rgen1: checkpoints exit
Browse files
runs/rgen1/best/lora_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:feb22cb1b96d40d4c019c05af7ee8b29c379ef05c63ee10e5e5ef4c5401f90a5
|
| 3 |
+
size 147843609
|
runs/rgen1/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": 60,
|
| 8 |
+
"val_metrics": {
|
| 9 |
+
"ce_encoder": 0.12605639823971915,
|
| 10 |
+
"ce_natural": 0.17520572880456878,
|
| 11 |
+
"ce_delta": -0.04914933056484964,
|
| 12 |
+
"ce_delta_easy": -0.05533516131767687,
|
| 13 |
+
"ce_delta_medium": -0.04799248812991329,
|
| 14 |
+
"ce_delta_very_hard": -0.033994804569068594,
|
| 15 |
+
"ce_delta_hard": -0.04345498918232876
|
| 16 |
+
}
|
| 17 |
+
}
|
runs/rgen1/latest/lora_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bdb9781bbb3ce1f2f038cd33d9ba075087e7f36d8b4d11e62f43534661a94b9a
|
| 3 |
+
size 147843609
|
runs/rgen1/latest/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0e8e866afea261ca5bfab510ccdc49b6796e0e1c74c7f85a2c33b2e4444213d6
|
| 3 |
+
size 286123933
|
runs/rgen1/latest/rng.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4a7f2dd3586282df0da3a52292b030e6c923f68d346687e9d9afd0cb1bdcc51d
|
| 3 |
+
size 10565
|
runs/rgen1/latest/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:52505970ce43f53a04aaadc70899b3b28a93790f8dcd6722810fd182d18a9d25
|
| 3 |
+
size 1401
|
runs/rgen1/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": 90,
|
| 8 |
+
"val_metrics": {
|
| 9 |
+
"ce_encoder": 0.12770924802687864,
|
| 10 |
+
"ce_natural": 0.17520572880456878,
|
| 11 |
+
"ce_delta": -0.047496480777690156,
|
| 12 |
+
"ce_delta_easy": -0.05319544801534509,
|
| 13 |
+
"ce_delta_medium": -0.04655357021234667,
|
| 14 |
+
"ce_delta_very_hard": -0.03203565688906777,
|
| 15 |
+
"ce_delta_hard": -0.042780363069756154
|
| 16 |
+
}
|
| 17 |
+
}
|