Training in progress, step 60000, checkpoint
Browse files
last-checkpoint/adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3555504
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4431a569e16eb6669f33895a534eefcadcc6125a0fcf8243e551a17cbe6b776e
|
| 3 |
size 3555504
|
last-checkpoint/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7141515
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5c6c8f3d8741838a0a0349b3de3a3bd22b9e28163b525970af04b332fd7c349
|
| 3 |
size 7141515
|
last-checkpoint/rng_state.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14645
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7dc3fe5842c642250bc8fc58f8686955286eff74353ef4cfdb5c91b975ca0c5b
|
| 3 |
size 14645
|
last-checkpoint/scaler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1383
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:31d1175317fe37be2b1293cac47d5356f1e200c454ceace1ce4e9dd0eded0aa5
|
| 3 |
size 1383
|
last-checkpoint/scheduler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1465
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d63d134ebc965a606838bb6f037c88484af3b1856e8a790a7b2291b1079013ce
|
| 3 |
size 1465
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -2,9 +2,9 @@
|
|
| 2 |
"best_global_step": 52000,
|
| 3 |
"best_metric": 0.9908199660129274,
|
| 4 |
"best_model_checkpoint": "CodeGenDetect-CodeBert_Lora/checkpoint-52000",
|
| 5 |
-
"epoch": 1.
|
| 6 |
"eval_steps": 4000,
|
| 7 |
-
"global_step":
|
| 8 |
"is_hyper_param_search": false,
|
| 9 |
"is_local_process_zero": true,
|
| 10 |
"is_world_process_zero": true,
|
|
@@ -4096,6 +4096,298 @@
|
|
| 4096 |
"eval_samples_per_second": 130.777,
|
| 4097 |
"eval_steps_per_second": 8.174,
|
| 4098 |
"step": 56000
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4099 |
}
|
| 4100 |
],
|
| 4101 |
"logging_steps": 100,
|
|
@@ -4110,7 +4402,7 @@
|
|
| 4110 |
"early_stopping_threshold": 0.0
|
| 4111 |
},
|
| 4112 |
"attributes": {
|
| 4113 |
-
"early_stopping_patience_counter":
|
| 4114 |
}
|
| 4115 |
},
|
| 4116 |
"TrainerControl": {
|
|
@@ -4124,7 +4416,7 @@
|
|
| 4124 |
"attributes": {}
|
| 4125 |
}
|
| 4126 |
},
|
| 4127 |
-
"total_flos": 2.
|
| 4128 |
"train_batch_size": 16,
|
| 4129 |
"trial_name": null,
|
| 4130 |
"trial_params": null
|
|
|
|
| 2 |
"best_global_step": 52000,
|
| 3 |
"best_metric": 0.9908199660129274,
|
| 4 |
"best_model_checkpoint": "CodeGenDetect-CodeBert_Lora/checkpoint-52000",
|
| 5 |
+
"epoch": 1.92,
|
| 6 |
"eval_steps": 4000,
|
| 7 |
+
