| 79 |
-
"loss": 4.1279,
|
| 80 |
-
"step": 6000
|
| 81 |
-
},
|
| 82 |
-
{
|
| 83 |
-
"epoch": 0.1,
|
| 84 |
-
"learning_rate": 1.9355369769817423e-05,
|
| 85 |
-
"loss": 4.0802,
|
| 86 |
-
"step": 6500
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 0.1,
|
| 90 |
-
"learning_rate": 1.930578282903415e-05,
|
| 91 |
-
"loss": 4.0298,
|
| 92 |
-
"step": 7000
|
| 93 |
-
},
|
| 94 |
-
{
|
| 95 |
-
"epoch": 0.11,
|
| 96 |
-
"learning_rate": 1.925619588825087e-05,
|
| 97 |
-
"loss": 3.9697,
|
| 98 |
-
"step": 7500
|
| 99 |
-
},
|
| 100 |
-
{
|
| 101 |
-
"epoch": 0.12,
|
| 102 |
-
"learning_rate": 1.9206608947467594e-05,
|
| 103 |
-
"loss": 3.9584,
|
| 104 |
-
"step": 8000
|
| 105 |
-
},
|
| 106 |
-
{
|
| 107 |
-
"epoch": 0.13,
|
| 108 |
-
"learning_rate": 1.915702200668432e-05,
|
| 109 |
-
"loss": 3.9196,
|
| 110 |
-
"step": 8500
|
| 111 |
-
},
|
| 112 |
-
{
|
| 113 |
-
"epoch": 0.13,
|
| 114 |
-
"learning_rate": 1.9107435065901046e-05,
|
| 115 |
-
"loss": 3.9081,
|
| 116 |
-
"step": 9000
|
| 117 |
-
},
|
| 118 |
-
{
|
| 119 |
-
"epoch": 0.14,
|
| 120 |
-
"learning_rate": 1.905784812511777e-05,
|
| 121 |
-
"loss": 3.8419,
|
| 122 |
-
"step": 9500
|
| 123 |
-
},
|
| 124 |
-
{
|
| 125 |
-
"epoch": 0.15,
|
| 126 |
-
"learning_rate": 1.9008261184334497e-05,
|
| 127 |
-
"loss": 3.8363,
|
| 128 |
-
"step": 10000
|
| 129 |
-
},
|
| 130 |
-
{
|
| 131 |
-
"epoch": 0.16,
|
| 132 |
-
"learning_rate": 1.895867424355122e-05,
|
| 133 |
-
"loss": 3.8047,
|
| 134 |
-
"step": 10500
|
| 135 |
-
},
|
| 136 |
-
{
|
| 137 |
-
"epoch": 0.16,
|
| 138 |
-
"learning_rate": 1.8909087302767945e-05,
|
| 139 |
-
"loss": 3.7728,
|
| 140 |
-
"step": 11000
|
| 141 |
-
},
|
| 142 |
-
{
|
| 143 |
-
"epoch": 0.17,
|
| 144 |
-
"learning_rate": 1.8859500361984668e-05,
|
| 145 |
-
"loss": 3.7731,
|
| 146 |
-
"step": 11500
|
| 147 |
-
},
|
| 148 |
-
{
|
| 149 |
-
"epoch": 0.18,
|
| 150 |
-
"learning_rate": 1.8809913421201393e-05,
|
| 151 |
-
"loss": 3.7408,
|
| 152 |
-
"step": 12000
|
| 153 |
-
},
|
| 154 |
-
{
|
| 155 |
-
"epoch": 0.19,
|
| 156 |
-
"learning_rate": 1.876032648041812e-05,
|
| 157 |
-
"loss": 3.7027,
|
| 158 |
-
"step": 12500
|
| 159 |
-
},
|
| 160 |
-
{
|
| 161 |
-
"epoch": 0.19,
|
| 162 |
-
"learning_rate": 1.8710739539634845e-05,
|
| 163 |
-
"loss": 3.6865,
|
| 164 |
-
"step": 13000
|
| 165 |
-
},
|
| 166 |
-
{
|
| 167 |
-
"epoch": 0.2,
|
| 168 |
-
"learning_rate": 1.8661152598851567e-05,
|
| 169 |
-
"loss": 3.6456,
|
| 170 |
-
"step": 13500
|
| 171 |
-
},
|
| 172 |
-
{
|
| 173 |
-
"epoch": 0.21,
|
| 174 |
-
"learning_rate": 1.8611565658068293e-05,
|
| 175 |
-
"loss": 3.6539,
|
| 176 |
-
"step": 14000
|
| 177 |
-
},
|
| 178 |
-
{
|
| 179 |
-
"epoch": 0.22,
|
| 180 |
-
"learning_rate": 1.8561978717285016e-05,
|
| 181 |
-
"loss": 3.6222,
|
| 182 |
-
"step": 14500
|
| 183 |
-
},
|
| 184 |
-
{
|
| 185 |
-
"epoch": 0.22,
|
| 186 |
-
"learning_rate": 1.851239177650174e-05,
|
| 187 |
-
"loss": 3.6127,
|
| 188 |
-
"step": 15000
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"epoch": 0.23,
|
| 192 |
-
"learning_rate": 1.8462804835718467e-05,
|
| 193 |
-
"loss": 3.6133,
|
| 194 |
-
"step": 15500
|
| 195 |
-
},
|
| 196 |
-
{
|
| 197 |
-
"epoch": 0.24,
|
| 198 |
-
"learning_rate": 1.8413217894935193e-05,
|
| 199 |
-
"loss": 3.5863,
|
| 200 |
-
"step": 16000
|
| 201 |
-
},
|
| 202 |
-
{
|
| 203 |
-
"epoch": 0.25,
|
| 204 |
-
"learning_rate": 1.8363630954151915e-05,
|
| 205 |
-
"loss": 3.5669,
|
| 206 |
-
"step": 16500
|
| 207 |
-
},
|
| 208 |
-
{
|
| 209 |
-
"epoch": 0.25,
|
| 210 |
-
"learning_rate": 1.831404401336864e-05,
|
| 211 |
-
"loss": 3.5518,
|
| 212 |
-
"step": 17000
|
| 213 |
-
},
|
| 214 |
-
{
|
| 215 |
-
"epoch": 0.26,
|
| 216 |
-
"learning_rate": 1.8264457072585367e-05,
|
| 217 |
-
"loss": 3.5368,
|
| 218 |
-
"step": 17500
|
| 219 |
-
},
|
| 220 |
-
{
|
| 221 |
-
"epoch": 0.27,
|
| 222 |
-
"learning_rate": 1.821487013180209e-05,
|
| 223 |
-
"loss": 3.5294,
|
| 224 |
-
"step": 18000
|
| 225 |
-
},
|
| 226 |
-
{
|
| 227 |
-
"epoch": 0.28,
|
| 228 |
-
"learning_rate": 1.8165283191018815e-05,
|
| 229 |
-
"loss": 3.5097,
|
| 230 |
-
"step": 18500
|
| 231 |
-
},
|
| 232 |
-
{
|
| 233 |
-
"epoch": 0.28,
|
| 234 |
-
"learning_rate": 1.811569625023554e-05,
|
| 235 |
-
"loss": 3.5198,
|
| 236 |
-
"step": 19000
|
| 237 |
-
},
|
| 238 |
-
{
|
| 239 |
-
"epoch": 0.29,
|
| 240 |
-
"learning_rate": 1.8066109309452263e-05,
|
| 241 |
-
"loss": 3.4702,
|
| 242 |
-
"step": 19500
|
| 243 |
-
},
|
| 244 |
-
{
|
| 245 |
-
"epoch": 0.3,
|
| 246 |
-
"learning_rate": 1.801652236866899e-05,
|
| 247 |
-
"loss": 3.485,
|
| 248 |
-
"step": 20000
|
| 249 |
-
},
|
| 250 |
-
{
|
| 251 |
-
"epoch": 0.3,
|
| 252 |
-
"learning_rate": 1.7966935427885715e-05,
|
| 253 |
-
"loss": 3.4853,
|
| 254 |
-
"step": 20500
|
| 255 |
-
},
|
| 256 |
-
{
|
| 257 |
-
"epoch": 0.31,
|
| 258 |
-
"learning_rate": 1.7917348487102437e-05,
|
| 259 |
-
"loss": 3.4395,
|
| 260 |
-
"step": 21000
|
| 261 |
-
},
|
| 262 |
-
{
|
| 263 |
-
"epoch": 0.32,
|
| 264 |
-
"learning_rate": 1.7867761546319163e-05,
|
| 265 |
-
"loss": 3.4515,
|
| 266 |
-
"step": 21500
|
| 267 |
-
},
|
| 268 |
-
{
|
| 269 |
-
"epoch": 0.33,
|
| 270 |
-
"learning_rate": 1.781817460553589e-05,
|
| 271 |
-
"loss": 3.4307,
|
| 272 |
-
"step": 22000
|
| 273 |
-
},
|
| 274 |
-
{
|
| 275 |
-
"epoch": 0.33,
|
| 276 |
-
"learning_rate": 1.776858766475261e-05,
|
| 277 |
-
"loss": 3.4343,
|
| 278 |
-
"step": 22500
|
| 279 |
-
},
|
| 280 |
-
{
|
| 281 |
-
"epoch": 0.34,
|
| 282 |
-
"learning_rate": 1.7719000723969337e-05,
|
| 283 |
-
"loss": 3.4053,
|
| 284 |
-
"step": 23000
|
| 285 |
-
},
|
| 286 |
-
{
|
| 287 |
-
"epoch": 0.35,
|
| 288 |
-
"learning_rate": 1.7669413783186063e-05,
|
| 289 |
-
"loss": 3.4008,
|
| 290 |
-
"step": 23500
|
| 291 |
-
},
|
| 292 |
-
{
|
| 293 |
-
"epoch": 0.36,
|
| 294 |
-
"learning_rate": 1.7619826842402785e-05,
|
| 295 |
-
"loss": 3.3951,
|
| 296 |
-
"step": 24000
|
| 297 |
-
},
|
| 298 |
-
{
|
| 299 |
-
"epoch": 0.36,
|
| 300 |
-
"learning_rate": 1.757023990161951e-05,
|
| 301 |
-
"loss": 3.3871,
|
| 302 |
-
"step": 24500
|
| 303 |
-
},
|
| 304 |
-
{
|
| 305 |
-
"epoch": 0.37,
|
| 306 |
-
"learning_rate": 1.7520652960836234e-05,
|
| 307 |
-
"loss": 3.3822,
|
| 308 |
-
"step": 25000
|
| 309 |
-
},
|
| 310 |
-
{
|
| 311 |
-
"epoch": 0.38,
|
| 312 |
-
"learning_rate": 1.747106602005296e-05,
|
| 313 |
-
"loss": 3.3816,
|
| 314 |
-
"step": 25500
|
| 315 |
-
},
|
| 316 |
-
{
|
| 317 |
-
"epoch": 0.39,
|
| 318 |
-
"learning_rate": 1.7421479079269685e-05,
|
| 319 |
-
"loss": 3.3759,
|
| 320 |
-
"step": 26000
|
| 321 |
-
},
|
| 322 |
-
{
|
| 323 |
-
"epoch": 0.39,
|
| 324 |
-
"learning_rate": 1.737189213848641e-05,
|
| 325 |
-
"loss": 3.3624,
|
| 326 |
-
"step": 26500
|
| 327 |
-
},
|
| 328 |
-
{
|
| 329 |
-
"epoch": 0.4,
|
| 330 |
-
"learning_rate": 1.7322305197703137e-05,
|
| 331 |
-
"loss": 3.3535,
|
| 332 |
-
"step": 27000
|
| 333 |
-
},
|
| 334 |
-
{
|
| 335 |
-
"epoch": 0.41,
|
| 336 |
-
"learning_rate": 1.727271825691986e-05,
|
| 337 |
-
"loss": 3.3366,
|
| 338 |
-
"step": 27500
|
| 339 |
-
},
|
| 340 |
-
{
|
| 341 |
-
"epoch": 0.42,
|
| 342 |
-
"learning_rate": 1.722313131613658e-05,
|
| 343 |
-
"loss": 3.3245,
|
| 344 |
-
"step": 28000
|
| 345 |
-
},
|
| 346 |
-
{
|
| 347 |
-
"epoch": 0.42,
|
| 348 |
-
"learning_rate": 1.7173544375353307e-05,
|
| 349 |
-
"loss": 3.3575,
|
| 350 |
-
"step": 28500
|
| 351 |
-
},
|
| 352 |
-
{
|
| 353 |
-
"epoch": 0.43,
|
| 354 |
-
"learning_rate": 1.7123957434570033e-05,
|
| 355 |
-
"loss": 3.3133,
|
| 356 |
-
"step": 29000
|
| 357 |
-
},
|
| 358 |
-
{
|
| 359 |
-
"epoch": 0.44,
|
| 360 |
-
"learning_rate": 1.707437049378676e-05,
|
| 361 |
-
"loss": 3.3124,
|
| 362 |
-
"step": 29500
|
| 363 |
-
},
|
| 364 |
-
{
|
| 365 |
-
"epoch": 0.45,
|
| 366 |
-
"learning_rate": 1.7024783553003485e-05,
|
| 367 |
-
"loss": 3.3295,
|
| 368 |
-
"step": 30000
|
| 369 |
-
},
|
| 370 |
-
{
|
| 371 |
-
"epoch": 0.45,
|
| 372 |
-
"learning_rate": 1.6975196612220207e-05,
|
| 373 |
-
"loss": 3.3192,
|
| 374 |
-
"step": 30500
|
| 375 |
-
},
|
| 376 |
-
{
|
| 377 |
-
"epoch": 0.46,
|
| 378 |
-
"learning_rate": 1.692560967143693e-05,
|
| 379 |
-
"loss": 3.3241,
|
| 380 |
-
"step": 31000
|
| 381 |
-
},
|
| 382 |
-
{
|
