Hugging Face's logo

mlfoundations-dev
/
hp_ablations_mistral_bsz256

Text Generation
Transformers
Safetensors
mistral
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
sedrickkeh commited on
Commit
31f09a9
·
verified ·
1 Parent(s): e52e129

Training in progress, epoch 2

Browse files
Files changed (4) hide show
  1. model-00001-of-00003.safetensors +1 -1
  2. model-00002-of-00003.safetensors +1 -1
  3. model-00003-of-00003.safetensors +1 -1
  4. trainer_log.jsonl +98 -0
model-00001-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8634408ab318a46c965a8d386e493cd3c41bdc44e2e256b1f8bdcf8290c63641
3
  size 4943162336
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5dfbe9afdb5bf0770c57450dfe73f5ac866bfe94f61231ccf8b9659491d7363c
3
  size 4943162336
model-00002-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b966bdcfe25bb67d730f601a528cd8d5353d7ee3f026a7147be03f3676b235f3
3
  size 4999819336
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:35846dc375a15b52d2ec4548eca539689a28cddf3dfd4efac5864b59093388e6
3
  size 4999819336
model-00003-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0986a97cdbe21d3e550522107622bd1bad0fa64e59c0d7bfce528bcbb5e10b0f
3
  size 4540516344
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:161d7d3d9dee36645b837c6c991957dc0154616ae2839fd72304c5b56c171166
3
  size 4540516344
trainer_log.jsonl CHANGED
@@ -198,3 +198,101 @@
198
  {"current_steps": 1970, "total_steps": 2961, "loss": 0.468, "learning_rate": 5e-06, "epoch": 1.9949367088607595, "percentage": 66.53, "elapsed_time": "15:26:00", "remaining_time": "7:45:49"}
199
  {"current_steps": 1975, "total_steps": 2961, "eval_loss": 0.1398227959871292, "epoch": 2.0, "percentage": 66.7, "elapsed_time": "15:36:44", "remaining_time": "7:47:39"}
200
  {"current_steps": 1980, "total_steps": 2961, "loss": 0.4085, "learning_rate": 5e-06, "epoch": 2.0050632911392405, "percentage": 66.87, "elapsed_time": "15:40:11", "remaining_time": "7:45:49"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
198
  {"current_steps": 1970, "total_steps": 2961, "loss": 0.468, "learning_rate": 5e-06, "epoch": 1.9949367088607595, "percentage": 66.53, "elapsed_time": "15:26:00", "remaining_time": "7:45:49"}
199
  {"current_steps": 1975, "total_steps": 2961, "eval_loss": 0.1398227959871292, "epoch": 2.0, "percentage": 66.7, "elapsed_time": "15:36:44", "remaining_time": "7:47:39"}
200
  {"current_steps": 1980, "total_steps": 2961, "loss": 0.4085, "learning_rate": 5e-06, "epoch": 2.0050632911392405, "percentage": 66.87, "elapsed_time": "15:40:11", "remaining_time": "7:45:49"}
201
+ {"current_steps": 1990, "total_steps": 2961, "loss": 0.3545, "learning_rate": 5e-06, "epoch": 2.0151898734177216, "percentage": 67.21, "elapsed_time": "15:44:51", "remaining_time": "7:41:01"}
202
+ {"current_steps": 2000, "total_steps": 2961, "loss": 0.3486, "learning_rate": 5e-06, "epoch": 2.0253164556962027, "percentage": 67.54, "elapsed_time": "15:49:30", "remaining_time": "7:36:14"}
203
+ {"current_steps": 2010, "total_steps": 2961, "loss": 0.3485, "learning_rate": 5e-06, "epoch": 2.0354430379746837, "percentage": 67.88, "elapsed_time": "15:54:10", "remaining_time": "7:31:27"}
