Training in progress, step 1500
Browse files- adapter_model.safetensors +1 -1
- trainer_log.jsonl +51 -0
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 645976488
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:474afdde13b45825aaa1e00f7ef60f5109d1cc33fdd07da07dfbb299b2104720
|
| 3 |
size 645976488
|
trainer_log.jsonl
CHANGED
|
@@ -99,3 +99,54 @@
|
|
| 99 |
{"current_steps": 980, "total_steps": 1770, "loss": 1.6289, "lr": 2.467953871623073e-05, "epoch": 1.6610169491525424, "percentage": 55.37, "elapsed_time": "0:33:46", "remaining_time": "0:27:13"}
|
| 100 |
{"current_steps": 990, "total_steps": 1770, "loss": 1.6339, "lr": 2.4186642632977697e-05, "epoch": 1.6779661016949152, "percentage": 55.93, "elapsed_time": "0:34:06", "remaining_time": "0:26:52"}
|
| 101 |
{"current_steps": 1000, "total_steps": 1770, "loss": 1.6172, "lr": 2.3694062876185738e-05, "epoch": 1.694915254237288, "percentage": 56.5, "elapsed_time": "0:34:26", "remaining_time": "0:26:31"}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 99 |
{"current_steps": 980, "total_steps": 1770, "loss": 1.6289, "lr": 2.467953871623073e-05, "epoch": 1.6610169491525424, "percentage": 55.37, "elapsed_time": "0:33:46", "remaining_time": "0:27:13"}
|
| 100 |
{"current_steps": 990, "total_steps": 1770, "loss": 1.6339, "lr": 2.4186642632977697e-05, "epoch": 1.6779661016949152, "percentage": 55.93, "elapsed_time": "0:34:06", "remaining_time": "0:26:52"}
|
| 101 |
{"current_steps": 1000, "total_steps": 1770, "loss": 1.6172, "lr": 2.3694062876185738e-05, "epoch": 1.694915254237288, "percentage": 56.5, "elapsed_time": "0:34:26", "remaining_time": "0:26:31"}
|
| 102 |
+
{"current_steps": 1010, "total_steps": 1770, "loss": 1.6169, "lr": 2.3201991017257007e-05, "epoch": 1.711864406779661, "percentage": 57.06, "elapsed_time": "0:35:11", "remaining_time": "0:26:28"}
|
| 103 |
+
{"current_steps": 1020, "total_steps": 1770, "loss": 1.6156, "lr": 2.2710618430064843e-05, "epoch": 1.7288135593220337, "percentage": 57.63, "elapsed_time": "0:35:31", "remaining_time": "0:26:07"}
|
| 104 |
+
{"current_steps": 1030, "total_steps": 1770, "loss": 1.6348, "lr": 2.222013621652565e-05, "epoch": 1.7457627118644068, "percentage": 58.19, "elapsed_time": "0:35:51", "remaining_time": "0:25:45"}
|
| 105 |
+
{"current_steps": 1040, "total_steps": 1770, "loss": 1.6162, "lr": 2.173073513227667e-05, "epoch": 1.7627118644067796, "percentage": 58.76, "elapsed_time": "0:36:12", "remaining_time": "0:25:24"}
|
| 106 |
+
{"current_steps": 1050, "total_steps": 1770, "loss": 1.6162, "lr": 2.1242605512488248e-05, "epoch": 1.7796610169491527, "percentage": 59.32, "elapsed_time": "0:36:32", "remaining_time": "0:25:03"}
|
| 107 |
+
{"current_steps": 1060, "total_steps": 1770, "loss": 1.6146, "lr": 2.0755937197839802e-05, "epoch": 1.7966101694915255, "percentage": 59.89, "elapsed_time": "0:36:52", "remaining_time": "0:24:42"}
|
| 108 |
+
{"current_steps": 1070, "total_steps": 1770, "loss": 1.6178, "lr": 2.0270919460688055e-05, "epoch": 1.8135593220338984, "percentage": 60.45, "elapsed_time": "0:37:13", "remaining_time": "0:24:21"}
