rspec2: checkpoints exit
Browse files
runs/rspec2/best/lora_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38208e1fab744c3e7adf03dd764eeaaf9d9fdf9ca6496968b8342994ffe7b321
|
| 3 |
+
size 147843609
|
runs/rspec2/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": 20,
|
| 8 |
+
"val_metrics": {
|
| 9 |
+
"ce_encoder": 0.14274850256142022,
|
| 10 |
+
"ce_natural": 0.17520572880456878,
|
| 11 |
+
"ce_delta": -0.03245722624314857,
|
| 12 |
+
"ce_delta_easy": -0.03531153876210209,
|
| 13 |
+
"ce_delta_medium": -0.032570773001432095,
|
| 14 |
+
"ce_delta_very_hard": -0.0257347858431911,
|
| 15 |
+
"ce_delta_hard": -0.028390619848336764
|
| 16 |
+
}
|
| 17 |
+
}
|
runs/rspec2/latest/lora_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e66940e2deb262985706db4ccddf57b709bd1bb951ecb76c512abead1bf61dac
|
| 3 |
+
size 147843609
|
runs/rspec2/latest/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:26531cfee8127b6ce8ff61e6252aa3e1bdb15a1d6ef14076d4fc87459dc95944
|
| 3 |
+
size 286123933
|
runs/rspec2/latest/rng.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aa215bdf7b69e9ae2b3a82a590c4fee4542848a0d81cd6dbc3e5d185649d2392
|
| 3 |
+
size 10565
|
runs/rspec2/latest/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0e4ea0f0e1a17f85cf352d2215bcce26843e0a9bfa1b9f6389d0092a2fd883b
|
| 3 |
+
size 1401
|
runs/rspec2/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": 50,
|
| 8 |
+
"val_metrics": {
|
| 9 |
+
"ce_encoder": 0.1460093880269576,
|
| 10 |
+
"ce_natural": 0.17520572880456878,
|
| 11 |
+
"ce_delta": -0.029196340777611192,
|
| 12 |
+
"ce_delta_easy": -0.03022114414174982,
|
| 13 |
+
"ce_delta_medium": -0.030726013024121507,
|
| 14 |
+
"ce_delta_very_hard": -0.023034356623797003,
|
| 15 |
+
"ce_delta_hard": -0.026690403482462516
|
| 16 |
+
}
|
| 17 |
+
}
|