| 383 |
-
"epoch": 0.47,
|
| 384 |
-
"learning_rate": 1.6876022730653655e-05,
|
| 385 |
-
"loss": 3.2989,
|
| 386 |
-
"step": 31500
|
| 387 |
-
},
|
| 388 |
-
{
|
| 389 |
-
"epoch": 0.48,
|
| 390 |
-
"learning_rate": 1.682643578987038e-05,
|
| 391 |
-
"loss": 3.2956,
|
| 392 |
-
"step": 32000
|
| 393 |
-
},
|
| 394 |
-
{
|
| 395 |
-
"epoch": 0.48,
|
| 396 |
-
"learning_rate": 1.6776848849087107e-05,
|
| 397 |
-
"loss": 3.2889,
|
| 398 |
-
"step": 32500
|
| 399 |
-
},
|
| 400 |
-
{
|
| 401 |
-
"epoch": 0.49,
|
| 402 |
-
"learning_rate": 1.6727261908303833e-05,
|
| 403 |
-
"loss": 3.2934,
|
| 404 |
-
"step": 33000
|
| 405 |
-
},
|
| 406 |
-
{
|
| 407 |
-
"epoch": 0.5,
|
| 408 |
-
"learning_rate": 1.6677674967520555e-05,
|
| 409 |
-
"loss": 3.2642,
|
| 410 |
-
"step": 33500
|
| 411 |
-
},
|
| 412 |
-
{
|
| 413 |
-
"epoch": 0.51,
|
| 414 |
-
"learning_rate": 1.6628088026737277e-05,
|
| 415 |
-
"loss": 3.2513,
|
| 416 |
-
"step": 34000
|
| 417 |
-
},
|
| 418 |
-
{
|
| 419 |
-
"epoch": 0.51,
|
| 420 |
-
"learning_rate": 1.6578501085954003e-05,
|
| 421 |
-
"loss": 3.2584,
|
| 422 |
-
"step": 34500
|
| 423 |
-
},
|
| 424 |
-
{
|
| 425 |
-
"epoch": 0.52,
|
| 426 |
-
"learning_rate": 1.652891414517073e-05,
|
| 427 |
-
"loss": 3.2576,
|
| 428 |
-
"step": 35000
|
| 429 |
-
},
|
| 430 |
-
{
|
| 431 |
-
"epoch": 0.53,
|
| 432 |
-
"learning_rate": 1.6479327204387455e-05,
|
| 433 |
-
"loss": 3.2532,
|
| 434 |
-
"step": 35500
|
| 435 |
-
},
|
| 436 |
-
{
|
| 437 |
-
"epoch": 0.54,
|
| 438 |
-
"learning_rate": 1.642974026360418e-05,
|
| 439 |
-
"loss": 3.2349,
|
| 440 |
-
"step": 36000
|
| 441 |
-
},
|
| 442 |
-
{
|
| 443 |
-
"epoch": 0.54,
|
| 444 |
-
"learning_rate": 1.6380153322820903e-05,
|
| 445 |
-
"loss": 3.2349,
|
| 446 |
-
"step": 36500
|
| 447 |
-
},
|
| 448 |
-
{
|
| 449 |
-
"epoch": 0.55,
|
| 450 |
-
"learning_rate": 1.6330566382037625e-05,
|
| 451 |
-
"loss": 3.2158,
|
| 452 |
-
"step": 37000
|
| 453 |
-
},
|
| 454 |
-
{
|
| 455 |
-
"epoch": 0.56,
|
| 456 |
-
"learning_rate": 1.628097944125435e-05,
|
| 457 |
-
"loss": 3.2309,
|
| 458 |
-
"step": 37500
|
| 459 |
-
},
|
| 460 |
-
{
|
| 461 |
-
"epoch": 0.57,
|
| 462 |
-
"learning_rate": 1.6231392500471077e-05,
|
| 463 |
-
"loss": 3.2227,
|
| 464 |
-
"step": 38000
|
| 465 |
-
},
|
| 466 |
-
{
|
| 467 |
-
"epoch": 0.57,
|
| 468 |
-
"learning_rate": 1.6181805559687803e-05,
|
| 469 |
-
"loss": 3.2134,
|
| 470 |
-
"step": 38500
|
| 471 |
-
},
|
| 472 |
-
{
|
| 473 |
-
"epoch": 0.58,
|
| 474 |
-
"learning_rate": 1.613221861890453e-05,
|
| 475 |
-
"loss": 3.2206,
|
| 476 |
-
"step": 39000
|
| 477 |
-
},
|
| 478 |
-
{
|
| 479 |
-
"epoch": 0.59,
|
| 480 |
-
"learning_rate": 1.608263167812125e-05,
|
| 481 |
-
"loss": 3.2002,
|
| 482 |
-
"step": 39500
|
| 483 |
-
},
|
| 484 |
-
{
|
| 485 |
-
"epoch": 0.6,
|
| 486 |
-
"learning_rate": 1.6033044737337973e-05,
|
| 487 |
-
"loss": 3.1988,
|
| 488 |
-
"step": 40000
|
| 489 |
-
},
|
| 490 |
-
{
|
| 491 |
-
"epoch": 0.6,
|
| 492 |
-
"learning_rate": 1.59834577965547e-05,
|
| 493 |
-
"loss": 3.2081,
|
| 494 |
-
"step": 40500
|
| 495 |
-
},
|
| 496 |
-
{
|
| 497 |
-
"epoch": 0.61,
|
| 498 |
-
"learning_rate": 1.5933870855771425e-05,
|
| 499 |
-
"loss": 3.1891,
|
| 500 |
-
"step": 41000
|
| 501 |
-
},
|
| 502 |
-
{
|
| 503 |
-
"epoch": 0.62,
|
| 504 |
-
"learning_rate": 1.588428391498815e-05,
|
| 505 |
-
"loss": 3.2007,
|
| 506 |
-
"step": 41500
|
| 507 |
-
},
|
| 508 |
-
{
|
| 509 |
-
"epoch": 0.62,
|
| 510 |
-
"learning_rate": 1.5834696974204877e-05,
|
| 511 |
-
"loss": 3.1948,
|
| 512 |
-
"step": 42000
|
| 513 |
-
},
|
| 514 |
-
{
|
| 515 |
-
"epoch": 0.63,
|
| 516 |
-
"learning_rate": 1.57851100334216e-05,
|
| 517 |
-
"loss": 3.1673,
|
| 518 |
-
"step": 42500
|
| 519 |
-
},
|
| 520 |
-
{
|
| 521 |
-
"epoch": 0.64,
|
| 522 |
-
"learning_rate": 1.5735523092638325e-05,
|
| 523 |
-
"loss": 3.158,
|
| 524 |
-
"step": 43000
|
| 525 |
-
},
|
| 526 |
-
{
|
| 527 |
-
"epoch": 0.65,
|
| 528 |
-
"learning_rate": 1.5685936151855047e-05,
|
| 529 |
-
"loss": 3.1561,
|
| 530 |
-
"step": 43500
|
| 531 |
-
},
|
| 532 |
-
{
|
| 533 |
-
"epoch": 0.65,
|
| 534 |
-
"learning_rate": 1.5636349211071773e-05,
|
| 535 |
-
"loss": 3.1734,
|
| 536 |
-
"step": 44000
|
| 537 |
-
},
|
| 538 |
-
{
|
| 539 |
-
"epoch": 0.66,
|
| 540 |
-
"learning_rate": 1.55867622702885e-05,
|
| 541 |
-
"loss": 3.1401,
|
| 542 |
-
"step": 44500
|
| 543 |
-
},
|
| 544 |
-
{
|
| 545 |
-
"epoch": 0.67,
|
| 546 |
-
"learning_rate": 1.5537175329505225e-05,
|
| 547 |
-
"loss": 3.1463,
|
| 548 |
-
"step": 45000
|
| 549 |
-
},
|
| 550 |
-
{
|
| 551 |
-
"epoch": 0.68,
|
| 552 |
-
"learning_rate": 1.5487588388721947e-05,
|
| 553 |
-
"loss": 3.1431,
|
| 554 |
-
"step": 45500
|
| 555 |
-
},
|
| 556 |
-
{
|
| 557 |
-
"epoch": 0.68,
|
| 558 |
-
"learning_rate": 1.5438001447938673e-05,
|
| 559 |
-
"loss": 3.1316,
|
| 560 |
-
"step": 46000
|
| 561 |
-
},
|
| 562 |
-
{
|
| 563 |
-
"epoch": 0.69,
|
| 564 |
-
"learning_rate": 1.5388414507155395e-05,
|
| 565 |
-
"loss": 3.1606,
|
| 566 |
-
"step": 46500
|
| 567 |
-
},
|
| 568 |
-
{
|
| 569 |
-
"epoch": 0.7,
|
| 570 |
-
"learning_rate": 1.533882756637212e-05,
|
| 571 |
-
"loss": 3.1362,
|
| 572 |
-
"step": 47000
|
| 573 |
-
},
|
| 574 |
-
{
|
| 575 |
-
"epoch": 0.71,
|
| 576 |
-
"learning_rate": 1.5289240625588847e-05,
|
| 577 |
-
"loss": 3.1335,
|
| 578 |
-
"step": 47500
|
| 579 |
-
},
|
| 580 |
-
{
|
| 581 |
-
"epoch": 0.71,
|
| 582 |
-
"learning_rate": 1.523965368480557e-05,
|
| 583 |
-
"loss": 3.149,
|
| 584 |
-
"step": 48000
|
| 585 |
-
},
|
| 586 |
-
{
|
| 587 |
-
"epoch": 0.72,
|
| 588 |
-
"learning_rate": 1.5190066744022297e-05,
|
| 589 |
-
"loss": 3.1293,
|
| 590 |
-
"step": 48500
|
| 591 |
-
},
|
| 592 |
-
{
|
| 593 |
-
"epoch": 0.73,
|
| 594 |
-
"learning_rate": 1.514047980323902e-05,
|
| 595 |
-
"loss": 3.1286,
|
| 596 |
-
"step": 49000
|
| 597 |
-
},
|
| 598 |
-
{
|
| 599 |
-
"epoch": 0.74,
|
| 600 |
-
"learning_rate": 1.5090892862455743e-05,
|
| 601 |
-
"loss": 3.1196,
|
| 602 |
-
"step": 49500
|
| 603 |
-
},
|
| 604 |
-
{
|
| 605 |
-
"epoch": 0.74,
|
| 606 |
-
"learning_rate": 1.5041305921672469e-05,
|
| 607 |
-
"loss": 3.1238,
|
| 608 |
-
"step": 50000
|
| 609 |
-
},
|
| 610 |
-
{
|
| 611 |
-
"epoch": 0.75,
|
| 612 |
-
"learning_rate": 1.4991718980889195e-05,
|
| 613 |
-
"loss": 3.1033,
|
| 614 |
-
"step": 50500
|
| 615 |
-
},
|
| 616 |
-
{
|
| 617 |
-
"epoch": 0.76,
|
| 618 |
-
"learning_rate": 1.4942132040105919e-05,
|
| 619 |
-
"loss": 3.1112,
|
| 620 |
-
"step": 51000
|
| 621 |
-
},
|
| 622 |
-
{
|
| 623 |
-
"epoch": 0.77,
|
| 624 |
-
"learning_rate": 1.4892545099322645e-05,
|
| 625 |
-
"loss": 3.0936,
|
| 626 |
-
"step": 51500
|
| 627 |
-
},
|
| 628 |
-
{
|
| 629 |
-
"epoch": 0.77,
|
| 630 |
-
"learning_rate": 1.4842958158539369e-05,
|
| 631 |
-
"loss": 3.107,
|
| 632 |
-
"step": 52000
|
| 633 |
-
},
|
| 634 |
-
{
|
| 635 |
-
"epoch": 0.78,
|
| 636 |
-
"learning_rate": 1.4793371217756094e-05,
|
| 637 |
-
"loss": 3.1063,
|
| 638 |
-
"step": 52500
|
| 639 |
-
},
|
| 640 |
-
{
|
| 641 |
-
"epoch": 0.79,
|
| 642 |
-
"learning_rate": 1.4743784276972817e-05,
|
| 643 |
-
"loss": 3.0639,
|
| 644 |
-
"step": 53000
|
| 645 |
-
},
|
| 646 |
-
{
|
| 647 |
-
"epoch": 0.8,
|
| 648 |
-
"learning_rate": 1.4694197336189543e-05,
|
| 649 |
-
"loss": 3.1028,
|
| 650 |
-
"step": 53500
|
| 651 |
-
},
|
| 652 |
-
{
|
| 653 |
-
"epoch": 0.8,
|
| 654 |
-
"learning_rate": 1.4644610395406267e-05,
|
| 655 |
-
"loss": 3.0821,
|
| 656 |
-
"step": 54000
|
| 657 |
-
},
|
| 658 |
-
{
|
| 659 |
-
"epoch": 0.81,
|
| 660 |
-
"learning_rate": 1.4595023454622992e-05,
|
| 661 |
-
"loss": 3.0596,
|
| 662 |
-
"step": 54500
|
| 663 |
-
},
|
| 664 |
-
{
|
| 665 |
-
"epoch": 0.82,
|
| 666 |
-
"learning_rate": 1.4545436513839717e-05,
|
| 667 |
-
"loss": 3.0787,
|
| 668 |
-
"step": 55000
|
| 669 |
-
},
|
| 670 |
-
{
|
| 671 |
-
"epoch": 0.83,
|
| 672 |
-
"learning_rate": 1.4495849573056442e-05,
|
| 673 |
-
"loss": 3.0755,
|
| 674 |
-
"step": 55500
|
| 675 |
-
},
|
| 676 |
-
{
|
| 677 |
-
"epoch": 0.83,
|
| 678 |
-
"learning_rate": 1.4446262632273165e-05,
|
| 679 |
-
"loss": 3.066,
|
| 680 |
-
"step": 56000
|
| 681 |
-
},
|
| 682 |
-
{
|
| 683 |
-
"epoch": 0.84,
|
| 684 |
-
"learning_rate": 1.439667569148989e-05,