204
+ {"current_steps": 2020, "total_steps": 2961, "loss": 0.3469, "learning_rate": 5e-06, "epoch": 2.0455696202531644, "percentage": 68.22, "elapsed_time": "15:58:50", "remaining_time": "7:26:40"}
205
+ {"current_steps": 2030, "total_steps": 2961, "loss": 0.3412, "learning_rate": 5e-06, "epoch": 2.0556962025316454, "percentage": 68.56, "elapsed_time": "16:03:30", "remaining_time": "7:21:53"}
206
+ {"current_steps": 2040, "total_steps": 2961, "loss": 0.3499, "learning_rate": 5e-06, "epoch": 2.0658227848101265, "percentage": 68.9, "elapsed_time": "16:08:10", "remaining_time": "7:17:05"}
207
+ {"current_steps": 2050, "total_steps": 2961, "loss": 0.3431, "learning_rate": 5e-06, "epoch": 2.0759493670886076, "percentage": 69.23, "elapsed_time": "16:12:49", "remaining_time": "7:12:19"}
208
+ {"current_steps": 2060, "total_steps": 2961, "loss": 0.3509, "learning_rate": 5e-06, "epoch": 2.0860759493670886, "percentage": 69.57, "elapsed_time": "16:17:29", "remaining_time": "7:07:32"}
209
+ {"current_steps": 2070, "total_steps": 2961, "loss": 0.3485, "learning_rate": 5e-06, "epoch": 2.0962025316455697, "percentage": 69.91, "elapsed_time": "16:22:09", "remaining_time": "7:02:45"}
210
+ {"current_steps": 2080, "total_steps": 2961, "loss": 0.3414, "learning_rate": 5e-06, "epoch": 2.1063291139240508, "percentage": 70.25, "elapsed_time": "16:26:48", "remaining_time": "6:57:58"}
211
+ {"current_steps": 2090, "total_steps": 2961, "loss": 0.356, "learning_rate": 5e-06, "epoch": 2.116455696202532, "percentage": 70.58, "elapsed_time": "16:31:28", "remaining_time": "6:53:11"}
212
+ {"current_steps": 2100, "total_steps": 2961, "loss": 0.351, "learning_rate": 5e-06, "epoch": 2.1265822784810124, "percentage": 70.92, "elapsed_time": "16:36:07", "remaining_time": "6:48:24"}
213
+ {"current_steps": 2110, "total_steps": 2961, "loss": 0.3475, "learning_rate": 5e-06, "epoch": 2.1367088607594935, "percentage": 71.26, "elapsed_time": "16:40:46", "remaining_time": "6:43:37"}
214
+ {"current_steps": 2120, "total_steps": 2961, "loss": 0.3507, "learning_rate": 5e-06, "epoch": 2.1468354430379746, "percentage": 71.6, "elapsed_time": "16:45:26", "remaining_time": "6:38:51"}
215
+ {"current_steps": 2130, "total_steps": 2961, "loss": 0.3544, "learning_rate": 5e-06, "epoch": 2.1569620253164556, "percentage": 71.94, "elapsed_time": "16:50:05", "remaining_time": "6:34:04"}
216
+ {"current_steps": 2140, "total_steps": 2961, "loss": 0.3514, "learning_rate": 5e-06, "epoch": 2.1670886075949367, "percentage": 72.27, "elapsed_time": "16:54:45", "remaining_time": "6:29:18"}
217
+ {"current_steps": 2150, "total_steps": 2961, "loss": 0.3525, "learning_rate": 5e-06, "epoch": 2.1772151898734178, "percentage": 72.61, "elapsed_time": "16:59:24", "remaining_time": "6:24:31"}
218
+ {"current_steps": 2160, "total_steps": 2961, "loss": 0.3507, "learning_rate": 5e-06, "epoch": 2.187341772151899, "percentage": 72.95, "elapsed_time": "17:04:04", "remaining_time": "6:19:45"}
219
+ {"current_steps": 2170, "total_steps": 2961, "loss": 0.3496, "learning_rate": 5e-06, "epoch": 2.19746835443038, "percentage": 73.29, "elapsed_time": "17:08:44", "remaining_time": "6:14:59"}