"global_step": 60000,
|
| 8 |
"is_hyper_param_search": false,
|
| 9 |
"is_local_process_zero": true,
|
| 10 |
"is_world_process_zero": true,
|
|
|
|
| 4096 |
"eval_samples_per_second": 130.777,
|
| 4097 |
"eval_steps_per_second": 8.174,
|
| 4098 |
"step": 56000
|
| 4099 |
+
},
|
| 4100 |
+
{
|
| 4101 |
+
"epoch": 1.7952,
|
| 4102 |
+
"grad_norm": 0.0088375024497509,
|
| 4103 |
+
"learning_rate": 1.2860481540930981e-05,
|
| 4104 |
+
"loss": 0.0303,
|
| 4105 |
+
"step": 56100
|
| 4106 |
+
},
|
| 4107 |
+
{
|
| 4108 |
+
"epoch": 1.7984,
|
| 4109 |
+
"grad_norm": 0.0020017733331769705,
|
| 4110 |
+
"learning_rate": 1.2847640449438204e-05,
|
| 4111 |
+
"loss": 0.0222,
|
| 4112 |
+
"step": 56200
|
| 4113 |
+
},
|
| 4114 |
+
{
|
| 4115 |
+
"epoch": 1.8016,
|
| 4116 |
+
"grad_norm": 0.0033814297057688236,
|
| 4117 |
+
"learning_rate": 1.2834799357945427e-05,
|
| 4118 |
+
"loss": 0.0399,
|
| 4119 |
+
"step": 56300
|
| 4120 |
+
},
|
| 4121 |
+
{
|
| 4122 |
+
"epoch": 1.8048,
|
| 4123 |
+
"grad_norm": 0.09427639096975327,
|
| 4124 |
+
"learning_rate": 1.282195826645265e-05,
|
| 4125 |
+
"loss": 0.0325,
|
| 4126 |
+
"step": 56400
|
| 4127 |
+
},
|
| 4128 |
+
{
|
| 4129 |
+
"epoch": 1.808,
|
| 4130 |
+
"grad_norm": 0.02629452385008335,
|
| 4131 |
+
"learning_rate": 1.2809117174959874e-05,
|
| 4132 |
+
"loss": 0.0265,
|
| 4133 |
+
"step": 56500
|
| 4134 |
+
},
|
| 4135 |
+
{
|
| 4136 |
+
"epoch": 1.8112,
|
| 4137 |
+
"grad_norm": 0.24074454605579376,
|
| 4138 |
+
"learning_rate": 1.2796276083467097e-05,
|
| 4139 |
+
"loss": 0.0473,
|
| 4140 |
+
"step": 56600
|
| 4141 |
+
},
|
| 4142 |
+
{
|
| 4143 |
+
"epoch": 1.8144,
|
| 4144 |
+
"grad_norm": 0.05172184109687805,
|
| 4145 |
+
"learning_rate": 1.278343499197432e-05,
|
| 4146 |
+
"loss": 0.0392,
|
| 4147 |
+
"step": 56700
|
| 4148 |
+
},
|
| 4149 |
+
{
|
| 4150 |
+
"epoch": 1.8176,
|
| 4151 |
+
"grad_norm": 0.042050667107105255,
|
| 4152 |
+
"learning_rate": 1.2770593900481543e-05,
|
| 4153 |
+
"loss": 0.0365,
|
| 4154 |
+
"step": 56800
|
| 4155 |
+
},
|
| 4156 |
+
{
|
| 4157 |
+
"epoch": 1.8208,
|
| 4158 |
+
"grad_norm": 0.011465741321444511,
|
| 4159 |
+
"learning_rate": 1.2757752808988766e-05,
|
| 4160 |
+
"loss": 0.0301,
|
| 4161 |
+
"step": 56900
|
| 4162 |
+
},
|
| 4163 |
+
{
|
| 4164 |
+
"epoch": 1.8239999999999998,
|
| 4165 |
+
"grad_norm": 1.5907392501831055,
|
| 4166 |
+
"learning_rate": 1.274491171749599e-05,
|
| 4167 |
+
"loss": 0.0489,
|
| 4168 |
+
"step": 57000
|
| 4169 |
+
},
|
| 4170 |
+
{
|
| 4171 |