|
| 685 |
-
"loss": 3.0695,
|
| 686 |
-
"step": 56500
|
| 687 |
-
},
|
| 688 |
-
{
|
| 689 |
-
"epoch": 0.85,
|
| 690 |
-
"learning_rate": 1.4347088750706615e-05,
|
| 691 |
-
"loss": 3.059,
|
| 692 |
-
"step": 57000
|
| 693 |
-
},
|
| 694 |
-
{
|
| 695 |
-
"epoch": 0.86,
|
| 696 |
-
"learning_rate": 1.429750180992334e-05,
|
| 697 |
-
"loss": 3.0628,
|
| 698 |
-
"step": 57500
|
| 699 |
-
},
|
| 700 |
-
{
|
| 701 |
-
"epoch": 0.86,
|
| 702 |
-
"learning_rate": 1.4247914869140065e-05,
|
| 703 |
-
"loss": 3.0733,
|
| 704 |
-
"step": 58000
|
| 705 |
-
},
|
| 706 |
-
{
|
| 707 |
-
"epoch": 0.87,
|
| 708 |
-
"learning_rate": 1.419832792835679e-05,
|
| 709 |
-
"loss": 3.0591,
|
| 710 |
-
"step": 58500
|
| 711 |
-
},
|
| 712 |
-
{
|
| 713 |
-
"epoch": 0.88,
|
| 714 |
-
"learning_rate": 1.4148740987573514e-05,
|
| 715 |
-
"loss": 3.0468,
|
| 716 |
-
"step": 59000
|
| 717 |
-
},
|
| 718 |
-
{
|
| 719 |
-
"epoch": 0.89,
|
| 720 |
-
"learning_rate": 1.4099154046790237e-05,
|
| 721 |
-
"loss": 3.0265,
|
| 722 |
-
"step": 59500
|
| 723 |
-
},
|
| 724 |
-
{
|
| 725 |
-
"epoch": 0.89,
|
| 726 |
-
"learning_rate": 1.4049567106006963e-05,
|
| 727 |
-
"loss": 3.0282,
|
| 728 |
-
"step": 60000
|
| 729 |
-
},
|
| 730 |
-
{
|
| 731 |
-
"epoch": 0.9,
|
| 732 |
-
"learning_rate": 1.3999980165223688e-05,
|
| 733 |
-
"loss": 3.0222,
|
| 734 |
-
"step": 60500
|
| 735 |
-
},
|
| 736 |
-
{
|
| 737 |
-
"epoch": 0.91,
|
| 738 |
-
"learning_rate": 1.3950393224440413e-05,
|
| 739 |
-
"loss": 3.0275,
|
| 740 |
-
"step": 61000
|
| 741 |
-
},
|
| 742 |
-
{
|
| 743 |
-
"epoch": 0.91,
|
| 744 |
-
"learning_rate": 1.3900806283657138e-05,
|
| 745 |
-
"loss": 3.0277,
|
| 746 |
-
"step": 61500
|
| 747 |
-
},
|
| 748 |
-
{
|
| 749 |
-
"epoch": 0.92,
|
| 750 |
-
"learning_rate": 1.3851219342873862e-05,
|
| 751 |
-
"loss": 3.0551,
|
| 752 |
-
"step": 62000
|
| 753 |
-
},
|
| 754 |
-
{
|
| 755 |
-
"epoch": 0.93,
|
| 756 |
-
"learning_rate": 1.3801632402090585e-05,
|
| 757 |
-
"loss": 3.0205,
|
| 758 |
-
"step": 62500
|
| 759 |
-
},
|
| 760 |
-
{
|
| 761 |
-
"epoch": 0.94,
|
| 762 |
-
"learning_rate": 1.375204546130731e-05,
|
| 763 |
-
"loss": 3.023,
|
| 764 |
-
"step": 63000
|
| 765 |
-
},
|
| 766 |
-
{
|
| 767 |
-
"epoch": 0.94,
|
| 768 |
-
"learning_rate": 1.3702458520524036e-05,
|
| 769 |
-
"loss": 3.0244,
|
| 770 |
-
"step": 63500
|
| 771 |
-
},
|
| 772 |
-
{
|
| 773 |
-
"epoch": 0.95,
|
| 774 |
-
"learning_rate": 1.365287157974076e-05,
|
| 775 |
-
"loss": 3.0116,
|
| 776 |
-
"step": 64000
|
| 777 |
-
},
|
| 778 |
-
{
|
| 779 |
-
"epoch": 0.96,
|
| 780 |
-
"learning_rate": 1.3603284638957486e-05,
|
| 781 |
-
"loss": 3.0141,
|
| 782 |
-
"step": 64500
|
| 783 |
-
},
|
| 784 |
-
{
|
| 785 |
-
"epoch": 0.97,
|
| 786 |
-
"learning_rate": 1.355369769817421e-05,
|
| 787 |
-
"loss": 3.0284,
|
| 788 |
-
"step": 65000
|
| 789 |
-
},
|
| 790 |
-
{
|
| 791 |
-
"epoch": 0.97,
|
| 792 |
-
"learning_rate": 1.3504110757390933e-05,
|
| 793 |
-
"loss": 3.0236,
|
| 794 |
-
"step": 65500
|
| 795 |
-
},
|
| 796 |
-
{
|
| 797 |
-
"epoch": 0.98,
|
| 798 |
-
"learning_rate": 1.3454523816607659e-05,
|
| 799 |
-
"loss": 3.013,
|
| 800 |
-
"step": 66000
|
| 801 |
-
},
|
| 802 |
-
{
|
| 803 |
-
"epoch": 0.99,
|
| 804 |
-
"learning_rate": 1.3404936875824384e-05,
|
| 805 |
-
"loss": 3.0027,
|
| 806 |
-
"step": 66500
|
| 807 |
-
},
|
| 808 |
-
{
|
| 809 |
-
"epoch": 1.0,
|
| 810 |
-
"learning_rate": 1.3355349935041108e-05,
|
| 811 |
-
"loss": 3.0155,
|
| 812 |
-
"step": 67000
|
| 813 |
-
},
|
| 814 |
-
{
|
| 815 |
-
"epoch": 1.0,
|
| 816 |
-
"eval_bleu": 11.298551127218651,
|
| 817 |
-
"eval_loss": 2.3749005794525146,
|
| 818 |
-
"eval_runtime": 4929.9601,
|
| 819 |
-
"eval_samples_per_second": 8.201,
|
| 820 |
-
"eval_steps_per_second": 0.513,
|
| 821 |
-
"step": 67222
|
| 822 |
-
},
|
| 823 |
-
{
|
| 824 |
-
"epoch": 1.0,
|
| 825 |
-
"learning_rate": 1.3305762994257834e-05,
|
| 826 |
-
"loss": 3.0195,
|
| 827 |
-
"step": 67500
|
| 828 |
-
},
|
| 829 |
-
{
|
| 830 |
-
"epoch": 1.01,
|
| 831 |
-
"learning_rate": 1.3256176053474558e-05,
|
| 832 |
-
"loss": 2.9924,
|
| 833 |
-
"step": 68000
|
| 834 |
-
},
|
| 835 |
-
{
|
| 836 |
-
"epoch": 1.02,
|
| 837 |
-
"learning_rate": 1.3206589112691284e-05,
|
| 838 |
-
"loss": 2.997,
|
| 839 |
-
"step": 68500
|
| 840 |
-
},
|
| 841 |
-
{
|
| 842 |
-
"epoch": 1.03,
|
| 843 |
-
"learning_rate": 1.3157002171908007e-05,
|
| 844 |
-
"loss": 2.9694,
|
| 845 |
-
"step": 69000
|
| 846 |
-
},
|
| 847 |
-
{
|
| 848 |
-
"epoch": 1.03,
|
| 849 |
-
"learning_rate": 1.3107415231124732e-05,
|
| 850 |
-
"loss": 2.9804,
|
| 851 |
-
"step": 69500
|
| 852 |
-
},
|
| 853 |
-
{
|
| 854 |
-
"epoch": 1.04,
|
| 855 |
-
"learning_rate": 1.3057828290341456e-05,
|
| 856 |
-
"loss": 2.9879,
|
| 857 |
-
"step": 70000
|
| 858 |
-
},
|
| 859 |
-
{
|
| 860 |
-
"epoch": 1.05,
|
| 861 |
-
"learning_rate": 1.3008241349558182e-05,
|
| 862 |
-
"loss": 2.9919,
|
| 863 |
-
"step": 70500
|
| 864 |
-
},
|
| 865 |
-
{
|
| 866 |
-
"epoch": 1.06,
|
| 867 |
-
"learning_rate": 1.2958654408774906e-05,
|
| 868 |
-
"loss": 2.9875,
|
| 869 |
-
"step": 71000
|
| 870 |
-
},
|
| 871 |
-
{
|
| 872 |
-
"epoch": 1.06,
|
| 873 |
-
"learning_rate": 1.2909067467991632e-05,
|
| 874 |
-
"loss": 2.9912,
|
| 875 |
-
"step": 71500
|
| 876 |
-
},
|
| 877 |
-
{
|
| 878 |
-
"epoch": 1.07,
|
| 879 |
-
"learning_rate": 1.2859480527208354e-05,
|
| 880 |
-
"loss": 2.974,
|
| 881 |
-
"step": 72000
|
| 882 |
-
},
|
| 883 |
-
{
|
| 884 |
-
"epoch": 1.08,
|
| 885 |
-
"learning_rate": 1.280989358642508e-05,
|
| 886 |
-
"loss": 2.9581,
|
| 887 |
-
"step": 72500
|
| 888 |
-
},
|
| 889 |
-
{
|
| 890 |
-
"epoch": 1.09,
|
| 891 |
-
"learning_rate": 1.2760306645641804e-05,
|
| 892 |
-
"loss": 2.975,
|
| 893 |
-
"step": 73000
|
| 894 |
-
},
|
| 895 |
-
{
|
| 896 |
-
"epoch": 1.09,
|
| 897 |
-
"learning_rate": 1.271071970485853e-05,
|
| 898 |
-
"loss": 2.9737,
|
| 899 |
-
"step": 73500
|
| 900 |
-
},
|
| 901 |
-
{
|
| 902 |
-
"epoch": 1.1,
|
| 903 |
-
"learning_rate": 1.2661132764075254e-05,
|
| 904 |
-
"loss": 2.9722,
|
| 905 |
-
"step": 74000
|
| 906 |
-
},
|
| 907 |
-
{
|
| 908 |
-
"epoch": 1.11,
|
| 909 |
-
"learning_rate": 1.261154582329198e-05,
|
| 910 |
-
"loss": 2.9727,
|
| 911 |
-
"step": 74500
|
| 912 |
-
},
|
| 913 |
-
{
|
| 914 |
-
"epoch": 1.12,
|
| 915 |
-
"learning_rate": 1.2561958882508702e-05,
|
| 916 |
-
"loss": 2.9618,
|
| 917 |
-
"step": 75000
|
| 918 |
-
},
|
| 919 |
-
{
|
| 920 |
-
"epoch": 1.12,
|
| 921 |
-
"learning_rate": 1.2512371941725428e-05,
|
| 922 |
-
"loss": 2.9554,
|
| 923 |
-
"step": 75500
|
| 924 |
-
},
|
| 925 |
-
{
|
| 926 |
-
"epoch": 1.13,
|
| 927 |
-
"learning_rate": 1.2462785000942152e-05,
|
| 928 |
-
"loss": 2.961,
|
| 929 |
-
"step": 76000
|
| 930 |
-
},
|
| 931 |
-
{
|
| 932 |
-
"epoch": 1.14,
|
| 933 |
-
"learning_rate": 1.2413198060158878e-05,
|
| 934 |
-
"loss": 2.9627,
|
| 935 |
-
"step": 76500
|
| 936 |
-
},
|
| 937 |
-
{
|
| 938 |
-
"epoch": 1.15,
|
| 939 |
-
"learning_rate": 1.2363611119375602e-05,
|
| 940 |
-
"loss": 2.9896,
|
| 941 |
-
"step": 77000
|
| 942 |
-
},
|
| 943 |
-
{
|
| 944 |
-
"epoch": 1.15,
|
| 945 |
-
"learning_rate": 1.2314024178592328e-05,
|
| 946 |
-
"loss": 2.9433,
|
| 947 |
-
"step": 77500
|
| 948 |
-
},
|
| 949 |
-
{
|
| 950 |
-
"epoch": 1.16,
|
| 951 |
-
"learning_rate": 1.2264437237809052e-05,
|
| 952 |
-
"loss": 2.9329,
|
| 953 |
-
"step": 78000
|
| 954 |
-
},
|
| 955 |
-
{
|
| 956 |
-
"epoch": 1.17,
|
| 957 |
-
"learning_rate": 1.2214850297025776e-05,
|
| 958 |
-
"loss": 2.9552,
|
| 959 |
-
"step": 78500
|
| 960 |
-
},
|
| 961 |
-
{
|
| 962 |
-
"epoch": 1.18,
|
| 963 |
-
"learning_rate": 1.21652633562425e-05,
|
| 964 |
-
"loss": 2.9382,
|
| 965 |
-
"step": 79000
|
| 966 |
-
},
|
| 967 |
-
{
|
| 968 |
-
"epoch": 1.18,
|
| 969 |
-
"learning_rate": 1.2115676415459226e-05,
|
| 970 |
-
"loss": 2.9629,
|
| 971 |
-
"step": 79500
|
| 972 |
-
},
|
| 973 |
-
{
|
| 974 |
-
"epoch": 1.19,
|
| 975 |
-
"learning_rate": 1.206608947467595e-05,
|
| 976 |
-
"loss": 2.9555,
|
| 977 |
-
"step": 80000
|
| 978 |
-
},
|
| 979 |
-
{
|
| 980 |
-
"epoch": 1.2,
|
| 981 |
-
"learning_rate": 1.2016502533892676e-05,
|
| 982 |
-
"loss": 2.9364,
|
| 983 |
-
"step": 80500
|
| 984 |
-
},
|
| 985 |
-
{
|
| 986 |
-
