Training in progress, step 1100
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 sha256:0249ddb36a9d4dc7a7221c0517131da5d934fe1fa4a4a18f655e370827ae3e59
|
3 |
size 35668592
|
metrics.json
CHANGED
@@ -19,3 +19,4 @@
|
|
19 |
{"Step":950,"eval_loss":0.7525290847,"eval_runtime":56.698,"eval_samples_per_second":1.764,"eval_steps_per_second":0.229,"epoch":1.8375241779}
|
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}
|
|
|
|
19 |
{"Step":950,"eval_loss":0.7525290847,"eval_runtime":56.698,"eval_samples_per_second":1.764,"eval_steps_per_second":0.229,"epoch":1.8375241779}
|
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}
|
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,
|
@@ -307,6 +307,21 @@
|
|
307 |
"eval_samples_per_second": 1.768,
|
308 |
"eval_steps_per_second": 0.23,
|
309 |
"step": 1000
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
310 |
}
|
311 |
],
|
312 |
"logging_steps": 50,
|
@@ -326,7 +341,7 @@
|
|
326 |
"attributes": {}
|
327 |
}
|
328 |
},
|
329 |
-
"total_flos": 1.
|
330 |
"train_batch_size": 8,
|
331 |
"trial_name": null,
|
332 |
"trial_params": null
|
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
+
"epoch": 2.127659574468085,
|
5 |
"eval_steps": 50,
|
6 |
+
"global_step": 1100,
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
|
|
307 |
"eval_samples_per_second": 1.768,
|
308 |
"eval_steps_per_second": 0.23,
|
309 |
"step": 1000
|
310 |
+
},
|
311 |
+
{
|
312 |
+
"epoch": 2.0309477756286265,
|
313 |
+
"grad_norm": 0.04249171167612076,
|
314 |
+
"learning_rate": 0.00017956989247311828,
|
315 |
+
"loss": 0.7328,
|
316 |
+
"step": 1050
|
317 |
+
},
|
318 |
+
{
|
319 |
+
"epoch": 2.0309477756286265,
|
320 |
+
"eval_loss": 0.7514955997467041,
|
321 |
+
"eval_runtime": 60.5727,
|
322 |
+
"eval_samples_per_second": 1.651,
|
323 |
+
"eval_steps_per_second": 0.215,
|
324 |
+
"step": 1050
|
325 |
}
|
326 |
],
|
327 |
"logging_steps": 50,
|
|
|
341 |
"attributes": {}
|
342 |
}
|
343 |
},
|
344 |
+
"total_flos": 1.3621119889631232e+17,
|
345 |
"train_batch_size": 8,
|
346 |
"trial_name": null,
|
347 |
"trial_params": null
|