s2_hard_01: checkpoints exit
Browse files
runs/s2_hard_01/best/lora_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0798ce73a561039047453cb8c656189c211496c7ea1ee37db6478caad06ef867
|
| 3 |
+
size 147843609
|
runs/s2_hard_01/best/twin_config.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "Qwen/Qwen2.5-Math-1.5B",
|
| 3 |
+
"bidirectional": false,
|
| 4 |
+
"teacher_style": "solution_after_raw",
|
| 5 |
+
"lora_rank": 32,
|
| 6 |
+
"lora_alpha": 32.0,
|
| 7 |
+
"step": 50,
|
| 8 |
+
"val_metrics": {
|
| 9 |
+
"ce_encoder": 0.14390980980651677,
|
| 10 |
+
"ce_natural": 0.17520572880456878,
|
| 11 |
+
"ce_delta": -0.03129591899805202,
|
| 12 |
+
"ce_delta_easy": -0.03396672226074433,
|
| 13 |
+
"ce_delta_medium": -0.032309787887854886,
|
| 14 |
+
"ce_delta_very_hard": -0.020300319589386828,
|
| 15 |
+
"ce_delta_hard": -0.028107164460473462
|
| 16 |
+
}
|
| 17 |
+
}
|
runs/s2_hard_01/latest/lora_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74edb91aedfcef3b6dcef18b41184345822a42aff28707895091c923cd0517fc
|
| 3 |
+
size 147843609
|
runs/s2_hard_01/latest/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:64a6ae1e2d11eae184a792dfd8246500e9032a79c3dbd30fa6e771b696a89569
|
| 3 |
+
size 286123933
|
runs/s2_hard_01/latest/rng.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f145758ba06e80f6c32eded59ffefc15919e51ccec064da0aa31307d1cc6a4b0
|
| 3 |
+
size 10565
|
runs/s2_hard_01/latest/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:83e1374ad5651fb5b46e846db61713c5ed408681167d4d52ef840ffcd76759de
|
| 3 |
+
size 1401
|
runs/s2_hard_01/latest/twin_config.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "Qwen/Qwen2.5-Math-1.5B",
|
| 3 |
+
"bidirectional": false,
|
| 4 |
+
"teacher_style": "solution_after_raw",
|
| 5 |
+
"lora_rank": 32,
|
| 6 |
+
"lora_alpha": 32.0,
|
| 7 |
+
"step": 150,
|
| 8 |
+
"val_metrics": {
|
| 9 |
+
"ce_encoder": 0.18950135618473452,
|
| 10 |
+
"ce_natural": 0.17520572880456878,
|
| 11 |
+
"ce_delta": 0.014295627380165718,
|
| 12 |
+
"ce_delta_easy": 0.011189128311246991,
|
| 13 |
+
"ce_delta_medium": 0.015641754842955796,
|
| 14 |
+
"ce_delta_very_hard": 0.027599042689628015,
|
| 15 |
+
"ce_delta_hard": 0.01267043098360615
|
| 16 |
+
}
|
| 17 |
+
}
|