"epoch": 1.2,
|
| 987 |
-
"learning_rate": 1.19669155931094e-05,
|
| 988 |
-
"loss": 2.9296,
|
| 989 |
-
"step": 81000
|
| 990 |
-
},
|
| 991 |
-
{
|
| 992 |
-
"epoch": 1.21,
|
| 993 |
-
"learning_rate": 1.1917328652326124e-05,
|
| 994 |
-
"loss": 2.9483,
|
| 995 |
-
"step": 81500
|
| 996 |
-
},
|
| 997 |
-
{
|
| 998 |
-
"epoch": 1.22,
|
| 999 |
-
"learning_rate": 1.1867741711542848e-05,
|
| 1000 |
-
"loss": 2.9605,
|
| 1001 |
-
"step": 82000
|
| 1002 |
-
},
|
| 1003 |
-
{
|
| 1004 |
-
"epoch": 1.23,
|
| 1005 |
-
"learning_rate": 1.1818154770759574e-05,
|
| 1006 |
-
"loss": 2.928,
|
| 1007 |
-
"step": 82500
|
| 1008 |
-
},
|
| 1009 |
-
{
|
| 1010 |
-
"epoch": 1.23,
|
| 1011 |
-
"learning_rate": 1.1768567829976298e-05,
|
| 1012 |
-
"loss": 2.9216,
|
| 1013 |
-
"step": 83000
|
| 1014 |
-
},
|
| 1015 |
-
{
|
| 1016 |
-
"epoch": 1.24,
|
| 1017 |
-
"learning_rate": 1.1718980889193024e-05,
|
| 1018 |
-
"loss": 2.9402,
|
| 1019 |
-
"step": 83500
|
| 1020 |
-
},
|
| 1021 |
-
{
|
| 1022 |
-
"epoch": 1.25,
|
| 1023 |
-
"learning_rate": 1.1669393948409748e-05,
|
| 1024 |
-
"loss": 2.9311,
|
| 1025 |
-
"step": 84000
|
| 1026 |
-
},
|
| 1027 |
-
{
|
| 1028 |
-
"epoch": 1.26,
|
| 1029 |
-
"learning_rate": 1.1619807007626474e-05,
|
| 1030 |
-
"loss": 2.9537,
|
| 1031 |
-
"step": 84500
|
| 1032 |
-
},
|
| 1033 |
-
{
|
| 1034 |
-
"epoch": 1.26,
|
| 1035 |
-
"learning_rate": 1.1570220066843196e-05,
|
| 1036 |
-
"loss": 2.919,
|
| 1037 |
-
"step": 85000
|
| 1038 |
-
},
|
| 1039 |
-
{
|
| 1040 |
-
"epoch": 1.27,
|
| 1041 |
-
"learning_rate": 1.1520633126059922e-05,
|
| 1042 |
-
"loss": 2.918,
|
| 1043 |
-
"step": 85500
|
| 1044 |
-
},
|
| 1045 |
-
{
|
| 1046 |
-
"epoch": 1.28,
|
| 1047 |
-
"learning_rate": 1.1471046185276646e-05,
|
| 1048 |
-
"loss": 2.9339,
|
| 1049 |
-
"step": 86000
|
| 1050 |
-
},
|
| 1051 |
-
{
|
| 1052 |
-
"epoch": 1.29,
|
| 1053 |
-
"learning_rate": 1.1421459244493372e-05,
|
| 1054 |
-
"loss": 2.9071,
|
| 1055 |
-
"step": 86500
|
| 1056 |
-
},
|
| 1057 |
-
{
|
| 1058 |
-
"epoch": 1.29,
|
| 1059 |
-
"learning_rate": 1.1371872303710096e-05,
|
| 1060 |
-
"loss": 2.9397,
|
| 1061 |
-
"step": 87000
|
| 1062 |
-
},
|
| 1063 |
-
{
|
| 1064 |
-
"epoch": 1.3,
|
| 1065 |
-
"learning_rate": 1.1322285362926822e-05,
|
| 1066 |
-
"loss": 2.9225,
|
| 1067 |
-
"step": 87500
|
| 1068 |
-
},
|
| 1069 |
-
{
|
| 1070 |
-
"epoch": 1.31,
|
| 1071 |
-
"learning_rate": 1.1272698422143544e-05,
|
| 1072 |
-
"loss": 2.9248,
|
| 1073 |
-
"step": 88000
|
| 1074 |
-
},
|
| 1075 |
-
{
|
| 1076 |
-
"epoch": 1.32,
|
| 1077 |
-
"learning_rate": 1.122311148136027e-05,
|
| 1078 |
-
"loss": 2.9132,
|
| 1079 |
-
"step": 88500
|
| 1080 |
-
},
|
| 1081 |
-
{
|
| 1082 |
-
"epoch": 1.32,
|
| 1083 |
-
"learning_rate": 1.1173524540576994e-05,
|
| 1084 |
-
"loss": 2.8945,
|
| 1085 |
-
"step": 89000
|
| 1086 |
-
},
|
| 1087 |
-
{
|
| 1088 |
-
"epoch": 1.33,
|
| 1089 |
-
"learning_rate": 1.112393759979372e-05,
|
| 1090 |
-
"loss": 2.905,
|
| 1091 |
-
"step": 89500
|
| 1092 |
-
},
|
| 1093 |
-
{
|
| 1094 |
-
"epoch": 1.34,
|
| 1095 |
-
"learning_rate": 1.1074350659010444e-05,
|
| 1096 |
-
"loss": 2.9256,
|
| 1097 |
-
"step": 90000
|
| 1098 |
-
},
|
| 1099 |
-
{
|
| 1100 |
-
"epoch": 1.35,
|
| 1101 |
-
"learning_rate": 1.102476371822717e-05,
|
| 1102 |
-
"loss": 2.9089,
|
| 1103 |
-
"step": 90500
|
| 1104 |
-
},
|
| 1105 |
-
{
|
| 1106 |
-
"epoch": 1.35,
|
| 1107 |
-
"learning_rate": 1.0975176777443892e-05,
|
| 1108 |
-
"loss": 2.9104,
|
| 1109 |
-
"step": 91000
|
| 1110 |
-
},
|
| 1111 |
-
{
|
| 1112 |
-
"epoch": 1.36,
|
| 1113 |
-
"learning_rate": 1.0925589836660618e-05,
|
| 1114 |
-
"loss": 2.9226,
|
| 1115 |
-
"step": 91500
|
| 1116 |
-
},
|
| 1117 |
-
{
|
| 1118 |
-
"epoch": 1.37,
|
| 1119 |
-
"learning_rate": 1.0876002895877342e-05,
|
| 1120 |
-
"loss": 2.902,
|
| 1121 |
-
"step": 92000
|
| 1122 |
-
},
|
| 1123 |
-
{
|
| 1124 |
-
"epoch": 1.38,
|
| 1125 |
-
"learning_rate": 1.0826415955094068e-05,
|
| 1126 |
-
"loss": 2.8831,
|
| 1127 |
-
"step": 92500
|
| 1128 |
-
},
|
| 1129 |
-
{
|
| 1130 |
-
"epoch": 1.38,
|
| 1131 |
-
"learning_rate": 1.0776829014310792e-05,
|
| 1132 |
-
"loss": 2.906,
|
| 1133 |
-
"step": 93000
|
| 1134 |
-
},
|
| 1135 |
-
{
|
| 1136 |
-
"epoch": 1.39,
|
| 1137 |
-
"learning_rate": 1.0727242073527518e-05,
|
| 1138 |
-
"loss": 2.906,
|
| 1139 |
-
"step": 93500
|
| 1140 |
-
},
|
| 1141 |
-
{
|
| 1142 |
-
"epoch": 1.4,
|
| 1143 |
-
"learning_rate": 1.0677655132744242e-05,
|
| 1144 |
-
"loss": 2.8901,
|
| 1145 |
-
"step": 94000
|
| 1146 |
-
},
|
| 1147 |
-
{
|
| 1148 |
-
"epoch": 1.41,
|
| 1149 |
-
"learning_rate": 1.0628068191960966e-05,
|
| 1150 |
-
"loss": 2.9063,
|
| 1151 |
-
"step": 94500
|
| 1152 |
-
},
|
| 1153 |
-
{
|
| 1154 |
-
"epoch": 1.41,
|
| 1155 |
-
"learning_rate": 1.057848125117769e-05,
|
| 1156 |
-
"loss": 2.8765,
|
| 1157 |
-
"step": 95000
|
| 1158 |
-
},
|
| 1159 |
-
{
|
| 1160 |
-
"epoch": 1.42,
|
| 1161 |
-
"learning_rate": 1.0528894310394416e-05,
|
| 1162 |
-
"loss": 2.9022,
|
| 1163 |
-
"step": 95500
|
| 1164 |
-
},
|
| 1165 |
-
{
|
| 1166 |
-
"epoch": 1.43,
|
| 1167 |
-
"learning_rate": 1.047930736961114e-05,
|
| 1168 |
-
"loss": 2.8906,
|
| 1169 |
-
"step": 96000
|
| 1170 |
-
},
|
| 1171 |
-
{
|
| 1172 |
-
"epoch": 1.44,
|
| 1173 |
-
"learning_rate": 1.0429720428827866e-05,
|
| 1174 |
-
"loss": 2.8627,
|
| 1175 |
-
"step": 96500
|
| 1176 |
-
},
|
| 1177 |
-
{
|
| 1178 |
-
"epoch": 1.44,
|
| 1179 |
-
"learning_rate": 1.038013348804459e-05,
|
| 1180 |
-
"loss": 2.8789,
|
| 1181 |
-
"step": 97000
|
| 1182 |
-
},
|
| 1183 |
-
{
|
| 1184 |
-
"epoch": 1.45,
|
| 1185 |
-
"learning_rate": 1.0330546547261314e-05,
|
| 1186 |
-
"loss": 2.8782,
|
| 1187 |
-
"step": 97500
|
| 1188 |
-
},
|
| 1189 |
-
{
|
| 1190 |
-
"epoch": 1.46,
|
| 1191 |
-
"learning_rate": 1.0280959606478038e-05,
|
| 1192 |
-
"loss": 2.8706,
|
| 1193 |
-
"step": 98000
|
| 1194 |
-
},
|
| 1195 |
-
{
|
| 1196 |
-
"epoch": 1.47,
|
| 1197 |
-
"learning_rate": 1.0231372665694764e-05,
|
| 1198 |
-
"loss": 2.8434,
|
| 1199 |
-
"step": 98500
|
| 1200 |
-
},
|
| 1201 |
-
{
|
| 1202 |
-
"epoch": 1.47,
|
| 1203 |
-
"learning_rate": 1.0181785724911488e-05,
|
| 1204 |
-
"loss": 2.8851,
|
| 1205 |
-
"step": 99000
|
| 1206 |
-
},
|
| 1207 |
-
{
|
| 1208 |
-
"epoch": 1.48,
|
| 1209 |
-
"learning_rate": 1.0132198784128214e-05,
|
| 1210 |
-
"loss": 2.8806,
|
| 1211 |
-
"step": 99500
|
| 1212 |
-
},
|
| 1213 |
-
{
|
| 1214 |
-
"epoch": 1.49,
|
| 1215 |
-
"learning_rate": 1.0082611843344938e-05,
|
| 1216 |
-
"loss": 2.8695,
|
| 1217 |
-
"step": 100000
|
| 1218 |
-
},
|
| 1219 |
-
{
|
| 1220 |
-
"epoch": 1.5,
|
| 1221 |
-
"learning_rate": 1.0033024902561664e-05,
|
| 1222 |
-
"loss": 2.8775,
|
| 1223 |
-
"step": 100500
|
| 1224 |
-
},
|
| 1225 |
-
{
|
| 1226 |
-
"epoch": 1.5,
|
| 1227 |
-
"learning_rate": 9.983437961778388e-06,
|
| 1228 |
-
"loss": 2.8717,
|
| 1229 |
-
"step": 101000
|
| 1230 |
-
},
|
| 1231 |
-
{
|
| 1232 |
-
"epoch": 1.51,
|
| 1233 |
-
"learning_rate": 9.933851020995112e-06,
|
| 1234 |
-
"loss": 2.8616,
|
| 1235 |
-
"step": 101500
|
| 1236 |
-
},
|
| 1237 |
-
{
|
| 1238 |
-
"epoch": 1.52,
|
| 1239 |
-
"learning_rate": 9.884264080211836e-06,
|
| 1240 |
-
"loss": 2.8656,
|
| 1241 |
-
"step": 102000
|
| 1242 |
-
},
|
| 1243 |
-
{
|
| 1244 |
-
"epoch": 1.52,
|
| 1245 |
-
"learning_rate": 9.834677139428562e-06,
|
| 1246 |
-
"loss": 2.8867,
|
| 1247 |
-
"step": 102500
|
| 1248 |
-
},
|
| 1249 |
-
{
|
| 1250 |
-
"epoch": 1.53,
|
| 1251 |
-
"learning_rate": 9.785090198645286e-06,
|
| 1252 |
-
"loss": 2.8491,
|
| 1253 |
-
"step": 103000
|
| 1254 |
-
},
|
| 1255 |
-
{
|
| 1256 |
-
"epoch": 1.54,
|
| 1257 |
-
"learning_rate": 9.73550325786201e-06,
|
| 1258 |
-
"loss": 2.8716,
|
| 1259 |
-
"step": 103500
|
| 1260 |
-
},
|
| 1261 |
-
{
|
| 1262 |
-
"epoch": 1.55,
|
| 1263 |
-
"learning_rate": 9.685916317078736e-06,
|
| 1264 |
-
"loss": 2.8743,
|
| 1265 |
-
"step": 104000
|
| 1266 |
-
},
|
| 1267 |
-
{
|
| 1268 |
-
"epoch": 1.55,
|
| 1269 |
-
"learning_rate": 9.63632937629546e-06,
|
| 1270 |
-
"loss": 2.8503,
|
| 1271 |
-
"step": 104500
|
| 1272 |
-
},
|
| 1273 |
-
{
|
| 1274 |
-
"epoch": 1.56,
|
| 1275 |
-
"learning_rate": 9.586742435512184e-06,
|
| 1276 |
-
"loss": 2.8625,
|
| 1277 |
-
"step": 105000