|
| 109 |
+
{"current_steps": 1080, "total_steps": 1770, "loss": 1.6287, "lr": 1.9787740931456165e-05, "epoch": 1.8305084745762712, "percentage": 61.02, "elapsed_time": "0:37:33", "remaining_time": "0:23:59"}
|
| 110 |
+
{"current_steps": 1090, "total_steps": 1770, "loss": 1.6348, "lr": 1.9306589525272756e-05, "epoch": 1.847457627118644, "percentage": 61.58, "elapsed_time": "0:37:54", "remaining_time": "0:23:38"}
|
| 111 |
+
{"current_steps": 1100, "total_steps": 1770, "loss": 1.6295, "lr": 1.8827652368888888e-05, "epoch": 1.8644067796610169, "percentage": 62.15, "elapsed_time": "0:38:14", "remaining_time": "0:23:17"}
|
| 112 |
+
{"current_steps": 1110, "total_steps": 1770, "loss": 1.629, "lr": 1.835111572790183e-05, "epoch": 1.8813559322033897, "percentage": 62.71, "elapsed_time": "0:38:34", "remaining_time": "0:22:56"}
|
| 113 |
+
{"current_steps": 1120, "total_steps": 1770, "loss": 1.6127, "lr": 1.7877164934313628e-05, "epoch": 1.8983050847457628, "percentage": 63.28, "elapsed_time": "0:38:55", "remaining_time": "0:22:35"}
|
| 114 |
+
{"current_steps": 1130, "total_steps": 1770, "loss": 1.6345, "lr": 1.740598431445282e-05, "epoch": 1.9152542372881356, "percentage": 63.84, "elapsed_time": "0:39:15", "remaining_time": "0:22:14"}
|
| 115 |
+
{"current_steps": 1140, "total_steps": 1770, "loss": 1.6291, "lr": 1.6937757117287278e-05, "epoch": 1.9322033898305084, "percentage": 64.41, "elapsed_time": "0:39:36", "remaining_time": "0:21:53"}
|
| 116 |
+
{"current_steps": 1150, "total_steps": 1770, "loss": 1.6311, "lr": 1.6472665443155983e-05, "epoch": 1.9491525423728815, "percentage": 64.97, "elapsed_time": "0:39:56", "remaining_time": "0:21:32"}
|
| 117 |
+
{"current_steps": 1160, "total_steps": 1770, "loss": 1.6082, "lr": 1.6010890172947606e-05, "epoch": 1.9661016949152543, "percentage": 65.54, "elapsed_time": "0:40:16", "remaining_time": "0:21:10"}
|
| 118 |
+
{"current_steps": 1170, "total_steps": 1770, "loss": 1.6266, "lr": 1.5552610897753292e-05, "epoch": 1.9830508474576272, "percentage": 66.1, "elapsed_time": "0:40:37", "remaining_time": "0:20:49"}
|
| 119 |
+
{"current_steps": 1180, "total_steps": 1770, "loss": 1.6265, "lr": 1.509800584902108e-05, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "0:40:57", "remaining_time": "0:20:28"}
|
| 120 |
+
{"current_steps": 1190, "total_steps": 1770, "loss": 1.5776, "lr": 1.4647251829239139e-05, "epoch": 2.016949152542373, "percentage": 67.23, "elapsed_time": "0:41:17", "remaining_time": "0:20:07"}
|
| 121 |
+
{"current_steps": 1200, "total_steps": 1770, "loss": 1.5922, "lr": 1.4200524143174677e-05, "epoch": 2.0338983050847457, "percentage": 67.8, "elapsed_time": "0:41:38", "remaining_time": "0:19:46"}
|
| 122 |
+
{"current_steps": 1200, "total_steps": 1770, "eval_loss": 1.6305924654006958, "epoch": 2.0338983050847457, "percentage": 67.8, "elapsed_time": "0:41:42", "remaining_time": "0:19:48"}
|
| 123 |
+
{"current_steps": 1210, "total_steps": 1770, "loss": 1.5948, "lr": 1.3757996529695411e-05, "epoch": 2.0508474576271185, "percentage": 68.36, "elapsed_time": "0:42:02", "remaining_time": "0:19:27"}