220
+ {"current_steps": 2180, "total_steps": 2961, "loss": 0.3494, "learning_rate": 5e-06, "epoch": 2.207594936708861, "percentage": 73.62, "elapsed_time": "17:13:24", "remaining_time": "6:10:13"}
221
+ {"current_steps": 2190, "total_steps": 2961, "loss": 0.3553, "learning_rate": 5e-06, "epoch": 2.2177215189873416, "percentage": 73.96, "elapsed_time": "17:18:04", "remaining_time": "6:05:27"}
222
+ {"current_steps": 2200, "total_steps": 2961, "loss": 0.3607, "learning_rate": 5e-06, "epoch": 2.2278481012658227, "percentage": 74.3, "elapsed_time": "17:22:43", "remaining_time": "6:00:41"}
223
+ {"current_steps": 2210, "total_steps": 2961, "loss": 0.3486, "learning_rate": 5e-06, "epoch": 2.2379746835443037, "percentage": 74.64, "elapsed_time": "17:27:23", "remaining_time": "5:55:55"}
224
+ {"current_steps": 2220, "total_steps": 2961, "loss": 0.3592, "learning_rate": 5e-06, "epoch": 2.248101265822785, "percentage": 74.97, "elapsed_time": "17:32:03", "remaining_time": "5:51:09"}
225
+ {"current_steps": 2230, "total_steps": 2961, "loss": 0.3584, "learning_rate": 5e-06, "epoch": 2.258227848101266, "percentage": 75.31, "elapsed_time": "17:36:43", "remaining_time": "5:46:23"}
226
+ {"current_steps": 2240, "total_steps": 2961, "loss": 0.3554, "learning_rate": 5e-06, "epoch": 2.268354430379747, "percentage": 75.65, "elapsed_time": "17:41:22", "remaining_time": "5:41:37"}
227
+ {"current_steps": 2250, "total_steps": 2961, "loss": 0.3542, "learning_rate": 5e-06, "epoch": 2.278481012658228, "percentage": 75.99, "elapsed_time": "17:46:02", "remaining_time": "5:36:52"}
228
+ {"current_steps": 2260, "total_steps": 2961, "loss": 0.3621, "learning_rate": 5e-06, "epoch": 2.2886075949367086, "percentage": 76.33, "elapsed_time": "17:50:42", "remaining_time": "5:32:06"}
229
+ {"current_steps": 2270, "total_steps": 2961, "loss": 0.3524, "learning_rate": 5e-06, "epoch": 2.2987341772151897, "percentage": 76.66, "elapsed_time": "17:55:22", "remaining_time": "5:27:20"}
230
+ {"current_steps": 2280, "total_steps": 2961, "loss": 0.3563, "learning_rate": 5e-06, "epoch": 2.3088607594936708, "percentage": 77.0, "elapsed_time": "18:00:02", "remaining_time": "5:22:35"}
231
+ {"current_steps": 2290, "total_steps": 2961, "loss": 0.3565, "learning_rate": 5e-06, "epoch": 2.318987341772152, "percentage": 77.34, "elapsed_time": "18:04:41", "remaining_time": "5:17:49"}
232
+ {"current_steps": 2300, "total_steps": 2961, "loss": 0.3622, "learning_rate": 5e-06, "epoch": 2.329113924050633, "percentage": 77.68, "elapsed_time": "18:09:20", "remaining_time": "5:13:04"}
233
+ {"current_steps": 2310, "total_steps": 2961, "loss": 0.3589, "learning_rate": 5e-06, "epoch": 2.339240506329114, "percentage": 78.01, "elapsed_time": "18:14:00", "remaining_time": "5:08:18"}
234
+ {"current_steps": 2320, "total_steps": 2961, "loss": 0.3613, "learning_rate": 5e-06, "epoch": 2.349367088607595, "percentage": 78.35, "elapsed_time": "18:18:39", "remaining_time": "5:03:33"}
235