|
| 1278 |
-
},
|
| 1279 |
-
{
|
| 1280 |
-
"epoch": 1.57,
|
| 1281 |
-
"learning_rate": 9.53715549472891e-06,
|
| 1282 |
-
"loss": 2.8237,
|
| 1283 |
-
"step": 105500
|
| 1284 |
-
},
|
| 1285 |
-
{
|
| 1286 |
-
"epoch": 1.58,
|
| 1287 |
-
"learning_rate": 9.487568553945634e-06,
|
| 1288 |
-
"loss": 2.8619,
|
| 1289 |
-
"step": 106000
|
| 1290 |
-
},
|
| 1291 |
-
{
|
| 1292 |
-
"epoch": 1.58,
|
| 1293 |
-
"learning_rate": 9.437981613162358e-06,
|
| 1294 |
-
"loss": 2.8629,
|
| 1295 |
-
"step": 106500
|
| 1296 |
-
},
|
| 1297 |
-
{
|
| 1298 |
-
"epoch": 1.59,
|
| 1299 |
-
"learning_rate": 9.388394672379084e-06,
|
| 1300 |
-
"loss": 2.8441,
|
| 1301 |
-
"step": 107000
|
| 1302 |
-
},
|
| 1303 |
-
{
|
| 1304 |
-
"epoch": 1.6,
|
| 1305 |
-
"learning_rate": 9.338807731595808e-06,
|
| 1306 |
-
"loss": 2.8569,
|
| 1307 |
-
"step": 107500
|
| 1308 |
-
},
|
| 1309 |
-
{
|
| 1310 |
-
"epoch": 1.61,
|
| 1311 |
-
"learning_rate": 9.289220790812532e-06,
|
| 1312 |
-
"loss": 2.8511,
|
| 1313 |
-
"step": 108000
|
| 1314 |
-
},
|
| 1315 |
-
{
|
| 1316 |
-
"epoch": 1.61,
|
| 1317 |
-
"learning_rate": 9.239633850029258e-06,
|
| 1318 |
-
"loss": 2.8701,
|
| 1319 |
-
"step": 108500
|
| 1320 |
-
},
|
| 1321 |
-
{
|
| 1322 |
-
"epoch": 1.62,
|
| 1323 |
-
"learning_rate": 9.190046909245982e-06,
|
| 1324 |
-
"loss": 2.8572,
|
| 1325 |
-
"step": 109000
|
| 1326 |
-
},
|
| 1327 |
-
{
|
| 1328 |
-
"epoch": 1.63,
|
| 1329 |
-
"learning_rate": 9.140459968462706e-06,
|
| 1330 |
-
"loss": 2.8673,
|
| 1331 |
-
"step": 109500
|
| 1332 |
-
},
|
| 1333 |
-
{
|
| 1334 |
-
"epoch": 1.64,
|
| 1335 |
-
"learning_rate": 9.090873027679432e-06,
|
| 1336 |
-
"loss": 2.8621,
|
| 1337 |
-
"step": 110000
|
| 1338 |
-
},
|
| 1339 |
-
{
|
| 1340 |
-
"epoch": 1.64,
|
| 1341 |
-
"learning_rate": 9.041286086896156e-06,
|
| 1342 |
-
"loss": 2.8592,
|
| 1343 |
-
"step": 110500
|
| 1344 |
-
},
|
| 1345 |
-
{
|
| 1346 |
-
"epoch": 1.65,
|
| 1347 |
-
"learning_rate": 8.99169914611288e-06,
|
| 1348 |
-
"loss": 2.8582,
|
| 1349 |
-
"step": 111000
|
| 1350 |
-
},
|
| 1351 |
-
{
|
| 1352 |
-
"epoch": 1.66,
|
| 1353 |
-
"learning_rate": 8.942112205329606e-06,
|
| 1354 |
-
"loss": 2.8666,
|
| 1355 |
-
"step": 111500
|
| 1356 |
-
},
|
| 1357 |
-
{
|
| 1358 |
-
"epoch": 1.67,
|
| 1359 |
-
"learning_rate": 8.89252526454633e-06,
|
| 1360 |
-
"loss": 2.8588,
|
| 1361 |
-
"step": 112000
|
| 1362 |
-
},
|
| 1363 |
-
{
|
| 1364 |
-
"epoch": 1.67,
|
| 1365 |
-
"learning_rate": 8.842938323763054e-06,
|
| 1366 |
-
"loss": 2.8475,
|
| 1367 |
-
"step": 112500
|
| 1368 |
-
},
|
| 1369 |
-
{
|
| 1370 |
-
"epoch": 1.68,
|
| 1371 |
-
"learning_rate": 8.79335138297978e-06,
|
| 1372 |
-
"loss": 2.8357,
|
| 1373 |
-
"step": 113000
|
| 1374 |
-
},
|
| 1375 |
-
{
|
| 1376 |
-
"epoch": 1.69,
|
| 1377 |
-
"learning_rate": 8.743764442196504e-06,
|
| 1378 |
-
"loss": 2.8608,
|
| 1379 |
-
"step": 113500
|
| 1380 |
-
},
|
| 1381 |
-
{
|
| 1382 |
-
"epoch": 1.7,
|
| 1383 |
-
"learning_rate": 8.69417750141323e-06,
|
| 1384 |
-
"loss": 2.8532,
|
| 1385 |
-
"step": 114000
|
| 1386 |
-
},
|
| 1387 |
-
{
|
| 1388 |
-
"epoch": 1.7,
|
| 1389 |
-
"learning_rate": 8.644590560629953e-06,
|
| 1390 |
-
"loss": 2.8545,
|
| 1391 |
-
"step": 114500
|
| 1392 |
-
},
|
| 1393 |
-
{
|
| 1394 |
-
"epoch": 1.71,
|
| 1395 |
-
"learning_rate": 8.595003619846678e-06,
|
| 1396 |
-
"loss": 2.8277,
|
| 1397 |
-
"step": 115000
|
| 1398 |
-
},
|
| 1399 |
-
{
|
| 1400 |
-
"epoch": 1.72,
|
| 1401 |
-
"learning_rate": 8.545416679063403e-06,
|
| 1402 |
-
"loss": 2.8509,
|
| 1403 |
-
"step": 115500
|
| 1404 |
-
},
|
| 1405 |
-
{
|
| 1406 |
-
"epoch": 1.73,
|
| 1407 |
-
"learning_rate": 8.495829738280127e-06,
|
| 1408 |
-
"loss": 2.8413,
|
| 1409 |
-
"step": 116000
|
| 1410 |
-
},
|
| 1411 |
-
{
|
| 1412 |
-
"epoch": 1.73,
|
| 1413 |
-
"learning_rate": 8.446242797496852e-06,
|
| 1414 |
-
"loss": 2.838,
|
| 1415 |
-
"step": 116500
|
| 1416 |
-
},
|
| 1417 |
-
{
|
| 1418 |
-
"epoch": 1.74,
|
| 1419 |
-
"learning_rate": 8.396655856713577e-06,
|
| 1420 |
-
"loss": 2.8543,
|
| 1421 |
-
"step": 117000
|
| 1422 |
-
},
|
| 1423 |
-
{
|
| 1424 |
-
"epoch": 1.75,
|
| 1425 |
-
"learning_rate": 8.347068915930301e-06,
|
| 1426 |
-
"loss": 2.8347,
|
| 1427 |
-
"step": 117500
|
| 1428 |
-
},
|
| 1429 |
-
{
|
| 1430 |
-
"epoch": 1.76,
|
| 1431 |
-
"learning_rate": 8.297481975147026e-06,
|
| 1432 |
-
"loss": 2.8669,
|
| 1433 |
-
"step": 118000
|
| 1434 |
-
},
|
| 1435 |
-
{
|
| 1436 |
-
"epoch": 1.76,
|
| 1437 |
-
"learning_rate": 8.247895034363751e-06,
|
| 1438 |
-
"loss": 2.8228,
|
| 1439 |
-
"step": 118500
|
| 1440 |
-
},
|
| 1441 |
-
{
|
| 1442 |
-
"epoch": 1.77,
|
| 1443 |
-
"learning_rate": 8.198308093580475e-06,
|
| 1444 |
-
"loss": 2.8385,
|
| 1445 |
-
"step": 119000
|
| 1446 |
-
},
|
| 1447 |
-
{
|
| 1448 |
-
"epoch": 1.78,
|
| 1449 |
-
"learning_rate": 8.1487211527972e-06,
|
| 1450 |
-
"loss": 2.8257,
|
| 1451 |
-
"step": 119500
|
| 1452 |
-
},
|
| 1453 |
-
{
|
| 1454 |
-
"epoch": 1.79,
|
| 1455 |
-
"learning_rate": 8.099134212013925e-06,
|
| 1456 |
-
"loss": 2.8362,
|
| 1457 |
-
"step": 120000
|
| 1458 |
-
},
|
| 1459 |
-
{
|
| 1460 |
-
"epoch": 1.79,
|
| 1461 |
-
"learning_rate": 8.04954727123065e-06,
|
| 1462 |
-
"loss": 2.8319,
|
| 1463 |
-
"step": 120500
|
| 1464 |
-
},
|
| 1465 |
-
{
|
| 1466 |
-
"epoch": 1.8,
|
| 1467 |
-
"learning_rate": 7.999960330447374e-06,
|
| 1468 |
-
"loss": 2.8356,
|
| 1469 |
-
"step": 121000
|
| 1470 |
-
},
|
| 1471 |
-
{
|
| 1472 |
-
"epoch": 1.81,
|
| 1473 |
-
"learning_rate": 7.9503733896641e-06,
|
| 1474 |
-
"loss": 2.8199,
|
| 1475 |
-
"step": 121500
|
| 1476 |
-
},
|
| 1477 |
-
{
|
| 1478 |
-
"epoch": 1.81,
|
| 1479 |
-
"learning_rate": 7.900786448880823e-06,
|
| 1480 |
-
"loss": 2.8039,
|
| 1481 |
-
"step": 122000
|
| 1482 |
-
},
|
| 1483 |
-
{
|
| 1484 |
-
"epoch": 1.82,
|
| 1485 |
-
"learning_rate": 7.851199508097548e-06,
|
| 1486 |
-
"loss": 2.832,
|
| 1487 |
-
"step": 122500
|
| 1488 |
-
},
|
| 1489 |
-
{
|
| 1490 |
-
"epoch": 1.83,
|
| 1491 |
-
"learning_rate": 7.801612567314273e-06,
|
| 1492 |
-
"loss": 2.8125,
|
| 1493 |
-
"step": 123000
|
| 1494 |
-
},
|
| 1495 |
-
{
|
| 1496 |
-
"epoch": 1.84,
|
| 1497 |
-
"learning_rate": 7.752025626530997e-06,
|
| 1498 |
-
"loss": 2.8005,
|
| 1499 |
-
"step": 123500
|
| 1500 |
-
},
|
| 1501 |
-
{
|
| 1502 |
-
"epoch": 1.84,
|
| 1503 |
-
"learning_rate": 7.702438685747721e-06,
|
| 1504 |
-
"loss": 2.8402,
|
| 1505 |
-
"step": 124000
|
| 1506 |
-
},
|
| 1507 |
-
{
|
| 1508 |
-
"epoch": 1.85,
|
| 1509 |
-
"learning_rate": 7.652851744964447e-06,
|
| 1510 |
-
"loss": 2.8186,
|
| 1511 |
-
"step": 124500
|
| 1512 |
-
},
|
| 1513 |
-
{
|
| 1514 |
-
"epoch": 1.86,
|
| 1515 |
-
"learning_rate": 7.603264804181172e-06,
|
| 1516 |
-
"loss": 2.8296,
|
| 1517 |
-
"step": 125000
|
| 1518 |
-
},
|
| 1519 |
-
{
|
| 1520 |
-
"epoch": 1.87,
|
| 1521 |
-
"learning_rate": 7.5536778633978955e-06,
|
| 1522 |
-
"loss": 2.8193,
|
| 1523 |
-
"step": 125500
|
| 1524 |
-
},
|
| 1525 |
-
{
|
| 1526 |
-
"epoch": 1.87,
|
| 1527 |
-
"learning_rate": 7.50409092261462e-06,
|
| 1528 |
-
"loss": 2.8093,
|
| 1529 |
-
"step": 126000
|
| 1530 |
-
},
|
| 1531 |
-
{
|
| 1532 |
-
"epoch": 1.88,
|
| 1533 |
-
"learning_rate": 7.454503981831346e-06,
|
| 1534 |
-
"loss": 2.8383,
|
| 1535 |
-
"step": 126500
|
| 1536 |
-
},
|
| 1537 |
-
{
|
| 1538 |
-
"epoch": 1.89,
|
| 1539 |
-
"learning_rate": 7.4049170410480695e-06,
|
| 1540 |
-
"loss": 2.821,
|
| 1541 |
-
"step": 127000
|
| 1542 |
-
},
|
| 1543 |
-
{
|
| 1544 |
-
"epoch": 1.9,
|
| 1545 |
-
"learning_rate": 7.355330100264794e-06,
|
| 1546 |
-
"loss": 2.7976,
|
| 1547 |
-
"step": 127500
|
| 1548 |
-
},
|
| 1549 |
-
{
|
| 1550 |
-
"epoch": 1.9,
|
| 1551 |
-
"learning_rate": 7.30574315948152e-06,
|
| 1552 |
-
"loss": 2.8183,
|
| 1553 |
-
"step": 128000
|
| 1554 |
-
},
|
| 1555 |
-
{
|
| 1556 |
-
"epoch": 1.91,
|
| 1557 |
-
"learning_rate": 7.2561562186982434e-06,
|
| 1558 |
-
"loss": 2.8089,
|
| 1559 |
-
"step": 128500
|
| 1560 |
-
},
|
| 1561 |
-
{
|
| 1562 |
-
"epoch": 1.92,
|
| 1563 |
-
"learning_rate": 7.206569277914968e-06,
|
| 1564 |
-
"loss": 2.818,
|
| 1565 |
-
"step": 129000
|
| 1566 |
-
},
|
| 1567 |
-
{
|
| 1568 |
-
"epoch": 1.93,
|
| 1569 |
-
"learning_rate": 7.156982337131694e-06,
