Upload trainer_state.json
Browse files- trainer_state.json +9 -9
trainer_state.json
CHANGED
@@ -10,55 +10,55 @@
|
|
10 |
{
|
11 |
"epoch": 0.49,
|
12 |
"learning_rate": 4.3836291913214994e-05,
|
13 |
-
"loss":
|
14 |
"step": 500
|
15 |
},
|
16 |
{
|
17 |
"epoch": 0.99,
|
18 |
"learning_rate": 3.767258382642998e-05,
|
19 |
-
"loss": 2.
|
20 |
"step": 1000
|
21 |
},
|
22 |
{
|
23 |
"epoch": 1.48,
|
24 |
"learning_rate": 3.150887573964497e-05,
|
25 |
-
"loss": 2.
|
26 |
"step": 1500
|
27 |
},
|
28 |
{
|
29 |
"epoch": 1.97,
|
30 |
"learning_rate": 2.5345167652859964e-05,
|
31 |
-
"loss":
|
32 |
"step": 2000
|
33 |
},
|
34 |
{
|
35 |
"epoch": 2.47,
|
36 |
"learning_rate": 1.918145956607495e-05,
|
37 |
-
"loss":
|
38 |
"step": 2500
|
39 |
},
|
40 |
{
|
41 |
"epoch": 2.96,
|
42 |
"learning_rate": 1.3017751479289941e-05,
|
43 |
-
"loss": 1.
|
44 |
"step": 3000
|
45 |
},
|
46 |
{
|
47 |
"epoch": 3.45,
|
48 |
"learning_rate": 6.854043392504931e-06,
|
49 |
-
"loss": 1.
|
50 |
"step": 3500
|
51 |
},
|
52 |
{
|
53 |
"epoch": 3.94,
|
54 |
"learning_rate": 6.903353057199211e-07,
|
55 |
-
"loss": 1.
|
56 |
"step": 4000
|
57 |
}
|
58 |
],
|
59 |
"max_steps": 4056,
|
60 |
"num_train_epochs": 4,
|
61 |
-
"total_flos":
|
62 |
"trial_name": null,
|
63 |
"trial_params": null
|
64 |
}
|
|
|
10 |
{
|
11 |
"epoch": 0.49,
|
12 |
"learning_rate": 4.3836291913214994e-05,
|
13 |
+
"loss": 2.2309,
|
14 |
"step": 500
|
15 |
},
|
16 |
{
|
17 |
"epoch": 0.99,
|
18 |
"learning_rate": 3.767258382642998e-05,
|
19 |
+
"loss": 2.1438,
|
20 |
"step": 1000
|
21 |
},
|
22 |
{
|
23 |
"epoch": 1.48,
|
24 |
"learning_rate": 3.150887573964497e-05,
|
25 |
+
"loss": 2.0484,
|
26 |
"step": 1500
|
27 |
},
|
28 |
{
|
29 |
"epoch": 1.97,
|
30 |
"learning_rate": 2.5345167652859964e-05,
|
31 |
+
"loss": 1.9944,
|
32 |
"step": 2000
|
33 |
},
|
34 |
{
|
35 |
"epoch": 2.47,
|
36 |
"learning_rate": 1.918145956607495e-05,
|
37 |
+
"loss": 1.9404,
|
38 |
"step": 2500
|
39 |
},
|
40 |
{
|
41 |
"epoch": 2.96,
|
42 |
"learning_rate": 1.3017751479289941e-05,
|
43 |
+
"loss": 1.911,
|
44 |
"step": 3000
|
45 |
},
|
46 |
{
|
47 |
"epoch": 3.45,
|
48 |
"learning_rate": 6.854043392504931e-06,
|
49 |
+
"loss": 1.8689,
|
50 |
"step": 3500
|
51 |
},
|
52 |
{
|
53 |
"epoch": 3.94,
|
54 |
"learning_rate": 6.903353057199211e-07,
|
55 |
+
"loss": 1.8617,
|
56 |
"step": 4000
|
57 |
}
|
58 |
],
|
59 |
"max_steps": 4056,
|
60 |
"num_train_epochs": 4,
|
61 |
+
"total_flos": 7375470258588300.0,
|
62 |
"trial_name": null,
|
63 |
"trial_params": null
|
64 |
}
|