exp010: gr_head_apx
Browse files- .gitattributes +1 -0
- experiments/exp_010_component_potentials/runs/gr_head_apx/checkpoint.pt +3 -0
- experiments/exp_010_component_potentials/runs/gr_head_apx/checkpoint.pt.graft +3 -0
- experiments/exp_010_component_potentials/runs/gr_head_apx/result.json +37 -0
- experiments/exp_010_component_potentials/runs/gr_head_apx/snapshots.pt +3 -0
.gitattributes
CHANGED
|
@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
experiments/exp_010_component_potentials/runs/bk_swap_e2d/checkpoint.pt.swap filter=lfs diff=lfs merge=lfs -text
|
| 37 |
experiments/exp_010_component_potentials/runs/bk_swap_d2e/checkpoint.pt.swap filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
experiments/exp_010_component_potentials/runs/bk_swap_e2d/checkpoint.pt.swap filter=lfs diff=lfs merge=lfs -text
|
| 37 |
experiments/exp_010_component_potentials/runs/bk_swap_d2e/checkpoint.pt.swap filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
experiments/exp_010_component_potentials/runs/gr_head_apx/checkpoint.pt.graft filter=lfs diff=lfs merge=lfs -text
|
experiments/exp_010_component_potentials/runs/gr_head_apx/checkpoint.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bd36c2d7f39301afba8155a0eb9d25509ac85c4cb4c68907d9e7f94a1c3af2db
|
| 3 |
+
size 28205498
|
experiments/exp_010_component_potentials/runs/gr_head_apx/checkpoint.pt.graft
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8270554e5c003a7ec1631b09242667038d773b5e73f77a6f560f20e9930aca62
|
| 3 |
+
size 28065335
|
experiments/exp_010_component_potentials/runs/gr_head_apx/result.json
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"run": "gr_head_apx",
|
| 3 |
+
"family": "graft",
|
| 4 |
+
"dev_final": 0.008110761642456055,
|
| 5 |
+
"rank_final": 3.916935443878174,
|
| 6 |
+
"statute_final": "uniform",
|
| 7 |
+
"params": 6951239,
|
| 8 |
+
"minutes": 1.2,
|
| 9 |
+
"bpb": 2.9497957500444394,
|
| 10 |
+
"bpb_eval": NaN,
|
| 11 |
+
"coverage": 0.8038330078125,
|
| 12 |
+
"gate_acc": 0.8580322265625,
|
| 13 |
+
"mode_spread_deg": 84.7422936361316,
|
| 14 |
+
"mix_entropy": 0.9709887653217407,
|
| 15 |
+
"write_kl": 0.12128203362226486,
|
| 16 |
+
"kernel_aux": NaN,
|
| 17 |
+
"loss": 6.146056175231934,
|
| 18 |
+
"condition": {
|
| 19 |
+
"tag": "gr_head_apx",
|
| 20 |
+
"graft": "head",
|
| 21 |
+
"donor": "exp010/_donor/checkpoint.pt",
|
| 22 |
+
"dim": "384",
|
| 23 |
+
"n_layers": "4",
|
| 24 |
+
"n_heads": "6",
|
| 25 |
+
"K": "64",
|
| 26 |
+
"seq_len": "256",
|
| 27 |
+
"batch_size": "32",
|
| 28 |
+
"accum_steps": "1",
|
| 29 |
+
"d_point": "48",
|
| 30 |
+
"head": "hybrid",
|
| 31 |
+
"steps": "2000",
|
| 32 |
+
"log_every": "500",
|
| 33 |
+
"bank_scorer": "apmix",
|
| 34 |
+
"n_pointers": "4",
|
| 35 |
+
"bank_source": "atlas/bank_16384_tri.pt"
|
| 36 |
+
}
|
| 37 |
+
}
|
experiments/exp_010_component_potentials/runs/gr_head_apx/snapshots.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01a4f076a2eaf15f9f5076b78efddb2c6f8c9bf420c7835797ea6daeeaa53a23
|
| 3 |
+
size 9391
|