+ {"current_steps": 2330, "total_steps": 2961, "loss": 0.3617, "learning_rate": 5e-06, "epoch": 2.359493670886076, "percentage": 78.69, "elapsed_time": "18:23:19", "remaining_time": "4:58:47"}
236
+ {"current_steps": 2340, "total_steps": 2961, "loss": 0.3598, "learning_rate": 5e-06, "epoch": 2.369620253164557, "percentage": 79.03, "elapsed_time": "18:27:59", "remaining_time": "4:54:02"}
237
+ {"current_steps": 2350, "total_steps": 2961, "loss": 0.362, "learning_rate": 5e-06, "epoch": 2.379746835443038, "percentage": 79.37, "elapsed_time": "18:32:38", "remaining_time": "4:49:17"}
238
+ {"current_steps": 2360, "total_steps": 2961, "loss": 0.361, "learning_rate": 5e-06, "epoch": 2.389873417721519, "percentage": 79.7, "elapsed_time": "18:37:18", "remaining_time": "4:44:32"}
239
+ {"current_steps": 2370, "total_steps": 2961, "loss": 0.3639, "learning_rate": 5e-06, "epoch": 2.4, "percentage": 80.04, "elapsed_time": "18:41:58", "remaining_time": "4:39:46"}
240
+ {"current_steps": 2380, "total_steps": 2961, "loss": 0.3586, "learning_rate": 5e-06, "epoch": 2.410126582278481, "percentage": 80.38, "elapsed_time": "18:46:38", "remaining_time": "4:35:01"}
241
+ {"current_steps": 2390, "total_steps": 2961, "loss": 0.3648, "learning_rate": 5e-06, "epoch": 2.420253164556962, "percentage": 80.72, "elapsed_time": "18:51:17", "remaining_time": "4:30:16"}
242
+ {"current_steps": 2400, "total_steps": 2961, "loss": 0.3554, "learning_rate": 5e-06, "epoch": 2.430379746835443, "percentage": 81.05, "elapsed_time": "18:55:57", "remaining_time": "4:25:31"}
243
+ {"current_steps": 2410, "total_steps": 2961, "loss": 0.3531, "learning_rate": 5e-06, "epoch": 2.440506329113924, "percentage": 81.39, "elapsed_time": "19:00:37", "remaining_time": "4:20:46"}
244
+ {"current_steps": 2420, "total_steps": 2961, "loss": 0.3659, "learning_rate": 5e-06, "epoch": 2.4506329113924052, "percentage": 81.73, "elapsed_time": "19:05:16", "remaining_time": "4:16:01"}
245
+ {"current_steps": 2430, "total_steps": 2961, "loss": 0.3618, "learning_rate": 5e-06, "epoch": 2.460759493670886, "percentage": 82.07, "elapsed_time": "19:09:55", "remaining_time": "4:11:16"}
246
+ {"current_steps": 2440, "total_steps": 2961, "loss": 0.3651, "learning_rate": 5e-06, "epoch": 2.470886075949367, "percentage": 82.4, "elapsed_time": "19:14:35", "remaining_time": "4:06:31"}
247
+ {"current_steps": 2450, "total_steps": 2961, "loss": 0.3611, "learning_rate": 5e-06, "epoch": 2.481012658227848, "percentage": 82.74, "elapsed_time": "19:19:14", "remaining_time": "4:01:47"}
248
+ {"current_steps": 2460, "total_steps": 2961, "loss": 0.3655, "learning_rate": 5e-06, "epoch": 2.491139240506329, "percentage": 83.08, "elapsed_time": "19:23:54", "remaining_time": "3:57:02"}
249
+ {"current_steps": 2470, "total_steps": 2961, "loss": 0.3717, "learning_rate": 5e-06, "epoch": 2.50126582278481, "percentage": 83.42, "elapsed_time": "19:28:33", "remaining_time": "3:52:17"}
250
+ {"current_steps": 2480, "total_steps": 2961, "loss": 0.3616, "learning_rate": 5e-06, "epoch": 2.511392405063291, "percentage": 83.76, "elapsed_time": "19:33:12", "remaining_time": "3:47:32"}
251