|
| 1570 |
-
"loss": 2.8052,
|
| 1571 |
-
"step": 129500
|
| 1572 |
-
},
|
| 1573 |
-
{
|
| 1574 |
-
"epoch": 1.93,
|
| 1575 |
-
"learning_rate": 7.107395396348419e-06,
|
| 1576 |
-
"loss": 2.8183,
|
| 1577 |
-
"step": 130000
|
| 1578 |
-
},
|
| 1579 |
-
{
|
| 1580 |
-
"epoch": 1.94,
|
| 1581 |
-
"learning_rate": 7.057808455565142e-06,
|
| 1582 |
-
"loss": 2.8098,
|
| 1583 |
-
"step": 130500
|
| 1584 |
-
},
|
| 1585 |
-
{
|
| 1586 |
-
"epoch": 1.95,
|
| 1587 |
-
"learning_rate": 7.008221514781868e-06,
|
| 1588 |
-
"loss": 2.8155,
|
| 1589 |
-
"step": 131000
|
| 1590 |
-
},
|
| 1591 |
-
{
|
| 1592 |
-
"epoch": 1.96,
|
| 1593 |
-
"learning_rate": 6.958634573998593e-06,
|
| 1594 |
-
"loss": 2.8074,
|
| 1595 |
-
"step": 131500
|
| 1596 |
-
},
|
| 1597 |
-
{
|
| 1598 |
-
"epoch": 1.96,
|
| 1599 |
-
"learning_rate": 6.909047633215316e-06,
|
| 1600 |
-
"loss": 2.7913,
|
| 1601 |
-
"step": 132000
|
| 1602 |
-
},
|
| 1603 |
-
{
|
| 1604 |
-
"epoch": 1.97,
|
| 1605 |
-
"learning_rate": 6.859460692432042e-06,
|
| 1606 |
-
"loss": 2.8122,
|
| 1607 |
-
"step": 132500
|
| 1608 |
-
},
|
| 1609 |
-
{
|
| 1610 |
-
"epoch": 1.98,
|
| 1611 |
-
"learning_rate": 6.809873751648767e-06,
|
| 1612 |
-
"loss": 2.8327,
|
| 1613 |
-
"step": 133000
|
| 1614 |
-
},
|
| 1615 |
-
{
|
| 1616 |
-
"epoch": 1.99,
|
| 1617 |
-
"learning_rate": 6.76028681086549e-06,
|
| 1618 |
-
"loss": 2.7897,
|
| 1619 |
-
"step": 133500
|
| 1620 |
-
},
|
| 1621 |
-
{
|
| 1622 |
-
"epoch": 1.99,
|
| 1623 |
-
"learning_rate": 6.710699870082215e-06,
|
| 1624 |
-
"loss": 2.7777,
|
| 1625 |
-
"step": 134000
|
| 1626 |
-
},
|
| 1627 |
-
{
|
| 1628 |
-
"epoch": 2.0,
|
| 1629 |
-
"eval_bleu": 13.585366050482984,
|
| 1630 |
-
"eval_loss": 2.2518081665039062,
|
| 1631 |
-
"eval_runtime": 4182.4693,
|
| 1632 |
-
"eval_samples_per_second": 9.667,
|
| 1633 |
-
"eval_steps_per_second": 0.604,
|
| 1634 |
-
"step": 134444
|
| 1635 |
-
},
|
| 1636 |
-
{
|
| 1637 |
-
"epoch": 2.0,
|
| 1638 |
-
"learning_rate": 6.661112929298941e-06,
|
| 1639 |
-
"loss": 2.7994,
|
| 1640 |
-
"step": 134500
|
| 1641 |
-
},
|
| 1642 |
-
{
|
| 1643 |
-
"epoch": 2.01,
|
| 1644 |
-
"learning_rate": 6.611525988515664e-06,
|
| 1645 |
-
"loss": 2.8167,
|
| 1646 |
-
"step": 135000
|
| 1647 |
-
},
|
| 1648 |
-
{
|
| 1649 |
-
"epoch": 2.02,
|
| 1650 |
-
"learning_rate": 6.561939047732389e-06,
|
| 1651 |
-
"loss": 2.8123,
|
| 1652 |
-
"step": 135500
|
| 1653 |
-
},
|
| 1654 |
-
{
|
| 1655 |
-
"epoch": 2.02,
|
| 1656 |
-
"learning_rate": 6.512352106949115e-06,
|
| 1657 |
-
"loss": 2.7844,
|
| 1658 |
-
"step": 136000
|
| 1659 |
-
},
|
| 1660 |
-
{
|
| 1661 |
-
"epoch": 2.03,
|
| 1662 |
-
"learning_rate": 6.462765166165838e-06,
|
| 1663 |
-
"loss": 2.7956,
|
| 1664 |
-
"step": 136500
|
| 1665 |
-
},
|
| 1666 |
-
{
|
| 1667 |
-
"epoch": 2.04,
|
| 1668 |
-
"learning_rate": 6.413178225382563e-06,
|
| 1669 |
-
"loss": 2.7968,
|
| 1670 |
-
"step": 137000
|
| 1671 |
-
},
|
| 1672 |
-
{
|
| 1673 |
-
"epoch": 2.05,
|
| 1674 |
-
"learning_rate": 6.363591284599289e-06,
|
| 1675 |
-
"loss": 2.7916,
|
| 1676 |
-
"step": 137500
|
| 1677 |
-
},
|
| 1678 |
-
{
|
| 1679 |
-
"epoch": 2.05,
|
| 1680 |
-
"learning_rate": 6.314004343816014e-06,
|
| 1681 |
-
"loss": 2.7958,
|
| 1682 |
-
"step": 138000
|
| 1683 |
-
},
|
| 1684 |
-
{
|
| 1685 |
-
"epoch": 2.06,
|
| 1686 |
-
"learning_rate": 6.264417403032737e-06,
|
| 1687 |
-
"loss": 2.7855,
|
| 1688 |
-
"step": 138500
|
| 1689 |
-
},
|
| 1690 |
-
{
|
| 1691 |
-
"epoch": 2.07,
|
| 1692 |
-
"learning_rate": 6.214830462249463e-06,
|
| 1693 |
-
"loss": 2.7876,
|
| 1694 |
-
"step": 139000
|
| 1695 |
-
},
|
| 1696 |
-
{
|
| 1697 |
-
"epoch": 2.08,
|
| 1698 |
-
"learning_rate": 6.165243521466188e-06,
|
| 1699 |
-
"loss": 2.7724,
|
| 1700 |
-
"step": 139500
|
| 1701 |
-
},
|
| 1702 |
-
{
|
| 1703 |
-
"epoch": 2.08,
|
| 1704 |
-
"learning_rate": 6.115656580682911e-06,
|
| 1705 |
-
"loss": 2.8021,
|
| 1706 |
-
"step": 140000
|
| 1707 |
-
},
|
| 1708 |
-
{
|
| 1709 |
-
"epoch": 2.09,
|
| 1710 |
-
"learning_rate": 6.066069639899637e-06,
|
| 1711 |
-
"loss": 2.8024,
|
| 1712 |
-
"step": 140500
|
| 1713 |
-
},
|
| 1714 |
-
{
|
| 1715 |
-
"epoch": 2.1,
|
| 1716 |
-
"learning_rate": 6.016482699116362e-06,
|
| 1717 |
-
"loss": 2.7891,
|
| 1718 |
-
"step": 141000
|
| 1719 |
-
},
|
| 1720 |
-
{
|
| 1721 |
-
"epoch": 2.1,
|
| 1722 |
-
"learning_rate": 5.966895758333085e-06,
|
| 1723 |
-
"loss": 2.7592,
|
| 1724 |
-
"step": 141500
|
| 1725 |
-
},
|
| 1726 |
-
{
|
| 1727 |
-
"epoch": 2.11,
|
| 1728 |
-
"learning_rate": 5.917308817549811e-06,
|
| 1729 |
-
"loss": 2.7895,
|
| 1730 |
-
"step": 142000
|
| 1731 |
-
},
|
| 1732 |
-
{
|
| 1733 |
-
"epoch": 2.12,
|
| 1734 |
-
"learning_rate": 5.867721876766536e-06,
|
| 1735 |
-
"loss": 2.8106,
|
| 1736 |
-
"step": 142500
|
| 1737 |
-
},
|
| 1738 |
-
{
|
| 1739 |
-
"epoch": 2.13,
|
| 1740 |
-
"learning_rate": 5.818134935983259e-06,
|
| 1741 |
-
"loss": 2.7985,
|
| 1742 |
-
"step": 143000
|
| 1743 |
-
},
|
| 1744 |
-
{
|
| 1745 |
-
"epoch": 2.13,
|
| 1746 |
-
"learning_rate": 5.768547995199985e-06,
|
| 1747 |
-
"loss": 2.8137,
|
| 1748 |
-
"step": 143500
|
| 1749 |
-
},
|
| 1750 |
-
{
|
| 1751 |
-
"epoch": 2.14,
|
| 1752 |
-
"learning_rate": 5.71896105441671e-06,
|
| 1753 |
-
"loss": 2.7824,
|
| 1754 |
-
"step": 144000
|
| 1755 |
-
},
|
| 1756 |
-
{
|
| 1757 |
-
"epoch": 2.15,
|
| 1758 |
-
"learning_rate": 5.669374113633433e-06,
|
| 1759 |
-
"loss": 2.7878,
|
| 1760 |
-
"step": 144500
|
| 1761 |
-
},
|
| 1762 |
-
{
|
| 1763 |
-
"epoch": 2.16,
|
| 1764 |
-
"learning_rate": 5.619787172850158e-06,
|
| 1765 |
-
"loss": 2.7861,
|
| 1766 |
-
"step": 145000
|
| 1767 |
-
},
|
| 1768 |
-
{
|
| 1769 |
-
"epoch": 2.16,
|
| 1770 |
-
"learning_rate": 5.570200232066884e-06,
|
| 1771 |
-
"loss": 2.7886,
|
| 1772 |
-
"step": 145500
|
| 1773 |
-
},
|
| 1774 |
-
{
|
| 1775 |
-
"epoch": 2.17,
|
| 1776 |
-
"learning_rate": 5.520613291283607e-06,
|
| 1777 |
-
"loss": 2.7798,
|
| 1778 |
-
"step": 146000
|
| 1779 |
-
},
|
| 1780 |
-
{
|
| 1781 |
-
"epoch": 2.18,
|
| 1782 |
-
"learning_rate": 5.471026350500332e-06,
|
| 1783 |
-
"loss": 2.8015,
|
| 1784 |
-
"step": 146500
|
| 1785 |
-
},
|
| 1786 |
-
{
|
| 1787 |
-
"epoch": 2.19,
|
| 1788 |
-
"learning_rate": 5.421439409717058e-06,
|
| 1789 |
-
"loss": 2.8013,
|
| 1790 |
-
"step": 147000
|
| 1791 |
-
},
|
| 1792 |
-
{
|
| 1793 |
-
"epoch": 2.19,
|
| 1794 |
-
"learning_rate": 5.371852468933783e-06,
|
| 1795 |
-
"loss": 2.7567,
|
| 1796 |
-
"step": 147500
|
| 1797 |
-
},
|
| 1798 |
-
{
|
| 1799 |
-
"epoch": 2.2,
|
| 1800 |
-
"learning_rate": 5.322265528150506e-06,
|
| 1801 |
-
"loss": 2.8004,
|
| 1802 |
-
"step": 148000
|
| 1803 |
-
},
|
| 1804 |
-
{
|
| 1805 |
-
"epoch": 2.21,
|
| 1806 |
-
"learning_rate": 5.272678587367232e-06,
|
| 1807 |
-
"loss": 2.7817,
|
| 1808 |
-
"step": 148500
|
| 1809 |
-
},
|
| 1810 |
-
{
|
| 1811 |
-
"epoch": 2.22,
|
| 1812 |
-
"learning_rate": 5.223091646583957e-06,
|
| 1813 |
-
"loss": 2.7674,
|
| 1814 |
-
"step": 149000
|
| 1815 |
-
},
|
| 1816 |
-
{
|
| 1817 |
-
"epoch": 2.22,
|
| 1818 |
-
"learning_rate": 5.17350470580068e-06,
|
| 1819 |
-
"loss": 2.7882,
|
| 1820 |
-
"step": 149500
|
| 1821 |
-
},
|
| 1822 |
-
{
|
| 1823 |
-
"epoch": 2.23,
|
| 1824 |
-
"learning_rate": 5.123917765017406e-06,
|
| 1825 |
-
"loss": 2.7695,
|
| 1826 |
-
"step": 150000
|
| 1827 |
-
},
|
| 1828 |
-
{
|
| 1829 |
-
"epoch": 2.24,
|
| 1830 |
-
"learning_rate": 5.074330824234131e-06,
|
| 1831 |
-
"loss": 2.7708,
|
| 1832 |
-
"step": 150500
|
| 1833 |
-
},
|
| 1834 |
-
{
|
| 1835 |
-
"epoch": 2.25,
|
| 1836 |
-
"learning_rate": 5.024743883450854e-06,
|
| 1837 |
-
"loss": 2.7791,
|
| 1838 |
-
"step": 151000
|
| 1839 |
-
},
|
| 1840 |
-
{
|
| 1841 |
-
"epoch": 2.25,
|
| 1842 |
-
"learning_rate": 4.97515694266758e-06,
|
| 1843 |
-
"loss": 2.7998,
|
| 1844 |
-
"step": 151500
|
| 1845 |
-
},
|
| 1846 |
-
{
|
| 1847 |
-
"epoch": 2.26,
|
| 1848 |
-
"learning_rate": 4.925570001884304e-06,
|
| 1849 |
-
"loss": 2.7546,
|
| 1850 |
-
"step": 152000
|
| 1851 |
-
},
|
| 1852 |
-
{
|
| 1853 |
-
"epoch": 2.27,
|
| 1854 |
-
"learning_rate": 4.875983061101029e-06,
|
| 1855 |
-
"loss": 2.7576,
|
| 1856 |
-
"step": 152500
|
| 1857 |
-
},
|
| 1858 |
-
{
|
| 1859 |
-
"epoch": 2.28,
