Training in progress, step 12000, checkpoint
Browse files
checkpoint-12000/trainer_state.json
CHANGED
@@ -11,9 +11,9 @@
|
|
11 |
{
|
12 |
"epoch": 2.5,
|
13 |
"eval_loss": 7.232339382171631,
|
14 |
-
"eval_runtime":
|
15 |
-
"eval_samples_per_second":
|
16 |
-
"eval_steps_per_second":
|
17 |
"step": 2000
|
18 |
},
|
19 |
{
|
@@ -26,17 +26,17 @@
|
|
26 |
{
|
27 |
"epoch": 5.0,
|
28 |
"eval_loss": 5.976784706115723,
|
29 |
-
"eval_runtime":
|
30 |
-
"eval_samples_per_second":
|
31 |
-
"eval_steps_per_second":
|
32 |
"step": 4000
|
33 |
},
|
34 |
{
|
35 |
"epoch": 7.5,
|
36 |
"eval_loss": 5.826813220977783,
|
37 |
-
"eval_runtime":
|
38 |
-
"eval_samples_per_second":
|
39 |
-
"eval_steps_per_second":
|
40 |
"step": 6000
|
41 |
},
|
42 |
{
|
@@ -49,17 +49,17 @@
|
|
49 |
{
|
50 |
"epoch": 10.0,
|
51 |
"eval_loss": 5.744414806365967,
|
52 |
-
"eval_runtime":
|
53 |
-
"eval_samples_per_second":
|
54 |
-
"eval_steps_per_second":
|
55 |
"step": 8000
|
56 |
},
|
57 |
{
|
58 |
"epoch": 12.5,
|
59 |
"eval_loss": 5.670751571655273,
|
60 |
-
"eval_runtime":
|
61 |
-
"eval_samples_per_second":
|
62 |
-
"eval_steps_per_second":
|
63 |
"step": 10000
|
64 |
},
|
65 |
{
|
@@ -72,9 +72,9 @@
|
|
72 |
{
|
73 |
"epoch": 15.0,
|
74 |
"eval_loss": 5.57132625579834,
|
75 |
-
"eval_runtime":
|
76 |
-
"eval_samples_per_second":
|
77 |
-
"eval_steps_per_second":
|
78 |
"step": 12000
|
79 |
}
|
80 |
],
|
|
|
11 |
{
|
12 |
"epoch": 2.5,
|
13 |
"eval_loss": 7.232339382171631,
|
14 |
+
"eval_runtime": 3.1848,
|
15 |
+
"eval_samples_per_second": 648.395,
|
16 |
+
"eval_steps_per_second": 40.819,
|
17 |
"step": 2000
|
18 |
},
|
19 |
{
|
|
|
26 |
{
|
27 |
"epoch": 5.0,
|
28 |
"eval_loss": 5.976784706115723,
|
29 |
+
"eval_runtime": 3.0375,
|
30 |
+
"eval_samples_per_second": 679.836,
|
31 |
+
"eval_steps_per_second": 42.798,
|
32 |
"step": 4000
|
33 |
},
|
34 |
{
|
35 |
"epoch": 7.5,
|
36 |
"eval_loss": 5.826813220977783,
|
37 |
+
"eval_runtime": 2.9574,
|
38 |
+
"eval_samples_per_second": 698.238,
|
39 |
+
"eval_steps_per_second": 43.957,
|
40 |
"step": 6000
|
41 |
},
|
42 |
{
|
|
|
49 |
{
|
50 |
"epoch": 10.0,
|
51 |
"eval_loss": 5.744414806365967,
|
52 |
+
"eval_runtime": 2.5747,
|
53 |
+
"eval_samples_per_second": 802.048,
|
54 |
+
"eval_steps_per_second": 50.492,
|
55 |
"step": 8000
|
56 |
},
|
57 |
{
|
58 |
"epoch": 12.5,
|
59 |
"eval_loss": 5.670751571655273,
|
60 |
+
"eval_runtime": 1.618,
|
61 |
+
"eval_samples_per_second": 1276.247,
|
62 |
+
"eval_steps_per_second": 80.345,
|
63 |
"step": 10000
|
64 |
},
|
65 |
{
|
|
|
72 |
{
|
73 |
"epoch": 15.0,
|
74 |
"eval_loss": 5.57132625579834,
|
75 |
+
"eval_runtime": 2.872,
|
76 |
+
"eval_samples_per_second": 719.021,
|
77 |
+
"eval_steps_per_second": 45.265,
|
78 |
"step": 12000
|
79 |
}
|
80 |
],
|