Upload folder using huggingface_hub
Browse files- ckpt-step=002500-metric=0.00242256.pt +3 -0
- ckpt-step=003000-metric=0.00222944.pt +3 -0
- ckpt-step=003500-metric=0.00185913.pt +3 -0
- ckpt-step=004000-metric=0.00168045.pt +3 -0
- ckpt-step=004500-metric=0.00182081.pt +3 -0
- ckpt_index.json +38 -0
- inference.yaml +37 -0
- model.pt +3 -0
ckpt-step=002500-metric=0.00242256.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3ec51aaa9cf4c29ef89ec0ca19354f692deb55927164b602d60f6fbe9391070b
|
| 3 |
+
size 81908399
|
ckpt-step=003000-metric=0.00222944.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aee6cc2933e08912d2253dc0ccf32174f0ed8e42e4d4ee4d0407371173e02858
|
| 3 |
+
size 81908399
|
ckpt-step=003500-metric=0.00185913.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:694a5dd7167de57f0005d6dfbff019367f2caed24c9cf3a7f72b359ba3d8e125
|
| 3 |
+
size 81908399
|
ckpt-step=004000-metric=0.00168045.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e19af8b64d8da8d3c3f1664d465c5e6c03d3aeccd29f6650d04c88b3b00c1221
|
| 3 |
+
size 81908399
|
ckpt-step=004500-metric=0.00182081.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5143335a627d2c4bac4aa8540bf2d5fcd6a2fafd893d0b344084ff5bde9d3b8
|
| 3 |
+
size 81908399
|
ckpt_index.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"k": 5,
|
| 3 |
+
"mode": "min",
|
| 4 |
+
"metric_name": "loss",
|
| 5 |
+
"entries": [
|
| 6 |
+
{
|
| 7 |
+
"path": "experiments/jolly-dust-782/ckpt-step=004000-metric=0.00168045.pt",
|
| 8 |
+
"step": 4000,
|
| 9 |
+
"metric": 0.001680454245563507,
|
| 10 |
+
"ts": 1764635981.863727
|
| 11 |
+
},
|
| 12 |
+
{
|
| 13 |
+
"path": "experiments/jolly-dust-782/ckpt-step=004500-metric=0.00182081.pt",
|
| 14 |
+
"step": 4500,
|
| 15 |
+
"metric": 0.0018208065658608286,
|
| 16 |
+
"ts": 1764636370.650296
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"path": "experiments/jolly-dust-782/ckpt-step=003500-metric=0.00185913.pt",
|
| 20 |
+
"step": 3500,
|
| 21 |
+
"metric": 0.001859126122928864,
|
| 22 |
+
"ts": 1764635592.6594105
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"path": "experiments/jolly-dust-782/ckpt-step=003000-metric=0.00222944.pt",
|
| 26 |
+
"step": 3000,
|
| 27 |
+
"metric": 0.0022294362020048313,
|
| 28 |
+
"ts": 1764635203.0245426
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"path": "experiments/jolly-dust-782/ckpt-step=002500-metric=0.00242256.pt",
|
| 32 |
+
"step": 2500,
|
| 33 |
+
"metric": 0.0024225555885234594,
|
| 34 |
+
"ts": 1764634813.81682
|
| 35 |
+
}
|
| 36 |
+
],
|
| 37 |
+
"updated_at": 1764636371.449813
|
| 38 |
+
}
|
inference.yaml
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model:
|
| 2 |
+
model_id: "dit"
|
| 3 |
+
width: 24
|
| 4 |
+
height: 24
|
| 5 |
+
T: 1000
|
| 6 |
+
in_channels: 3
|
| 7 |
+
n_window: 30
|
| 8 |
+
patch_size: 3
|
| 9 |
+
n_heads: 20
|
| 10 |
+
d_model: 320
|
| 11 |
+
n_blocks: 8
|
| 12 |
+
C: 5000
|
| 13 |
+
bidirectional: false
|
| 14 |
+
nocompile: false
|
| 15 |
+
checkpoint: "experiments/jolly-dust-782"
|
| 16 |
+
rope_type: "rope"
|
| 17 |
+
use_flex: true
|
| 18 |
+
|
| 19 |
+
train:
|
| 20 |
+
lr1: 0.02
|
| 21 |
+
lr2: 3.0e-4
|
| 22 |
+
betas: [0.9, 0.95]
|
| 23 |
+
weight_decay: 1.0e-5
|
| 24 |
+
max_steps: 2500
|
| 25 |
+
warmup_steps: 100
|
| 26 |
+
batch_size: 64
|
| 27 |
+
noclip: false
|
| 28 |
+
duration: 1
|
| 29 |
+
fps: 30
|
| 30 |
+
debug: false
|
| 31 |
+
dtype: "bf16"
|
| 32 |
+
action_dropout: 0.2
|
| 33 |
+
|
| 34 |
+
wandb:
|
| 35 |
+
name: null
|
| 36 |
+
project: "toy-wm"
|
| 37 |
+
run_name: "causal-layers8-heads12-d384"
|
model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:27e06f0da92b9309122762df8e4f1819f63b1a34d99e27c68b9cd360c740cf26
|
| 3 |
+
size 39833499
|