|
| 124 |
+
{"current_steps": 1220, "total_steps": 1770, "loss": 1.5884, "lr": 1.331984109420001e-05, "epoch": 2.0677966101694913, "percentage": 68.93, "elapsed_time": "0:42:23", "remaining_time": "0:19:06"}
|
| 125 |
+
{"current_steps": 1230, "total_steps": 1770, "loss": 1.6067, "lr": 1.2886228241683749e-05, "epoch": 2.084745762711864, "percentage": 69.49, "elapsed_time": "0:42:43", "remaining_time": "0:18:45"}
|
| 126 |
+
{"current_steps": 1240, "total_steps": 1770, "loss": 1.6103, "lr": 1.2457326610465642e-05, "epoch": 2.1016949152542375, "percentage": 70.06, "elapsed_time": "0:43:03", "remaining_time": "0:18:24"}
|
| 127 |
+
{"current_steps": 1250, "total_steps": 1770, "loss": 1.5979, "lr": 1.2033303006602444e-05, "epoch": 2.1186440677966103, "percentage": 70.62, "elapsed_time": "0:43:24", "remaining_time": "0:18:03"}
|
| 128 |
+
{"current_steps": 1260, "total_steps": 1770, "loss": 1.5786, "lr": 1.1614322339015484e-05, "epoch": 2.135593220338983, "percentage": 71.19, "elapsed_time": "0:43:44", "remaining_time": "0:17:42"}
|
| 129 |
+
{"current_steps": 1270, "total_steps": 1770, "loss": 1.5903, "lr": 1.1200547555355067e-05, "epoch": 2.152542372881356, "percentage": 71.75, "elapsed_time": "0:44:04", "remaining_time": "0:17:21"}
|
| 130 |
+
{"current_steps": 1280, "total_steps": 1770, "loss": 1.5913, "lr": 1.0792139578627865e-05, "epoch": 2.169491525423729, "percentage": 72.32, "elapsed_time": "0:44:25", "remaining_time": "0:17:00"}
|
| 131 |
+
{"current_steps": 1290, "total_steps": 1770, "loss": 1.592, "lr": 1.0389257244611602e-05, "epoch": 2.1864406779661016, "percentage": 72.88, "elapsed_time": "0:44:45", "remaining_time": "0:16:39"}
|
| 132 |
+
{"current_steps": 1300, "total_steps": 1770, "loss": 1.5968, "lr": 9.992057240081487e-06, "epoch": 2.2033898305084745, "percentage": 73.45, "elapsed_time": "0:45:06", "remaining_time": "0:16:18"}
|
| 133 |
+
{"current_steps": 1310, "total_steps": 1770, "loss": 1.5893, "lr": 9.600694041872518e-06, "epoch": 2.2203389830508473, "percentage": 74.01, "elapsed_time": "0:45:26", "remaining_time": "0:15:57"}
|
| 134 |
+
{"current_steps": 1320, "total_steps": 1770, "loss": 1.5861, "lr": 9.215319856801158e-06, "epoch": 2.23728813559322, "percentage": 74.58, "elapsed_time": "0:45:46", "remaining_time": "0:15:36"}
|
| 135 |
+
{"current_steps": 1330, "total_steps": 1770, "loss": 1.606, "lr": 8.836084562469921e-06, "epoch": 2.2542372881355934, "percentage": 75.14, "elapsed_time": "0:46:07", "remaining_time": "0:15:15"}
|
| 136 |
+
{"current_steps": 1340, "total_steps": 1770, "loss": 1.5911, "lr": 8.463135648977773e-06, "epoch": 2.2711864406779663, "percentage": 75.71, "elapsed_time": "0:46:27", "remaining_time": "0:14:54"}
|
| 137 |
+
{"current_steps": 1350, "total_steps": 1770, "loss": 1.5832, "lr": 8.096618161559142e-06, "epoch": 2.288135593220339, "percentage": 76.27, "elapsed_time": "0:46:47", "remaining_time": "0:14:33"}
|
| 138 |
+