+
"epoch": 1.8272,
|
| 4172 |
+
"grad_norm": 0.01520006824284792,
|
| 4173 |
+
"learning_rate": 1.2732070626003212e-05,
|
| 4174 |
+
"loss": 0.0425,
|
| 4175 |
+
"step": 57100
|
| 4176 |
+
},
|
| 4177 |
+
{
|
| 4178 |
+
"epoch": 1.8304,
|
| 4179 |
+
"grad_norm": 0.03128599748015404,
|
| 4180 |
+
"learning_rate": 1.2719229534510436e-05,
|
| 4181 |
+
"loss": 0.016,
|
| 4182 |
+
"step": 57200
|
| 4183 |
+
},
|
| 4184 |
+
{
|
| 4185 |
+
"epoch": 1.8336000000000001,
|
| 4186 |
+
"grad_norm": 0.07054321467876434,
|
| 4187 |
+
"learning_rate": 1.2706388443017657e-05,
|
| 4188 |
+
"loss": 0.0455,
|
| 4189 |
+
"step": 57300
|
| 4190 |
+
},
|
| 4191 |
+
{
|
| 4192 |
+
"epoch": 1.8368,
|
| 4193 |
+
"grad_norm": 1.4368666410446167,
|
| 4194 |
+
"learning_rate": 1.269354735152488e-05,
|
| 4195 |
+
"loss": 0.0507,
|
| 4196 |
+
"step": 57400
|
| 4197 |
+
},
|
| 4198 |
+
{
|
| 4199 |
+
"epoch": 1.8399999999999999,
|
| 4200 |
+
"grad_norm": 0.012565890327095985,
|
| 4201 |
+
"learning_rate": 1.2680706260032103e-05,
|
| 4202 |
+
"loss": 0.041,
|
| 4203 |
+
"step": 57500
|
| 4204 |
+
},
|
| 4205 |
+
{
|
| 4206 |
+
"epoch": 1.8432,
|
| 4207 |
+
"grad_norm": 1.327050805091858,
|
| 4208 |
+
"learning_rate": 1.2667865168539326e-05,
|
| 4209 |
+
"loss": 0.0167,
|
| 4210 |
+
"step": 57600
|
| 4211 |
+
},
|
| 4212 |
+
{
|
| 4213 |
+
"epoch": 1.8464,
|
| 4214 |
+
"grad_norm": 5.873330593109131,
|
| 4215 |
+
"learning_rate": 1.265502407704655e-05,
|
| 4216 |
+
"loss": 0.0417,
|
| 4217 |
+
"step": 57700
|
| 4218 |
+
},
|
| 4219 |
+
{
|
| 4220 |
+
"epoch": 1.8496000000000001,
|
| 4221 |
+
"grad_norm": 1.0156079530715942,
|
| 4222 |
+
"learning_rate": 1.2642182985553773e-05,
|
| 4223 |
+
"loss": 0.051,
|
| 4224 |
+
"step": 57800
|
| 4225 |
+
},
|
| 4226 |
+
{
|
| 4227 |
+
"epoch": 1.8528,
|
| 4228 |
+
"grad_norm": 0.2399500459432602,
|
| 4229 |
+
"learning_rate": 1.2629341894060996e-05,
|
| 4230 |
+
"loss": 0.0235,
|
| 4231 |
+
"step": 57900
|
| 4232 |
+
},
|
| 4233 |
+
{
|
| 4234 |
+
"epoch": 1.8559999999999999,
|
| 4235 |
+
"grad_norm": 0.5918147563934326,
|
| 4236 |
+
"learning_rate": 1.2616500802568219e-05,
|
| 4237 |
+
"loss": 0.0383,
|
| 4238 |
+
"step": 58000
|
| 4239 |
+
},
|
| 4240 |
+
{
|
| 4241 |
+
"epoch": 1.8592,
|
| 4242 |
+
"grad_norm": 0.8496889472007751,
|
| 4243 |
+
"learning_rate": 1.2603659711075442e-05,
|
| 4244 |
+
"loss": 0.0329,
|
| 4245 |
+
"step": 58100
|
| 4246 |
+
},
|
| 4247 |
+
{
|
| 4248 |
+
"epoch": 1.8624,
|
| 4249 |
+
"grad_norm": 3.212545871734619,
