Training in progress, step 1197
Browse files- adapter_model.safetensors +1 -1
- trainer_log.jsonl +14 -0
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 819328
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77556f74e6647045066e2f3039d105e1d90d6a906d4de733ea6942e86a0e2ea4
|
| 3 |
size 819328
|
trainer_log.jsonl
CHANGED
|
@@ -245,3 +245,17 @@
|
|
| 245 |
{"current_steps": 1135, "total_steps": 1250, "loss": 0.3666, "lr": 0.0007801371805323276, "epoch": 9.08, "percentage": 90.8, "elapsed_time": "0:02:47", "remaining_time": "0:00:16", "throughput": 2654.34, "total_tokens": 445040}
|
| 246 |
{"current_steps": 1140, "total_steps": 1250, "loss": 0.3579, "lr": 0.0007148613558025102, "epoch": 9.12, "percentage": 91.2, "elapsed_time": "0:02:48", "remaining_time": "0:00:16", "throughput": 2655.66, "total_tokens": 446864}
|
| 247 |
{"current_steps": 1145, "total_steps": 1250, "loss": 0.3587, "lr": 0.0006523704471558306, "epoch": 9.16, "percentage": 91.6, "elapsed_time": "0:02:48", "remaining_time": "0:00:15", "throughput": 2659.2, "total_tokens": 449008}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 245 |
{"current_steps": 1135, "total_steps": 1250, "loss": 0.3666, "lr": 0.0007801371805323276, "epoch": 9.08, "percentage": 90.8, "elapsed_time": "0:02:47", "remaining_time": "0:00:16", "throughput": 2654.34, "total_tokens": 445040}
|
| 246 |
{"current_steps": 1140, "total_steps": 1250, "loss": 0.3579, "lr": 0.0007148613558025102, "epoch": 9.12, "percentage": 91.2, "elapsed_time": "0:02:48", "remaining_time": "0:00:16", "throughput": 2655.66, "total_tokens": 446864}
|
| 247 |
{"current_steps": 1145, "total_steps": 1250, "loss": 0.3587, "lr": 0.0006523704471558306, "epoch": 9.16, "percentage": 91.6, "elapsed_time": "0:02:48", "remaining_time": "0:00:15", "throughput": 2659.2, "total_tokens": 449008}
|
| 248 |
+
{"current_steps": 1150, "total_steps": 1250, "loss": 0.3458, "lr": 0.0005926766373187531, "epoch": 9.2, "percentage": 92.0, "elapsed_time": "0:02:49", "remaining_time": "0:00:14", "throughput": 2661.37, "total_tokens": 450832}
|
| 249 |
+
{"current_steps": 1155, "total_steps": 1250, "loss": 0.3488, "lr": 0.0005357915637177817, "epoch": 9.24, "percentage": 92.4, "elapsed_time": "0:02:49", "remaining_time": "0:00:13", "throughput": 2664.01, "total_tokens": 452784}
|
| 250 |
+
{"current_steps": 1160, "total_steps": 1250, "loss": 0.34, "lr": 0.00048172631621072045, "epoch": 9.28, "percentage": 92.8, "elapsed_time": "0:02:50", "remaining_time": "0:00:13", "throughput": 2667.78, "total_tokens": 454992}
|
| 251 |
+
{"current_steps": 1165, "total_steps": 1250, "loss": 0.3476, "lr": 0.00043049143492470017, "epoch": 9.32, "percentage": 93.2, "elapsed_time": "0:02:51", "remaining_time": "0:00:12", "throughput": 2668.66, "total_tokens": 456528}
|
| 252 |
+
{"current_steps": 1170, "total_steps": 1250, "loss": 0.3434, "lr": 0.00038209690820134145, "epoch": 9.36, "percentage": 93.6, "elapsed_time": "0:02:51", "remaining_time": "0:00:11", "throughput": 2672.4, "total_tokens": 458768}
|
| 253 |
+
{"current_steps": 1175, "total_steps": 1250, "loss": 0.3664, "lr": 0.0003365521706495234, "epoch": 9.4, "percentage": 94.0, "elapsed_time": "0:02:52", "remaining_time": "0:00:10", "throughput": 2675.44, "total_tokens": 460816}
|
| 254 |
+
{"current_steps": 1180, "total_steps": 1250, "loss": 0.3411, "lr": 0.00029386610130606504, "epoch": 9.44, "percentage": 94.4, "elapsed_time": "0:02:52", "remaining_time": "0:00:10", "throughput": 2676.17, "total_tokens": 462384}
|
| 255 |
+
{"current_steps": 1185, "total_steps": 1250, "loss": 0.3474, "lr": 0.00025404702190476856, "epoch": 9.48, "percentage": 94.8, "elapsed_time": "0:02:53", "remaining_time": "0:00:09", "throughput": 2678.71, "total_tokens": 464304}
|
| 256 |
+
{"current_steps": 1190, "total_steps": 1250, "loss": 0.3438, "lr": 0.00021710269525405834, "epoch": 9.52, "percentage": 95.2, "elapsed_time": "0:02:53", "remaining_time": "0:00:08", "throughput": 2681.88, "total_tokens": 466384}
|
| 257 |
+
{"current_steps": 1195, "total_steps": 1250, "loss": 0.3406, "lr": 0.00018304032372361666, "epoch": 9.56, "percentage": 95.6, "elapsed_time": "0:02:54", "remaining_time": "0:00:08", "throughput": 2684.67, "total_tokens": 468368}
|
| 258 |
+
{"current_steps": 1197, "total_steps": 1250, "eval_loss": 0.3528577983379364, "epoch": 9.576, "percentage": 95.76, "elapsed_time": "0:02:55", "remaining_time": "0:00:07", "throughput": 2670.33, "total_tokens": 469104}
|
| 259 |
+
{"current_steps": 1200, "total_steps": 1250, "loss": 0.3416, "lr": 0.00015186654784026365, "epoch": 9.6, "percentage": 96.0, "elapsed_time": "0:02:57", "remaining_time": "0:00:07", "throughput": 2655.46, "total_tokens": 470192}
|
| 260 |
+
{"current_steps": 1205, "total_steps": 1250, "loss": 0.3411, "lr": 0.00012358744499337603, "epoch": 9.64, "percentage": 96.4, "elapsed_time": "0:02:57", "remaining_time": "0:00:06", "throughput": 2658.4, "total_tokens": 472240}
|
| 261 |
+
{"current_steps": 1210, "total_steps": 1250, "loss": 0.3502, "lr": 9.820852825008664e-05, "epoch": 9.68, "percentage": 96.8, "elapsed_time": "0:02:58", "remaining_time": "0:00:05", "throughput": 2660.92, "total_tokens": 474160}
|