Training in progress, step 945
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:a38c94112767020291eb7ca4e0b9f3d318b21f491dcb64eb3355a00d4d39f714
|
| 3 |
size 819328
|
trainer_log.jsonl
CHANGED
|
@@ -191,3 +191,17 @@
|
|
| 191 |
{"current_steps": 885, "total_steps": 1250, "loss": 0.3407, "lr": 0.007175919064573383, "epoch": 7.08, "percentage": 70.8, "elapsed_time": "0:02:10", "remaining_time": "0:00:53", "throughput": 2653.97, "total_tokens": 347232}
|
| 192 |
{"current_steps": 890, "total_steps": 1250, "loss": 0.484, "lr": 0.006997996493425461, "epoch": 7.12, "percentage": 71.2, "elapsed_time": "0:02:11", "remaining_time": "0:00:53", "throughput": 2658.83, "total_tokens": 349504}
|
| 193 |
{"current_steps": 895, "total_steps": 1250, "loss": 0.3891, "lr": 0.0068216339287162486, "epoch": 7.16, "percentage": 71.6, "elapsed_time": "0:02:11", "remaining_time": "0:00:52", "throughput": 2661.47, "total_tokens": 351296}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 191 |
{"current_steps": 885, "total_steps": 1250, "loss": 0.3407, "lr": 0.007175919064573383, "epoch": 7.08, "percentage": 70.8, "elapsed_time": "0:02:10", "remaining_time": "0:00:53", "throughput": 2653.97, "total_tokens": 347232}
|
| 192 |
{"current_steps": 890, "total_steps": 1250, "loss": 0.484, "lr": 0.006997996493425461, "epoch": 7.12, "percentage": 71.2, "elapsed_time": "0:02:11", "remaining_time": "0:00:53", "throughput": 2658.83, "total_tokens": 349504}
|
| 193 |
{"current_steps": 895, "total_steps": 1250, "loss": 0.3891, "lr": 0.0068216339287162486, "epoch": 7.16, "percentage": 71.6, "elapsed_time": "0:02:11", "remaining_time": "0:00:52", "throughput": 2661.47, "total_tokens": 351296}
|
| 194 |
+
{"current_steps": 900, "total_steps": 1250, "loss": 0.4381, "lr": 0.006646865752677185, "epoch": 7.2, "percentage": 72.0, "elapsed_time": "0:02:12", "remaining_time": "0:00:51", "throughput": 2664.72, "total_tokens": 353248}
|
| 195 |
+
{"current_steps": 905, "total_steps": 1250, "loss": 0.4013, "lr": 0.00647372603671046, "epoch": 7.24, "percentage": 72.4, "elapsed_time": "0:02:13", "remaining_time": "0:00:50", "throughput": 2668.44, "total_tokens": 355232}
|
| 196 |
+
{"current_steps": 910, "total_steps": 1250, "loss": 0.3963, "lr": 0.0063022485347467615, "epoch": 7.28, "percentage": 72.8, "elapsed_time": "0:02:13", "remaining_time": "0:00:49", "throughput": 2672.78, "total_tokens": 357376}
|
| 197 |
+
{"current_steps": 915, "total_steps": 1250, "loss": 0.303, "lr": 0.00613246667666487, "epoch": 7.32, "percentage": 73.2, "elapsed_time": "0:02:14", "remaining_time": "0:00:49", "throughput": 2675.67, "total_tokens": 359232}
|
| 198 |
+
{"current_steps": 920, "total_steps": 1250, "loss": 0.4389, "lr": 0.005964413561774424, "epoch": 7.36, "percentage": 73.6, "elapsed_time": "0:02:14", "remaining_time": "0:00:48", "throughput": 2677.79, "total_tokens": 361024}
|
| 199 |
+
{"current_steps": 925, "total_steps": 1250, "loss": 0.3223, "lr": 0.0057981219523631404, "epoch": 7.4, "percentage": 74.0, "elapsed_time": "0:02:15", "remaining_time": "0:00:47", "throughput": 2682.2, "total_tokens": 363200}
|
| 200 |
+
{"current_steps": 930, "total_steps": 1250, "loss": 0.4041, "lr": 0.005633624267309767, "epoch": 7.44, "percentage": 74.4, "elapsed_time": "0:02:15", "remaining_time": "0:00:46", "throughput": 2685.48, "total_tokens": 365152}
|
| 201 |
+
{"current_steps": 935, "total_steps": 1250, "loss": 0.3497, "lr": 0.005470952575763933, "epoch": 7.48, "percentage": 74.8, "elapsed_time": "0:02:16", "remaining_time": "0:00:46", "throughput": 2689.83, "total_tokens": 367328}
|
| 202 |
+
{"current_steps": 940, "total_steps": 1250, "loss": 0.3507, "lr": 0.0053101385908942405, "epoch": 7.52, "percentage": 75.2, "elapsed_time": "0:02:17", "remaining_time": "0:00:45", "throughput": 2692.9, "total_tokens": 369248}
|
| 203 |
+
{"current_steps": 945, "total_steps": 1250, "loss": 0.3561, "lr": 0.0051512136637056555, "epoch": 7.5600000000000005, "percentage": 75.6, "elapsed_time": "0:02:17", "remaining_time": "0:00:44", "throughput": 2695.16, "total_tokens": 371040}
|
| 204 |
+
{"current_steps": 945, "total_steps": 1250, "eval_loss": 0.35614675283432007, "epoch": 7.5600000000000005, "percentage": 75.6, "elapsed_time": "0:02:18", "remaining_time": "0:00:44", "throughput": 2676.43, "total_tokens": 371040}
|
| 205 |
+
{"current_steps": 950, "total_steps": 1250, "loss": 0.3589, "lr": 0.004994208776927635, "epoch": 7.6, "percentage": 76.0, "elapsed_time": "0:02:20", "remaining_time": "0:00:44", "throughput": 2662.14, "total_tokens": 372928}
|
| 206 |
+
{"current_steps": 955, "total_steps": 1250, "loss": 0.3537, "lr": 0.004839154538973943, "epoch": 7.64, "percentage": 76.4, "elapsed_time": "0:02:20", "remaining_time": "0:00:43", "throughput": 2665.09, "total_tokens": 374816}
|
| 207 |
+
{"current_steps": 960, "total_steps": 1250, "loss": 0.3578, "lr": 0.00468608117797549, "epoch": 7.68, "percentage": 76.8, "elapsed_time": "0:02:21", "remaining_time": "0:00:42", "throughput": 2667.96, "total_tokens": 376704}
|