{"current_steps": 1360, "total_steps": 1770, "loss": 1.5868, "lr": 7.736674644173695e-06, "epoch": 2.305084745762712, "percentage": 76.84, "elapsed_time": "0:47:08", "remaining_time": "0:14:12"}
|
| 139 |
+
{"current_steps": 1370, "total_steps": 1770, "loss": 1.5788, "lr": 7.383445084068907e-06, "epoch": 2.3220338983050848, "percentage": 77.4, "elapsed_time": "0:47:28", "remaining_time": "0:13:51"}
|
| 140 |
+
{"current_steps": 1380, "total_steps": 1770, "loss": 1.5769, "lr": 7.037066857337058e-06, "epoch": 2.3389830508474576, "percentage": 77.97, "elapsed_time": "0:47:48", "remaining_time": "0:13:30"}
|
| 141 |
+
{"current_steps": 1390, "total_steps": 1770, "loss": 1.5676, "lr": 6.69767467548762e-06, "epoch": 2.3559322033898304, "percentage": 78.53, "elapsed_time": "0:48:09", "remaining_time": "0:13:09"}
|
| 142 |
+
{"current_steps": 1400, "total_steps": 1770, "loss": 1.5896, "lr": 6.365400533056012e-06, "epoch": 2.3728813559322033, "percentage": 79.1, "elapsed_time": "0:48:29", "remaining_time": "0:12:48"}
|
| 143 |
+
{"current_steps": 1410, "total_steps": 1770, "loss": 1.5698, "lr": 6.040373656269041e-06, "epoch": 2.389830508474576, "percentage": 79.66, "elapsed_time": "0:48:49", "remaining_time": "0:12:28"}
|
| 144 |
+
{"current_steps": 1420, "total_steps": 1770, "loss": 1.5881, "lr": 5.722720452786928e-06, "epoch": 2.406779661016949, "percentage": 80.23, "elapsed_time": "0:49:10", "remaining_time": "0:12:07"}
|
| 145 |
+
{"current_steps": 1430, "total_steps": 1770, "loss": 1.5864, "lr": 5.412564462541544e-06, "epoch": 2.423728813559322, "percentage": 80.79, "elapsed_time": "0:49:30", "remaining_time": "0:11:46"}
|
| 146 |
+
{"current_steps": 1440, "total_steps": 1770, "loss": 1.592, "lr": 5.110026309689922e-06, "epoch": 2.440677966101695, "percentage": 81.36, "elapsed_time": "0:49:50", "remaining_time": "0:11:25"}
|
| 147 |
+
{"current_steps": 1450, "total_steps": 1770, "loss": 1.5912, "lr": 4.815223655701812e-06, "epoch": 2.457627118644068, "percentage": 81.92, "elapsed_time": "0:50:11", "remaining_time": "0:11:04"}
|
| 148 |
+
{"current_steps": 1460, "total_steps": 1770, "loss": 1.5732, "lr": 4.5282711535994115e-06, "epoch": 2.4745762711864407, "percentage": 82.49, "elapsed_time": "0:50:31", "remaining_time": "0:10:43"}
|
| 149 |
+
{"current_steps": 1470, "total_steps": 1770, "loss": 1.5994, "lr": 4.2492804033671145e-06, "epoch": 2.4915254237288136, "percentage": 83.05, "elapsed_time": "0:50:51", "remaining_time": "0:10:22"}
|
| 150 |
+
{"current_steps": 1480, "total_steps": 1770, "loss": 1.5906, "lr": 3.978359908548746e-06, "epoch": 2.5084745762711864, "percentage": 83.62, "elapsed_time": "0:51:12", "remaining_time": "0:10:01"}
|
| 151 |
+
{"current_steps": 1490, "total_steps": 1770, "loss": 1.593, "lr": 3.71561503404885e-06, "epoch": 2.5254237288135593, "percentage": 84.18, "elapsed_time": "0:51:32", "remaining_time": "0:09:41"}
|
| 152 |
+
{"current_steps": 1500, "total_steps": 1770, "loss": 1.6067, "lr": 3.4611479651548457e-06, "epoch": 2.542372881355932, "percentage": 84.75, "elapsed_time": "0:51:52", "remaining_time": "0:09:20"}
|