+ {"current_steps": 2490, "total_steps": 2961, "loss": 0.3618, "learning_rate": 5e-06, "epoch": 2.5215189873417723, "percentage": 84.09, "elapsed_time": "19:37:51", "remaining_time": "3:42:48"}
252
+ {"current_steps": 2500, "total_steps": 2961, "loss": 0.3608, "learning_rate": 5e-06, "epoch": 2.5316455696202533, "percentage": 84.43, "elapsed_time": "19:42:31", "remaining_time": "3:38:03"}
253
+ {"current_steps": 2510, "total_steps": 2961, "loss": 0.3564, "learning_rate": 5e-06, "epoch": 2.5417721518987344, "percentage": 84.77, "elapsed_time": "19:47:10", "remaining_time": "3:33:18"}
254
+ {"current_steps": 2520, "total_steps": 2961, "loss": 0.3603, "learning_rate": 5e-06, "epoch": 2.5518987341772155, "percentage": 85.11, "elapsed_time": "19:51:49", "remaining_time": "3:28:34"}
255
+ {"current_steps": 2530, "total_steps": 2961, "loss": 0.3633, "learning_rate": 5e-06, "epoch": 2.562025316455696, "percentage": 85.44, "elapsed_time": "19:56:29", "remaining_time": "3:23:49"}
256
+ {"current_steps": 2540, "total_steps": 2961, "loss": 0.3658, "learning_rate": 5e-06, "epoch": 2.572151898734177, "percentage": 85.78, "elapsed_time": "20:01:08", "remaining_time": "3:19:05"}
257
+ {"current_steps": 2550, "total_steps": 2961, "loss": 0.3706, "learning_rate": 5e-06, "epoch": 2.5822784810126582, "percentage": 86.12, "elapsed_time": "20:05:47", "remaining_time": "3:14:20"}
258
+ {"current_steps": 2560, "total_steps": 2961, "loss": 0.3618, "learning_rate": 5e-06, "epoch": 2.5924050632911393, "percentage": 86.46, "elapsed_time": "20:10:27", "remaining_time": "3:09:36"}
259
+ {"current_steps": 2570, "total_steps": 2961, "loss": 0.3719, "learning_rate": 5e-06, "epoch": 2.6025316455696204, "percentage": 86.8, "elapsed_time": "20:15:07", "remaining_time": "3:04:52"}
260
+ {"current_steps": 2580, "total_steps": 2961, "loss": 0.3695, "learning_rate": 5e-06, "epoch": 2.6126582278481014, "percentage": 87.13, "elapsed_time": "20:19:46", "remaining_time": "3:00:07"}
261
+ {"current_steps": 2590, "total_steps": 2961, "loss": 0.3671, "learning_rate": 5e-06, "epoch": 2.622784810126582, "percentage": 87.47, "elapsed_time": "20:24:26", "remaining_time": "2:55:23"}
262
+ {"current_steps": 2600, "total_steps": 2961, "loss": 0.3648, "learning_rate": 5e-06, "epoch": 2.632911392405063, "percentage": 87.81, "elapsed_time": "20:29:06", "remaining_time": "2:50:39"}
263
+ {"current_steps": 2610, "total_steps": 2961, "loss": 0.3675, "learning_rate": 5e-06, "epoch": 2.643037974683544, "percentage": 88.15, "elapsed_time": "20:33:45", "remaining_time": "2:45:55"}
264
+ {"current_steps": 2620, "total_steps": 2961, "loss": 0.3638, "learning_rate": 5e-06, "epoch": 2.6531645569620252, "percentage": 88.48, "elapsed_time": "20:38:25", "remaining_time": "2:41:11"}
265
+ {"current_steps": 2630, "total_steps": 2961, "loss": 0.3683, "learning_rate": 5e-06, "epoch": 2.6632911392405063, "percentage": 88.82, "elapsed_time": "20:43:05", "remaining_time": "2:36:26"}
266
+ {"current_steps": 2640, "total_steps": 2961, "loss": 0.3685, "learning_rate": 5e-06, "epoch": 2.6734177215189874, "percentage": 89.16, "elapsed_time": "20:47:45", "remaining_time": "2:31:42"}
267