|
| 4250 |
+
"learning_rate": 1.2590818619582665e-05,
|
| 4251 |
+
"loss": 0.0363,
|
| 4252 |
+
"step": 58200
|
| 4253 |
+
},
|
| 4254 |
+
{
|
| 4255 |
+
"epoch": 1.8656000000000001,
|
| 4256 |
+
"grad_norm": 0.2524072527885437,
|
| 4257 |
+
"learning_rate": 1.2577977528089888e-05,
|
| 4258 |
+
"loss": 0.0443,
|
| 4259 |
+
"step": 58300
|
| 4260 |
+
},
|
| 4261 |
+
{
|
| 4262 |
+
"epoch": 1.8688,
|
| 4263 |
+
"grad_norm": 0.09237070381641388,
|
| 4264 |
+
"learning_rate": 1.2565136436597111e-05,
|
| 4265 |
+
"loss": 0.0394,
|
| 4266 |
+
"step": 58400
|
| 4267 |
+
},
|
| 4268 |
+
{
|
| 4269 |
+
"epoch": 1.8719999999999999,
|
| 4270 |
+
"grad_norm": 0.08935701102018356,
|
| 4271 |
+
"learning_rate": 1.2552295345104335e-05,
|
| 4272 |
+
"loss": 0.0251,
|
| 4273 |
+
"step": 58500
|
| 4274 |
+
},
|
| 4275 |
+
{
|
| 4276 |
+
"epoch": 1.8752,
|
| 4277 |
+
"grad_norm": 0.03011847473680973,
|
| 4278 |
+
"learning_rate": 1.2539454253611558e-05,
|
| 4279 |
+
"loss": 0.0364,
|
| 4280 |
+
"step": 58600
|
| 4281 |
+
},
|
| 4282 |
+
{
|
| 4283 |
+
"epoch": 1.8784,
|
| 4284 |
+
"grad_norm": 0.00672591058537364,
|
| 4285 |
+
"learning_rate": 1.2526613162118781e-05,
|
| 4286 |
+
"loss": 0.0453,
|
| 4287 |
+
"step": 58700
|
| 4288 |
+
},
|
| 4289 |
+
{
|
| 4290 |
+
"epoch": 1.8816000000000002,
|
| 4291 |
+
"grad_norm": 0.01110851764678955,
|
| 4292 |
+
"learning_rate": 1.2513772070626004e-05,
|
| 4293 |
+
"loss": 0.0316,
|
| 4294 |
+
"step": 58800
|
| 4295 |
+
},
|
| 4296 |
+
{
|
| 4297 |
+
"epoch": 1.8848,
|
| 4298 |
+
"grad_norm": 13.021896362304688,
|
| 4299 |
+
"learning_rate": 1.2500930979133227e-05,
|
| 4300 |
+
"loss": 0.0438,
|
| 4301 |
+
"step": 58900
|
| 4302 |
+
},
|
| 4303 |
+
{
|
| 4304 |
+
"epoch": 1.888,
|
| 4305 |
+
"grad_norm": 0.01617710292339325,
|
| 4306 |
+
"learning_rate": 1.248808988764045e-05,
|
| 4307 |
+
"loss": 0.0281,
|
| 4308 |
+
"step": 59000
|
| 4309 |
+
},
|
| 4310 |
+
{
|
| 4311 |
+
"epoch": 1.8912,
|
| 4312 |
+
"grad_norm": 0.126037135720253,
|
| 4313 |
+
"learning_rate": 1.2475248796147675e-05,
|
| 4314 |
+
"loss": 0.0361,
|
| 4315 |
+
"step": 59100
|
| 4316 |
+
},
|
| 4317 |
+
{
|
| 4318 |
+
"epoch": 1.8944,
|
| 4319 |
+
"grad_norm": 0.1865486055612564,
|
| 4320 |
+
"learning_rate": 1.2462407704654898e-05,
|
| 4321 |
+
"loss": 0.0294,
|
| 4322 |
+
"step": 59200
|
| 4323 |
+
},
|
| 4324 |
+
{
|
| 4325 |
+
"epoch": 1.8976,
|
| 4326 |
+
"grad_norm": 24.230648040771484,
|
| 4327 |
+
"learning_rate": 1.2449566613162121e-05,
|