|
| 1860 |
-
"learning_rate": 4.826396120317754e-06,
|
| 1861 |
-
"loss": 2.7917,
|
| 1862 |
-
"step": 153000
|
| 1863 |
-
},
|
| 1864 |
-
{
|
| 1865 |
-
"epoch": 2.28,
|
| 1866 |
-
"learning_rate": 4.776809179534479e-06,
|
| 1867 |
-
"loss": 2.7828,
|
| 1868 |
-
"step": 153500
|
| 1869 |
-
},
|
| 1870 |
-
{
|
| 1871 |
-
"epoch": 2.29,
|
| 1872 |
-
"learning_rate": 4.727222238751203e-06,
|
| 1873 |
-
"loss": 2.7884,
|
| 1874 |
-
"step": 154000
|
| 1875 |
-
},
|
| 1876 |
-
{
|
| 1877 |
-
"epoch": 2.3,
|
| 1878 |
-
"learning_rate": 4.677635297967928e-06,
|
| 1879 |
-
"loss": 2.7807,
|
| 1880 |
-
"step": 154500
|
| 1881 |
-
},
|
| 1882 |
-
{
|
| 1883 |
-
"epoch": 2.31,
|
| 1884 |
-
"learning_rate": 4.628048357184653e-06,
|
| 1885 |
-
"loss": 2.7652,
|
| 1886 |
-
"step": 155000
|
| 1887 |
-
},
|
| 1888 |
-
{
|
| 1889 |
-
"epoch": 2.31,
|
| 1890 |
-
"learning_rate": 4.578461416401377e-06,
|
| 1891 |
-
"loss": 2.7918,
|
| 1892 |
-
"step": 155500
|
| 1893 |
-
},
|
| 1894 |
-
{
|
| 1895 |
-
"epoch": 2.32,
|
| 1896 |
-
"learning_rate": 4.528874475618102e-06,
|
| 1897 |
-
"loss": 2.7771,
|
| 1898 |
-
"step": 156000
|
| 1899 |
-
},
|
| 1900 |
-
{
|
| 1901 |
-
"epoch": 2.33,
|
| 1902 |
-
"learning_rate": 4.479287534834827e-06,
|
| 1903 |
-
"loss": 2.7561,
|
| 1904 |
-
"step": 156500
|
| 1905 |
-
},
|
| 1906 |
-
{
|
| 1907 |
-
"epoch": 2.34,
|
| 1908 |
-
"learning_rate": 4.429700594051551e-06,
|
| 1909 |
-
"loss": 2.7789,
|
| 1910 |
-
"step": 157000
|
| 1911 |
-
},
|
| 1912 |
-
{
|
| 1913 |
-
"epoch": 2.34,
|
| 1914 |
-
"learning_rate": 4.380113653268276e-06,
|
| 1915 |
-
"loss": 2.7629,
|
| 1916 |
-
"step": 157500
|
| 1917 |
-
},
|
| 1918 |
-
{
|
| 1919 |
-
"epoch": 2.35,
|
| 1920 |
-
"learning_rate": 4.330526712485001e-06,
|
| 1921 |
-
"loss": 2.7672,
|
| 1922 |
-
"step": 158000
|
| 1923 |
-
},
|
| 1924 |
-
{
|
| 1925 |
-
"epoch": 2.36,
|
| 1926 |
-
"learning_rate": 4.280939771701725e-06,
|
| 1927 |
-
"loss": 2.7643,
|
| 1928 |
-
"step": 158500
|
| 1929 |
-
},
|
| 1930 |
-
{
|
| 1931 |
-
"epoch": 2.37,
|
| 1932 |
-
"learning_rate": 4.23135283091845e-06,
|
| 1933 |
-
"loss": 2.7836,
|
| 1934 |
-
"step": 159000
|
| 1935 |
-
},
|
| 1936 |
-
{
|
| 1937 |
-
"epoch": 2.37,
|
| 1938 |
-
"learning_rate": 4.181765890135175e-06,
|
| 1939 |
-
"loss": 2.7742,
|
| 1940 |
-
"step": 159500
|
| 1941 |
-
},
|
| 1942 |
-
{
|
| 1943 |
-
"epoch": 2.38,
|
| 1944 |
-
"learning_rate": 4.132178949351899e-06,
|
| 1945 |
-
"loss": 2.7504,
|
| 1946 |
-
"step": 160000
|
| 1947 |
-
},
|
| 1948 |
-
{
|
| 1949 |
-
"epoch": 2.39,
|
| 1950 |
-
"learning_rate": 4.082592008568624e-06,
|
| 1951 |
-
"loss": 2.7738,
|
| 1952 |
-
"step": 160500
|
| 1953 |
-
},
|
| 1954 |
-
{
|
| 1955 |
-
"epoch": 2.4,
|
| 1956 |
-
"learning_rate": 4.033005067785349e-06,
|
| 1957 |
-
"loss": 2.7741,
|
| 1958 |
-
"step": 161000
|
| 1959 |
-
},
|
| 1960 |
-
{
|
| 1961 |
-
"epoch": 2.4,
|
| 1962 |
-
"learning_rate": 3.983418127002074e-06,
|
| 1963 |
-
"loss": 2.768,
|
| 1964 |
-
"step": 161500
|
| 1965 |
-
},
|
| 1966 |
-
{
|
| 1967 |
-
"epoch": 2.41,
|
| 1968 |
-
"learning_rate": 3.933831186218798e-06,
|
| 1969 |
-
"loss": 2.7874,
|
| 1970 |
-
"step": 162000
|
| 1971 |
-
},
|
| 1972 |
-
{
|
| 1973 |
-
"epoch": 2.42,
|
| 1974 |
-
"learning_rate": 3.884244245435523e-06,
|
| 1975 |
-
"loss": 2.7733,
|
| 1976 |
-
"step": 162500
|
| 1977 |
-
},
|
| 1978 |
-
{
|
| 1979 |
-
"epoch": 2.42,
|
| 1980 |
-
"learning_rate": 3.834657304652248e-06,
|
| 1981 |
-
"loss": 2.7677,
|
| 1982 |
-
"step": 163000
|
| 1983 |
-
},
|
| 1984 |
-
{
|
| 1985 |
-
"epoch": 2.43,
|
| 1986 |
-
"learning_rate": 3.7850703638689717e-06,
|
| 1987 |
-
"loss": 2.7601,
|
| 1988 |
-
"step": 163500
|
| 1989 |
-
},
|
| 1990 |
-
{
|
| 1991 |
-
"epoch": 2.44,
|
| 1992 |
-
"learning_rate": 3.735483423085696e-06,
|
| 1993 |
-
"loss": 2.7832,
|
| 1994 |
-
"step": 164000
|
| 1995 |
-
},
|
| 1996 |
-
{
|
| 1997 |
-
"epoch": 2.45,
|
| 1998 |
-
"learning_rate": 3.685896482302421e-06,
|
| 1999 |
-
"loss": 2.7704,
|
| 2000 |
-
"step": 164500
|
| 2001 |
-
},
|
| 2002 |
-
{
|
| 2003 |
-
"epoch": 2.45,
|
| 2004 |
-
"learning_rate": 3.6363095415191457e-06,
|
| 2005 |
-
"loss": 2.7698,
|
| 2006 |
-
"step": 165000
|
| 2007 |
-
},
|
| 2008 |
-
{
|
| 2009 |
-
"epoch": 2.46,
|
| 2010 |
-
"learning_rate": 3.5867226007358706e-06,
|
| 2011 |
-
"loss": 2.7724,
|
| 2012 |
-
"step": 165500
|
| 2013 |
-
},
|
| 2014 |
-
{
|
| 2015 |
-
"epoch": 2.47,
|
| 2016 |
-
"learning_rate": 3.537135659952595e-06,
|
| 2017 |
-
"loss": 2.7647,
|
| 2018 |
-
"step": 166000
|
| 2019 |
-
},
|
| 2020 |
-
{
|
| 2021 |
-
"epoch": 2.48,
|
| 2022 |
-
"learning_rate": 3.4875487191693196e-06,
|
| 2023 |
-
"loss": 2.7719,
|
| 2024 |
-
"step": 166500
|
| 2025 |
-
},
|
| 2026 |
-
{
|
| 2027 |
-
"epoch": 2.48,
|
| 2028 |
-
"learning_rate": 3.4379617783860446e-06,
|
| 2029 |
-
"loss": 2.7578,
|
| 2030 |
-
"step": 167000
|
| 2031 |
-
},
|
| 2032 |
-
{
|
| 2033 |
-
"epoch": 2.49,
|
| 2034 |
-
"learning_rate": 3.388374837602769e-06,
|
| 2035 |
-
"loss": 2.7478,
|
| 2036 |
-
"step": 167500
|
| 2037 |
-
},
|
| 2038 |
-
{
|
| 2039 |
-
"epoch": 2.5,
|
| 2040 |
-
"learning_rate": 3.3387878968194936e-06,
|
| 2041 |
-
"loss": 2.7747,
|
| 2042 |
-
"step": 168000
|
| 2043 |
-
},
|
| 2044 |
-
{
|
| 2045 |
-
"epoch": 2.51,
|
| 2046 |
-
"learning_rate": 3.2892009560362186e-06,
|
| 2047 |
-
"loss": 2.7533,
|
| 2048 |
-
"step": 168500
|
| 2049 |
-
},
|
| 2050 |
-
{
|
| 2051 |
-
"epoch": 2.51,
|
| 2052 |
-
"learning_rate": 3.239614015252943e-06,
|
| 2053 |
-
"loss": 2.7488,
|
| 2054 |
-
"step": 169000
|
| 2055 |
-
},
|
| 2056 |
-
{
|
| 2057 |
-
"epoch": 2.52,
|
| 2058 |
-
"learning_rate": 3.190027074469668e-06,
|
| 2059 |
-
"loss": 2.7577,
|
| 2060 |
-
"step": 169500
|
| 2061 |
-
},
|
| 2062 |
-
{
|
| 2063 |
-
"epoch": 2.53,
|
| 2064 |
-
"learning_rate": 3.1404401336863925e-06,
|
| 2065 |
-
"loss": 2.7646,
|
| 2066 |
-
"step": 170000
|
| 2067 |
-
},
|
| 2068 |
-
{
|
| 2069 |
-
"epoch": 2.54,
|
| 2070 |
-
"learning_rate": 3.090853192903117e-06,
|
| 2071 |
-
"loss": 2.7556,
|
| 2072 |
-
"step": 170500
|
| 2073 |
-
},
|
| 2074 |
-
{
|
| 2075 |
-
"epoch": 2.54,
|
| 2076 |
-
"learning_rate": 3.041266252119842e-06,
|
| 2077 |
-
"loss": 2.7578,
|
| 2078 |
-
"step": 171000
|
| 2079 |
-
},
|
| 2080 |
-
{
|
| 2081 |
-
"epoch": 2.55,
|
| 2082 |
-
"learning_rate": 2.9916793113365665e-06,
|
| 2083 |
-
"loss": 2.7428,
|
| 2084 |
-
"step": 171500
|
| 2085 |
-
},
|
| 2086 |
-
{
|
| 2087 |
-
"epoch": 2.56,
|
| 2088 |
-
"learning_rate": 2.942092370553291e-06,
|
| 2089 |
-
"loss": 2.7487,
|
| 2090 |
-
"step": 172000
|
| 2091 |
-
},
|
| 2092 |
-
{
|
| 2093 |
-
"epoch": 2.57,
|
| 2094 |
-
"learning_rate": 2.892505429770016e-06,
|
| 2095 |
-
"loss": 2.7457,
|
| 2096 |
-
"step": 172500
|
| 2097 |
-
},
|
| 2098 |
-
{
|
| 2099 |
-
"epoch": 2.57,
|
| 2100 |
-
"learning_rate": 2.8429184889867405e-06,
|
| 2101 |
-
"loss": 2.7366,
|
| 2102 |
-
"step": 173000
|
| 2103 |
-
},
|
| 2104 |
-
{
|
| 2105 |
-
"epoch": 2.58,
|
| 2106 |
-
"learning_rate": 2.7933315482034655e-06,
|
| 2107 |
-
"loss": 2.7497,
|
| 2108 |
-
"step": 173500
|
| 2109 |
-
},
|
| 2110 |
-
{
|
| 2111 |
-
"epoch": 2.59,
|
| 2112 |
-
"learning_rate": 2.74374460742019e-06,
|
| 2113 |
-
"loss": 2.7467,
|
| 2114 |
-
"step": 174000
|
| 2115 |
-
},
|
| 2116 |
-
{
|
| 2117 |
-
"epoch": 2.6,
|
| 2118 |
-
"learning_rate": 2.6941576666369145e-06,
|
| 2119 |
-
"loss": 2.7633,
|
| 2120 |
-
"step": 174500
|
| 2121 |
-
},
|
| 2122 |
-
{
|
| 2123 |
-
"epoch": 2.6,
|
| 2124 |
-
"learning_rate": 2.6445707258536394e-06,
|
| 2125 |
-
"loss": 2.7681,
|
| 2126 |
-
"step": 175000
|
| 2127 |
-
},
|
| 2128 |
-
{
|
| 2129 |
-
"epoch": 2.61,
|
| 2130 |
-
"learning_rate": 2.594983785070364e-06,
|
| 2131 |
-
"loss": 2.7552,
|
| 2132 |
-
"step": 175500
|
| 2133 |
-
},
|
| 2134 |
-
{
|
| 2135 |
-
"epoch": 2.62,
|
| 2136 |
-
"learning_rate": 2.5453968442870885e-06,
|
| 2137 |
-
"loss": 2.7539,
|
| 2138 |
-
"step": 176000
|
| 2139 |
-
},
|
| 2140 |
-
{
|
| 2141 |
-
"epoch": 2.63,
|
| 2142 |
-
"learning_rate": 2.4958099035038134e-06,
|
| 2143 |
-
"loss": 2.7393,
|
| 2144 |
-
"step": 176500
|
| 2145 |
-
},
|
| 2146 |
-
{
|
| 2147 |
-
"epoch": 2.63,
|
| 2148 |
-
"learning_rate": 2.446222962720538e-06,
|
| 2149 |
-