+ {"current_steps": 2650, "total_steps": 2961, "loss": 0.3684, "learning_rate": 5e-06, "epoch": 2.6835443037974684, "percentage": 89.5, "elapsed_time": "20:52:24", "remaining_time": "2:26:58"}
268
+ {"current_steps": 2660, "total_steps": 2961, "loss": 0.3723, "learning_rate": 5e-06, "epoch": 2.6936708860759495, "percentage": 89.83, "elapsed_time": "20:57:04", "remaining_time": "2:22:14"}
269
+ {"current_steps": 2670, "total_steps": 2961, "loss": 0.3671, "learning_rate": 5e-06, "epoch": 2.7037974683544306, "percentage": 90.17, "elapsed_time": "21:01:44", "remaining_time": "2:17:30"}
270
+ {"current_steps": 2680, "total_steps": 2961, "loss": 0.3715, "learning_rate": 5e-06, "epoch": 2.7139240506329116, "percentage": 90.51, "elapsed_time": "21:06:23", "remaining_time": "2:12:46"}
271
+ {"current_steps": 2690, "total_steps": 2961, "loss": 0.3674, "learning_rate": 5e-06, "epoch": 2.7240506329113923, "percentage": 90.85, "elapsed_time": "21:11:03", "remaining_time": "2:08:03"}
272
+ {"current_steps": 2700, "total_steps": 2961, "loss": 0.3701, "learning_rate": 5e-06, "epoch": 2.7341772151898733, "percentage": 91.19, "elapsed_time": "21:15:43", "remaining_time": "2:03:19"}
273
+ {"current_steps": 2710, "total_steps": 2961, "loss": 0.3692, "learning_rate": 5e-06, "epoch": 2.7443037974683544, "percentage": 91.52, "elapsed_time": "21:20:23", "remaining_time": "1:58:35"}
274
+ {"current_steps": 2720, "total_steps": 2961, "loss": 0.3605, "learning_rate": 5e-06, "epoch": 2.7544303797468355, "percentage": 91.86, "elapsed_time": "21:25:02", "remaining_time": "1:53:51"}
275
+ {"current_steps": 2730, "total_steps": 2961, "loss": 0.3681, "learning_rate": 5e-06, "epoch": 2.7645569620253165, "percentage": 92.2, "elapsed_time": "21:29:42", "remaining_time": "1:49:07"}
276
+ {"current_steps": 2740, "total_steps": 2961, "loss": 0.3732, "learning_rate": 5e-06, "epoch": 2.7746835443037976, "percentage": 92.54, "elapsed_time": "21:34:22", "remaining_time": "1:44:24"}
277
+ {"current_steps": 2750, "total_steps": 2961, "loss": 0.3666, "learning_rate": 5e-06, "epoch": 2.7848101265822782, "percentage": 92.87, "elapsed_time": "21:39:02", "remaining_time": "1:39:40"}
278
+ {"current_steps": 2760, "total_steps": 2961, "loss": 0.3706, "learning_rate": 5e-06, "epoch": 2.7949367088607593, "percentage": 93.21, "elapsed_time": "21:43:41", "remaining_time": "1:34:56"}
279
+ {"current_steps": 2770, "total_steps": 2961, "loss": 0.3679, "learning_rate": 5e-06, "epoch": 2.8050632911392404, "percentage": 93.55, "elapsed_time": "21:48:21", "remaining_time": "1:30:12"}
280
+ {"current_steps": 2780, "total_steps": 2961, "loss": 0.3761, "learning_rate": 5e-06, "epoch": 2.8151898734177214, "percentage": 93.89, "elapsed_time": "21:53:01", "remaining_time": "1:25:29"}
281
+ {"current_steps": 2790, "total_steps": 2961, "loss": 0.3697, "learning_rate": 5e-06, "epoch": 2.8253164556962025, "percentage": 94.22, "elapsed_time": "21:57:40", "remaining_time": "1:20:45"}
282
+ {"current_steps": 2800, "total_steps": 2961, "loss": 0.3613, "learning_rate": 5e-06, "epoch": 2.8354430379746836, "percentage": 94.56, "elapsed_time": "22:02:20", "remaining_time": "1:16:02"}
283