| 4328 |
+
"loss": 0.0418,
|
| 4329 |
+
"step": 59300
|
| 4330 |
+
},
|
| 4331 |
+
{
|
| 4332 |
+
"epoch": 1.9008,
|
| 4333 |
+
"grad_norm": 0.008240806870162487,
|
| 4334 |
+
"learning_rate": 1.2436725521669345e-05,
|
| 4335 |
+
"loss": 0.0369,
|
| 4336 |
+
"step": 59400
|
| 4337 |
+
},
|
| 4338 |
+
{
|
| 4339 |
+
"epoch": 1.904,
|
| 4340 |
+
"grad_norm": 0.13642114400863647,
|
| 4341 |
+
"learning_rate": 1.2423884430176568e-05,
|
| 4342 |
+
"loss": 0.0264,
|
| 4343 |
+
"step": 59500
|
| 4344 |
+
},
|
| 4345 |
+
{
|
| 4346 |
+
"epoch": 1.9072,
|
| 4347 |
+
"grad_norm": 7.658884525299072,
|
| 4348 |
+
"learning_rate": 1.2411043338683787e-05,
|
| 4349 |
+
"loss": 0.0358,
|
| 4350 |
+
"step": 59600
|
| 4351 |
+
},
|
| 4352 |
+
{
|
| 4353 |
+
"epoch": 1.9104,
|
| 4354 |
+
"grad_norm": 0.019852541387081146,
|
| 4355 |
+
"learning_rate": 1.239820224719101e-05,
|
| 4356 |
+
"loss": 0.0431,
|
| 4357 |
+
"step": 59700
|
| 4358 |
+
},
|
| 4359 |
+
{
|
| 4360 |
+
"epoch": 1.9136,
|
| 4361 |
+
"grad_norm": 0.2717871367931366,
|
| 4362 |
+
"learning_rate": 1.2385361155698234e-05,
|
| 4363 |
+
"loss": 0.0429,
|
| 4364 |
+
"step": 59800
|
| 4365 |
+
},
|
| 4366 |
+
{
|
| 4367 |
+
"epoch": 1.9167999999999998,
|
| 4368 |
+
"grad_norm": 0.020897701382637024,
|
| 4369 |
+
"learning_rate": 1.2372520064205458e-05,
|
| 4370 |
+
"loss": 0.0372,
|
| 4371 |
+
"step": 59900
|
| 4372 |
+
},
|
| 4373 |
+
{
|
| 4374 |
+
"epoch": 1.92,
|
| 4375 |
+
"grad_norm": 0.02349219284951687,
|
| 4376 |
+
"learning_rate": 1.2359678972712682e-05,
|
| 4377 |
+
"loss": 0.0441,
|
| 4378 |
+
"step": 60000
|
| 4379 |
+
},
|
| 4380 |
+
{
|
| 4381 |
+
"epoch": 1.92,
|
| 4382 |
+
"eval_accuracy": 0.99049,
|
| 4383 |
+
"eval_f1": 0.9904909638313073,
|
| 4384 |
+
"eval_loss": 0.03547532483935356,
|
| 4385 |
+
"eval_precision": 0.9905025162088448,
|
| 4386 |
+
"eval_recall": 0.99049,
|
| 4387 |
+
"eval_runtime": 764.9462,
|
| 4388 |
+
"eval_samples_per_second": 130.728,
|
| 4389 |
+
"eval_steps_per_second": 8.171,
|
| 4390 |
+
"step": 60000
|
| 4391 |
}
|
| 4392 |
],
|
| 4393 |
"logging_steps": 100,
|
|
|
|
| 4402 |
"early_stopping_threshold": 0.0
|
| 4403 |
},
|
| 4404 |
"attributes": {
|
| 4405 |
+
"early_stopping_patience_counter": 2
|
| 4406 |
}
|
| 4407 |
},
|
| 4408 |
"TrainerControl": {
|
|
|
|
| 4416 |
"attributes": {}
|
| 4417 |
}
|
| 4418 |
},
|
| 4419 |
+
"total_flos": 2.5479625112655552e+17,
|
| 4420 |
"train_batch_size": 16,
|
| 4421 |
"trial_name": null,
|
| 4422 |
"trial_params": null
|