Training in progress, step 1150
Browse files- adapter_model.safetensors +1 -1
- metrics.json +1 -0
- state.json +18 -3
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 35668592
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7479b5aff4fb6157ffd37e56061e1b2c5f90d14f4601420aadd7008a772e6a3b
|
3 |
size 35668592
|
metrics.json
CHANGED
@@ -20,3 +20,4 @@
|
|
20 |
{"Step":1000,"eval_loss":0.751624763,"eval_runtime":56.5619,"eval_samples_per_second":1.768,"eval_steps_per_second":0.23,"epoch":1.9342359768}
|
21 |
{"Step":1050,"eval_loss":0.7514955997,"eval_runtime":60.5727,"eval_samples_per_second":1.651,"eval_steps_per_second":0.215,"epoch":2.0309477756}
|
22 |
{"Step":1100,"eval_loss":0.7509506345,"eval_runtime":56.8661,"eval_samples_per_second":1.759,"eval_steps_per_second":0.229,"epoch":2.1276595745}
|
|
|
|
20 |
{"Step":1000,"eval_loss":0.751624763,"eval_runtime":56.5619,"eval_samples_per_second":1.768,"eval_steps_per_second":0.23,"epoch":1.9342359768}
|
21 |
{"Step":1050,"eval_loss":0.7514955997,"eval_runtime":60.5727,"eval_samples_per_second":1.651,"eval_steps_per_second":0.215,"epoch":2.0309477756}
|
22 |
{"Step":1100,"eval_loss":0.7509506345,"eval_runtime":56.8661,"eval_samples_per_second":1.759,"eval_steps_per_second":0.229,"epoch":2.1276595745}
|
23 |
+
{"Step":1150,"eval_loss":0.7505396008,"eval_runtime":56.5726,"eval_samples_per_second":1.768,"eval_steps_per_second":0.23,"epoch":2.2243713733}
|
state.json
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
-
"epoch": 2.
|
5 |
"eval_steps": 50,
|
6 |
-
"global_step":
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
@@ -322,6 +322,21 @@
|
|
322 |
"eval_samples_per_second": 1.651,
|
323 |
"eval_steps_per_second": 0.215,
|
324 |
"step": 1050
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
325 |
}
|
326 |
],
|
327 |
"logging_steps": 50,
|
@@ -341,7 +356,7 @@
|
|
341 |
"attributes": {}
|
342 |
}
|
343 |
},
|
344 |
-
"total_flos": 1.
|
345 |
"train_batch_size": 8,
|
346 |
"trial_name": null,
|
347 |
"trial_params": null
|
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
+
"epoch": 2.2243713733075436,
|
5 |
"eval_steps": 50,
|
6 |
+
"global_step": 1150,
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
|
|
322 |
"eval_samples_per_second": 1.651,
|
323 |
"eval_steps_per_second": 0.215,
|
324 |
"step": 1050
|
325 |
+
},
|
326 |
+
{
|
327 |
+
"epoch": 2.127659574468085,
|
328 |
+
"grad_norm": 0.04245985299348831,
|
329 |
+
"learning_rate": 0.00016164874551971326,
|
330 |
+
"loss": 0.7264,
|
331 |
+
"step": 1100
|
332 |
+
},
|
333 |
+
{
|
334 |
+
"epoch": 2.127659574468085,
|
335 |
+
"eval_loss": 0.7509506344795227,
|
336 |
+
"eval_runtime": 56.8661,
|
337 |
+
"eval_samples_per_second": 1.759,
|
338 |
+
"eval_steps_per_second": 0.229,
|
339 |
+
"step": 1100
|
340 |
}
|
341 |
],
|
342 |
"logging_steps": 50,
|
|
|
356 |
"attributes": {}
|
357 |
}
|
358 |
},
|
359 |
+
"total_flos": 1.426061997597696e+17,
|
360 |
"train_batch_size": 8,
|
361 |
"trial_name": null,
|
362 |
"trial_params": null
|