+ {"current_steps": 2810, "total_steps": 2961, "loss": 0.3773, "learning_rate": 5e-06, "epoch": 2.8455696202531646, "percentage": 94.9, "elapsed_time": "22:07:00", "remaining_time": "1:11:18"}
284
+ {"current_steps": 2820, "total_steps": 2961, "loss": 0.3699, "learning_rate": 5e-06, "epoch": 2.8556962025316457, "percentage": 95.24, "elapsed_time": "22:11:40", "remaining_time": "1:06:35"}
285
+ {"current_steps": 2830, "total_steps": 2961, "loss": 0.3711, "learning_rate": 5e-06, "epoch": 2.8658227848101268, "percentage": 95.58, "elapsed_time": "22:16:19", "remaining_time": "1:01:51"}
286
+ {"current_steps": 2840, "total_steps": 2961, "loss": 0.3691, "learning_rate": 5e-06, "epoch": 2.875949367088608, "percentage": 95.91, "elapsed_time": "22:20:59", "remaining_time": "0:57:08"}
287
+ {"current_steps": 2850, "total_steps": 2961, "loss": 0.3747, "learning_rate": 5e-06, "epoch": 2.8860759493670884, "percentage": 96.25, "elapsed_time": "22:25:39", "remaining_time": "0:52:24"}
288
+ {"current_steps": 2860, "total_steps": 2961, "loss": 0.3674, "learning_rate": 5e-06, "epoch": 2.8962025316455695, "percentage": 96.59, "elapsed_time": "22:30:19", "remaining_time": "0:47:41"}
289
+ {"current_steps": 2870, "total_steps": 2961, "loss": 0.373, "learning_rate": 5e-06, "epoch": 2.9063291139240506, "percentage": 96.93, "elapsed_time": "22:34:59", "remaining_time": "0:42:57"}
290
+ {"current_steps": 2880, "total_steps": 2961, "loss": 0.3723, "learning_rate": 5e-06, "epoch": 2.9164556962025316, "percentage": 97.26, "elapsed_time": "22:39:38", "remaining_time": "0:38:14"}
291
+ {"current_steps": 2890, "total_steps": 2961, "loss": 0.3796, "learning_rate": 5e-06, "epoch": 2.9265822784810127, "percentage": 97.6, "elapsed_time": "22:44:18", "remaining_time": "0:33:31"}
292
+ {"current_steps": 2900, "total_steps": 2961, "loss": 0.3828, "learning_rate": 5e-06, "epoch": 2.9367088607594938, "percentage": 97.94, "elapsed_time": "22:48:58", "remaining_time": "0:28:47"}
293
+ {"current_steps": 2910, "total_steps": 2961, "loss": 0.3713, "learning_rate": 5e-06, "epoch": 2.946835443037975, "percentage": 98.28, "elapsed_time": "22:53:37", "remaining_time": "0:24:04"}
294
+ {"current_steps": 2920, "total_steps": 2961, "loss": 0.369, "learning_rate": 5e-06, "epoch": 2.9569620253164555, "percentage": 98.62, "elapsed_time": "22:58:17", "remaining_time": "0:19:21"}
295
+ {"current_steps": 2930, "total_steps": 2961, "loss": 0.3721, "learning_rate": 5e-06, "epoch": 2.9670886075949365, "percentage": 98.95, "elapsed_time": "23:02:57", "remaining_time": "0:14:37"}
296
+ {"current_steps": 2940, "total_steps": 2961, "loss": 0.3759, "learning_rate": 5e-06, "epoch": 2.9772151898734176, "percentage": 99.29, "elapsed_time": "23:07:37", "remaining_time": "0:09:54"}
297
+ {"current_steps": 2950, "total_steps": 2961, "loss": 0.3746, "learning_rate": 5e-06, "epoch": 2.9873417721518987, "percentage": 99.63, "elapsed_time": "23:12:16", "remaining_time": "0:05:11"}
298
+ {"current_steps": 2960, "total_steps": 2961, "loss": 0.3711, "learning_rate": 5e-06, "epoch": 2.9974683544303797, "percentage": 99.97, "elapsed_time": "23:16:56", "remaining_time": "0:00:28"}