"loss": 2.7727,
|
| 2150 |
-
"step": 177000
|
| 2151 |
-
},
|
| 2152 |
-
{
|
| 2153 |
-
"epoch": 2.64,
|
| 2154 |
-
"learning_rate": 2.396636021937263e-06,
|
| 2155 |
-
"loss": 2.736,
|
| 2156 |
-
"step": 177500
|
| 2157 |
-
},
|
| 2158 |
-
{
|
| 2159 |
-
"epoch": 2.65,
|
| 2160 |
-
"learning_rate": 2.3470490811539874e-06,
|
| 2161 |
-
"loss": 2.7546,
|
| 2162 |
-
"step": 178000
|
| 2163 |
-
},
|
| 2164 |
-
{
|
| 2165 |
-
"epoch": 2.66,
|
| 2166 |
-
"learning_rate": 2.297462140370712e-06,
|
| 2167 |
-
"loss": 2.7601,
|
| 2168 |
-
"step": 178500
|
| 2169 |
-
},
|
| 2170 |
-
{
|
| 2171 |
-
"epoch": 2.66,
|
| 2172 |
-
"learning_rate": 2.247875199587437e-06,
|
| 2173 |
-
"loss": 2.7456,
|
| 2174 |
-
"step": 179000
|
| 2175 |
-
},
|
| 2176 |
-
{
|
| 2177 |
-
"epoch": 2.67,
|
| 2178 |
-
"learning_rate": 2.1982882588041614e-06,
|
| 2179 |
-
"loss": 2.76,
|
| 2180 |
-
"step": 179500
|
| 2181 |
-
},
|
| 2182 |
-
{
|
| 2183 |
-
"epoch": 2.68,
|
| 2184 |
-
"learning_rate": 2.1487013180208863e-06,
|
| 2185 |
-
"loss": 2.7396,
|
| 2186 |
-
"step": 180000
|
| 2187 |
-
},
|
| 2188 |
-
{
|
| 2189 |
-
"epoch": 2.69,
|
| 2190 |
-
"learning_rate": 2.099114377237611e-06,
|
| 2191 |
-
"loss": 2.761,
|
| 2192 |
-
"step": 180500
|
| 2193 |
-
},
|
| 2194 |
-
{
|
| 2195 |
-
"epoch": 2.69,
|
| 2196 |
-
"learning_rate": 2.0495274364543354e-06,
|
| 2197 |
-
"loss": 2.7603,
|
| 2198 |
-
"step": 181000
|
| 2199 |
-
},
|
| 2200 |
-
{
|
| 2201 |
-
"epoch": 2.7,
|
| 2202 |
-
"learning_rate": 1.9999404956710603e-06,
|
| 2203 |
-
"loss": 2.7614,
|
| 2204 |
-
"step": 181500
|
| 2205 |
-
},
|
| 2206 |
-
{
|
| 2207 |
-
"epoch": 2.71,
|
| 2208 |
-
"learning_rate": 1.950353554887785e-06,
|
| 2209 |
-
"loss": 2.7638,
|
| 2210 |
-
"step": 182000
|
| 2211 |
-
},
|
| 2212 |
-
{
|
| 2213 |
-
"epoch": 2.71,
|
| 2214 |
-
"learning_rate": 1.9007666141045096e-06,
|
| 2215 |
-
"loss": 2.7806,
|
| 2216 |
-
"step": 182500
|
| 2217 |
-
},
|
| 2218 |
-
{
|
| 2219 |
-
"epoch": 2.72,
|
| 2220 |
-
"learning_rate": 1.8511796733212343e-06,
|
| 2221 |
-
"loss": 2.7561,
|
| 2222 |
-
"step": 183000
|
| 2223 |
-
},
|
| 2224 |
-
{
|
| 2225 |
-
"epoch": 2.73,
|
| 2226 |
-
"learning_rate": 1.801592732537959e-06,
|
| 2227 |
-
"loss": 2.7473,
|
| 2228 |
-
"step": 183500
|
| 2229 |
-
},
|
| 2230 |
-
{
|
| 2231 |
-
"epoch": 2.74,
|
| 2232 |
-
"learning_rate": 1.7520057917546838e-06,
|
| 2233 |
-
"loss": 2.7405,
|
| 2234 |
-
"step": 184000
|
| 2235 |
-
},
|
| 2236 |
-
{
|
| 2237 |
-
"epoch": 2.74,
|
| 2238 |
-
"learning_rate": 1.7024188509714083e-06,
|
| 2239 |
-
"loss": 2.742,
|
| 2240 |
-
"step": 184500
|
| 2241 |
-
},
|
| 2242 |
-
{
|
| 2243 |
-
"epoch": 2.75,
|
| 2244 |
-
"learning_rate": 1.652831910188133e-06,
|
| 2245 |
-
"loss": 2.7387,
|
| 2246 |
-
"step": 185000
|
| 2247 |
-
},
|
| 2248 |
-
{
|
| 2249 |
-
"epoch": 2.76,
|
| 2250 |
-
"learning_rate": 1.6032449694048577e-06,
|
| 2251 |
-
"loss": 2.7681,
|
| 2252 |
-
"step": 185500
|
| 2253 |
-
},
|
| 2254 |
-
{
|
| 2255 |
-
"epoch": 2.77,
|
| 2256 |
-
"learning_rate": 1.5536580286215825e-06,
|
| 2257 |
-
"loss": 2.7599,
|
| 2258 |
-
"step": 186000
|
| 2259 |
-
},
|
| 2260 |
-
{
|
| 2261 |
-
"epoch": 2.77,
|
| 2262 |
-
"learning_rate": 1.504071087838307e-06,
|
| 2263 |
-
"loss": 2.7497,
|
| 2264 |
-
"step": 186500
|
| 2265 |
-
},
|
| 2266 |
-
{
|
| 2267 |
-
"epoch": 2.78,
|
| 2268 |
-
"learning_rate": 1.4544841470550317e-06,
|
| 2269 |
-
"loss": 2.7417,
|
| 2270 |
-
"step": 187000
|
| 2271 |
-
},
|
| 2272 |
-
{
|
| 2273 |
-
"epoch": 2.79,
|
| 2274 |
-
"learning_rate": 1.4048972062717565e-06,
|
| 2275 |
-
"loss": 2.739,
|
| 2276 |
-
"step": 187500
|
| 2277 |
-
},
|
| 2278 |
-
{
|
| 2279 |
-
"epoch": 2.8,
|
| 2280 |
-
"learning_rate": 1.3553102654884812e-06,
|
| 2281 |
-
"loss": 2.7633,
|
| 2282 |
-
"step": 188000
|
| 2283 |
-
},
|
| 2284 |
-
{
|
| 2285 |
-
"epoch": 2.8,
|
| 2286 |
-
"learning_rate": 1.3057233247052057e-06,
|
| 2287 |
-
"loss": 2.7516,
|
| 2288 |
-
"step": 188500
|
| 2289 |
-
},
|
| 2290 |
-
{
|
| 2291 |
-
"epoch": 2.81,
|
| 2292 |
-
"learning_rate": 1.2561363839219304e-06,
|
| 2293 |
-
"loss": 2.7638,
|
| 2294 |
-
"step": 189000
|
| 2295 |
-
},
|
| 2296 |
-
{
|
| 2297 |
-
"epoch": 2.82,
|
| 2298 |
-
"learning_rate": 1.2065494431386552e-06,
|
| 2299 |
-
"loss": 2.7397,
|
| 2300 |
-
"step": 189500
|
| 2301 |
-
},
|
| 2302 |
-
{
|
| 2303 |
-
"epoch": 2.83,
|
| 2304 |
-
"learning_rate": 1.1569625023553797e-06,
|
| 2305 |
-
"loss": 2.7614,
|
| 2306 |
-
"step": 190000
|
| 2307 |
-
},
|
| 2308 |
-
{
|
| 2309 |
-
"epoch": 2.83,
|
| 2310 |
-
"learning_rate": 1.1073755615721044e-06,
|
| 2311 |
-
"loss": 2.7235,
|
| 2312 |
-
"step": 190500
|
| 2313 |
-
},
|
| 2314 |
-
{
|
| 2315 |
-
"epoch": 2.84,
|
| 2316 |
-
"learning_rate": 1.0577886207888292e-06,
|
| 2317 |
-
"loss": 2.7401,
|
| 2318 |
-
"step": 191000
|
| 2319 |
-
},
|
| 2320 |
-
{
|
| 2321 |
-
"epoch": 2.85,
|
| 2322 |
-
"learning_rate": 1.0082016800055539e-06,
|
| 2323 |
-
"loss": 2.7573,
|
| 2324 |
-
"step": 191500
|
| 2325 |
-
},
|
| 2326 |
-
{
|
| 2327 |
-
"epoch": 2.86,
|
| 2328 |
-
"learning_rate": 9.586147392222784e-07,
|
| 2329 |
-
"loss": 2.7452,
|
| 2330 |
-
"step": 192000
|
| 2331 |
-
},
|
| 2332 |
-
{
|
| 2333 |
-
"epoch": 2.86,
|
| 2334 |
-
"learning_rate": 9.090277984390032e-07,
|
| 2335 |
-
"loss": 2.7505,
|
| 2336 |
-
"step": 192500
|
| 2337 |
-
},
|
| 2338 |
-
{
|
| 2339 |
-
"epoch": 2.87,
|
| 2340 |
-
"learning_rate": 8.594408576557279e-07,
|
| 2341 |
-
"loss": 2.7233,
|
| 2342 |
-
"step": 193000
|
| 2343 |
-
},
|
| 2344 |
-
{
|
| 2345 |
-
"epoch": 2.88,
|
| 2346 |
-
"learning_rate": 8.098539168724526e-07,
|
| 2347 |
-
"loss": 2.7409,
|
| 2348 |
-
"step": 193500
|
| 2349 |
-
},
|
| 2350 |
-
{
|
| 2351 |
-
"epoch": 2.89,
|
| 2352 |
-
"learning_rate": 7.602669760891772e-07,
|
| 2353 |
-
"loss": 2.7424,
|
| 2354 |
-
"step": 194000
|
| 2355 |
-
},
|
| 2356 |
-
{
|
| 2357 |
-
"epoch": 2.89,
|
| 2358 |
-
"learning_rate": 7.10680035305902e-07,
|
| 2359 |
-
"loss": 2.7563,
|
| 2360 |
-
"step": 194500
|
| 2361 |
-
},
|
| 2362 |
-
{
|
| 2363 |
-
"epoch": 2.9,
|
| 2364 |
-
"learning_rate": 6.610930945226266e-07,
|
| 2365 |
-
"loss": 2.7478,
|
| 2366 |
-
"step": 195000
|
| 2367 |
-
},
|
| 2368 |
-
{
|
| 2369 |
-
"epoch": 2.91,
|
| 2370 |
-
"learning_rate": 6.115061537393512e-07,
|
| 2371 |
-
"loss": 2.7555,
|
| 2372 |
-
"step": 195500
|
| 2373 |
-
},
|
| 2374 |
-
{
|
| 2375 |
-
"epoch": 2.92,
|
| 2376 |
-
"learning_rate": 5.619192129560759e-07,
|
| 2377 |
-
"loss": 2.7679,
|
| 2378 |
-
"step": 196000
|
| 2379 |
-
},
|
| 2380 |
-
{
|
| 2381 |
-
"epoch": 2.92,
|
| 2382 |
-
"learning_rate": 5.123322721728006e-07,
|
| 2383 |
-
"loss": 2.7219,
|
| 2384 |
-
"step": 196500
|
| 2385 |
-
},
|
| 2386 |
-
{
|
| 2387 |
-
"epoch": 2.93,
|
| 2388 |
-
"learning_rate": 4.6274533138952524e-07,
|
| 2389 |
-
"loss": 2.7283,
|
| 2390 |
-
"step": 197000
|
| 2391 |
-
},
|
| 2392 |
-
{
|
| 2393 |
-
"epoch": 2.94,
|
| 2394 |
-
"learning_rate": 4.131583906062499e-07,
|
| 2395 |
-
"loss": 2.7703,
|
| 2396 |
-
"step": 197500
|
| 2397 |
-
},
|
| 2398 |
-
{
|
| 2399 |
-
"epoch": 2.95,
|
| 2400 |
-
"learning_rate": 3.6357144982297465e-07,
|
| 2401 |
-
"loss": 2.7606,
|
| 2402 |
-
"step": 198000
|
| 2403 |
-
},
|
| 2404 |
-
{
|
| 2405 |
-
"epoch": 2.95,
|
| 2406 |
-
"learning_rate": 3.1398450903969933e-07,
|
| 2407 |
-
"loss": 2.7247,
|
| 2408 |
-
"step": 198500
|
| 2409 |
-
}
|
| 2410 |
-
],
|
| 2411 |
-
"max_steps": 201666,
|
| 2412 |
-
"num_train_epochs": 3,
|
| 2413 |
-
"total_flos": 2.3390721608830157e+17,
|
| 2414 |
-
"trial_name": null,
|
| 2415 |
-
"trial_params": null
|
| 2416 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
last-checkpoint/training_args.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:b850b2abccaa9cb6b0c14aaa89493109b9ac553fd3b9208268f6ca150829be14
|
| 3 |
-
size 3643
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1200739717
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:78c954b7531becf7ee955cd4be1cd0399bb0f33cb2a0ea78c8d99d44f686badc
|
| 3 |
size 1200739717
|
runs/Feb07_05-04-07_5214b674e698/events.out.tfevents.1675746342.5214b674e698.342.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:13f9d0277d514d6e2ad74ecd63999ef4b383df03978d356722e96e08fd2bcc67
|
| 3 |
+
size 69836
|