Training in progress, step 1200
Browse files- adapter_model.safetensors +1 -1
- trainer_log.jsonl +11 -0
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 29034840
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:419197c7b895b88c7afacbe056bbf0e1d22ab3d4ba5caafffba674acfa991533
|
3 |
size 29034840
|
trainer_log.jsonl
CHANGED
@@ -251,3 +251,14 @@
|
|
251 |
{"current_steps": 1145, "total_steps": 3400, "loss": 0.0001, "lr": 7.915270142049566e-05, "epoch": 0.5897501931496266, "percentage": 33.68, "elapsed_time": "5:39:24", "remaining_time": "11:08:26", "throughput": 561.35, "total_tokens": 11431680}
|
252 |
{"current_steps": 1150, "total_steps": 3400, "loss": 0.0127, "lr": 7.89548083375394e-05, "epoch": 0.5923255215039918, "percentage": 33.82, "elapsed_time": "5:40:50", "remaining_time": "11:06:51", "throughput": 561.43, "total_tokens": 11481600}
|
253 |
{"current_steps": 1150, "total_steps": 3400, "eval_loss": 1.5566551155643538e-05, "epoch": 0.5923255215039918, "percentage": 33.82, "elapsed_time": "5:41:09", "remaining_time": "11:07:29", "throughput": 560.9, "total_tokens": 11481600}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
251 |
{"current_steps": 1145, "total_steps": 3400, "loss": 0.0001, "lr": 7.915270142049566e-05, "epoch": 0.5897501931496266, "percentage": 33.68, "elapsed_time": "5:39:24", "remaining_time": "11:08:26", "throughput": 561.35, "total_tokens": 11431680}
|
252 |
{"current_steps": 1150, "total_steps": 3400, "loss": 0.0127, "lr": 7.89548083375394e-05, "epoch": 0.5923255215039918, "percentage": 33.82, "elapsed_time": "5:40:50", "remaining_time": "11:06:51", "throughput": 561.43, "total_tokens": 11481600}
|
253 |
{"current_steps": 1150, "total_steps": 3400, "eval_loss": 1.5566551155643538e-05, "epoch": 0.5923255215039918, "percentage": 33.82, "elapsed_time": "5:41:09", "remaining_time": "11:07:29", "throughput": 560.9, "total_tokens": 11481600}
|
254 |
+
{"current_steps": 1155, "total_steps": 3400, "loss": 0.0003, "lr": 7.875623046909544e-05, "epoch": 0.5949008498583569, "percentage": 33.97, "elapsed_time": "5:42:40", "remaining_time": "11:06:04", "throughput": 560.85, "total_tokens": 11531520}
|
255 |
+
{"current_steps": 1160, "total_steps": 3400, "loss": 0.0, "lr": 7.855697251155967e-05, "epoch": 0.5974761782127221, "percentage": 34.12, "elapsed_time": "5:44:07", "remaining_time": "11:04:30", "throughput": 560.92, "total_tokens": 11581440}
|
256 |
+
{"current_steps": 1165, "total_steps": 3400, "loss": 0.0414, "lr": 7.835703917741212e-05, "epoch": 0.6000515065670873, "percentage": 34.26, "elapsed_time": "5:45:33", "remaining_time": "11:02:55", "throughput": 561.0, "total_tokens": 11631360}
|
257 |
+
{"current_steps": 1170, "total_steps": 3400, "loss": 0.02, "lr": 7.81564351951057e-05, "epoch": 0.6026268349214525, "percentage": 34.41, "elapsed_time": "5:46:58", "remaining_time": "11:01:20", "throughput": 561.09, "total_tokens": 11681280}
|
258 |
+
{"current_steps": 1175, "total_steps": 3400, "loss": 0.0002, "lr": 7.795516530895414e-05, "epoch": 0.6052021632758177, "percentage": 34.56, "elapsed_time": "5:48:23", "remaining_time": "10:59:44", "throughput": 561.2, "total_tokens": 11731200}
|
259 |
+
{"current_steps": 1180, "total_steps": 3400, "loss": 0.0026, "lr": 7.775323427901993e-05, "epoch": 0.6077774916301828, "percentage": 34.71, "elapsed_time": "5:49:49", "remaining_time": "10:58:08", "throughput": 561.28, "total_tokens": 11781120}
|
260 |
+
{"current_steps": 1185, "total_steps": 3400, "loss": 0.0168, "lr": 7.755064688100171e-05, "epoch": 0.610352819984548, "percentage": 34.85, "elapsed_time": "5:51:14", "remaining_time": "10:56:32", "throughput": 561.38, "total_tokens": 11831040}
|
261 |
+
{"current_steps": 1190, "total_steps": 3400, "loss": 0.0037, "lr": 7.734740790612136e-05, "epoch": 0.6129281483389132, "percentage": 35.0, "elapsed_time": "5:52:40", "remaining_time": "10:54:57", "throughput": 561.47, "total_tokens": 11880960}
|
262 |
+
{"current_steps": 1195, "total_steps": 3400, "loss": 0.0007, "lr": 7.714352216101055e-05, "epoch": 0.6155034766932784, "percentage": 35.15, "elapsed_time": "5:54:05", "remaining_time": "10:53:21", "throughput": 561.58, "total_tokens": 11930880}
|
263 |
+
{"current_steps": 1200, "total_steps": 3400, "loss": 0.0038, "lr": 7.693899446759727e-05, "epoch": 0.6180788050476436, "percentage": 35.29, "elapsed_time": "5:55:30", "remaining_time": "10:51:46", "throughput": 561.67, "total_tokens": 11980800}
|
264 |
+
{"current_steps": 1200, "total_steps": 3400, "eval_loss": 0.0022160401567816734, "epoch": 0.6180788050476436, "percentage": 35.29, "elapsed_time": "5:55:49", "remaining_time": "10:52:20", "throughput": 561.17, "total_tokens": 11980800}
|