add vs1/ckpt/Gclean_smoke
Browse files- vs1/ckpt/Gclean_smoke/step_2/heads.pt +3 -0
- vs1/ckpt/Gclean_smoke/step_2/meta.json +97 -0
- vs1/ckpt/Gclean_smoke/step_2/new_rows.pt +3 -0
- vs1/ckpt/Gclean_smoke/step_2/optim_state.pt +3 -0
- vs1/ckpt/Gclean_smoke/step_2/trainable_params.pt +3 -0
- vs1/ckpt/Gclean_smoke/step_3/heads.pt +3 -0
- vs1/ckpt/Gclean_smoke/step_3/meta.json +97 -0
- vs1/ckpt/Gclean_smoke/step_3/new_rows.pt +3 -0
- vs1/ckpt/Gclean_smoke/step_3/optim_state.pt +3 -0
- vs1/ckpt/Gclean_smoke/step_3/trainable_params.pt +3 -0
vs1/ckpt/Gclean_smoke/step_2/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7414a1491111937460c83d89a60e2f74fe76e5b6ebaf860c25fe9e3056d8accf
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean_smoke/step_2/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 2,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 00:18:27 KST",
|
| 9 |
+
"cfg": {
|
| 10 |
+
"variant": "Gclean",
|
| 11 |
+
"base_variant": "G",
|
| 12 |
+
"fmt": "clean (지시서 53 §3, 49 §4.3)",
|
| 13 |
+
"p1": false,
|
| 14 |
+
"p1_ids": null,
|
| 15 |
+
"lora_r": 16,
|
| 16 |
+
"lora_alpha": 32,
|
| 17 |
+
"lora_dropout": 0.05,
|
| 18 |
+
"lora_targets": [
|
| 19 |
+
"q_proj",
|
| 20 |
+
"k_proj",
|
| 21 |
+
"v_proj",
|
| 22 |
+
"o_proj",
|
| 23 |
+
"gate_proj",
|
| 24 |
+
"up_proj",
|
| 25 |
+
"down_proj"
|
| 26 |
+
],
|
| 27 |
+
"lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
|
| 28 |
+
"micro_batch": 1,
|
| 29 |
+
"grad_accum": 8,
|
| 30 |
+
"lr": 0.0001,
|
| 31 |
+
"head_lr": 0.0001,
|
| 32 |
+
"new_token_lr": 0.0001,
|
| 33 |
+
"scheduler": "constant, linear warmup 10 steps (lr_k = 1e-4·min(1,k/10))",
|
| 34 |
+
"warmup_steps": 10,
|
| 35 |
+
"weight_decay": 0.0,
|
| 36 |
+
"max_grad_norm": 1.0,
|
| 37 |
+
"optimizer": "AdamW",
|
| 38 |
+
"seed": 0,
|
| 39 |
+
"bf16": true,
|
| 40 |
+
"grad_ckpt": true,
|
| 41 |
+
"frame_rule": "D16",
|
| 42 |
+
"n_input_frames": 32,
|
| 43 |
+
"grid_t": 16,
|
| 44 |
+
"frame_hw": [
|
| 45 |
+
480,
|
| 46 |
+
640
|
| 47 |
+
],
|
| 48 |
+
"base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
|
| 49 |
+
"processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
|
| 50 |
+
"lam": 1.0,
|
| 51 |
+
"n_samples": 5000,
|
| 52 |
+
"n_steps_full": 625,
|
| 53 |
+
"steps_this_run": 3,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean_smoke",
|
| 58 |
+
"git_commit": "2440a377907a57e6edb492efc284b87e0d8a369f",
|
| 59 |
+
"workers": 4,
|
| 60 |
+
"new_tokens": [
|
| 61 |
+
"<box>",
|
| 62 |
+
"<cam>",
|
| 63 |
+
"<rel>"
|
| 64 |
+
],
|
| 65 |
+
"new_token_ids": [
|
| 66 |
+
151669,
|
| 67 |
+
151670,
|
| 68 |
+
151671
|
| 69 |
+
],
|
| 70 |
+
"heads": [
|
| 71 |
+
"box",
|
| 72 |
+
"cam(GeoVR CameraDec)",
|
| 73 |
+
"rel"
|
| 74 |
+
],
|
| 75 |
+
"cam_init": "GeoVR model.camera_token.weight[1]",
|
| 76 |
+
"cam_head": "GeoVR camera_head (CameraDec 2560, fp32, ckpt)",
|
| 77 |
+
"cam_target": "gconv.cam_g: first-frame w2c [s·t, quat xyzw w>=0], s=VGGT scale",
|
| 78 |
+
"cam_loss": "|dt|.clamp(100).mean + min(|q̂-q|,|q̂+q|).mean, no FoV, no 2.5x",
|
| 79 |
+
"scale_s1": false,
|
| 80 |
+
"scale_file_sha256": "9e2f0a5abe5f1dbc4f4f6a074fd5584f668c209302f997b778e373bfea42bb95",
|
| 81 |
+
"cam_mask": {
|
| 82 |
+
"n_pair": 4000,
|
| 83 |
+
"n_valid": 4000,
|
| 84 |
+
"frame0_invalid": 0,
|
| 85 |
+
"scale_none": 0,
|
| 86 |
+
"scale_missing_scene": 0,
|
| 87 |
+
"frames_mismatch": 0,
|
| 88 |
+
"s1_used": false,
|
| 89 |
+
"n_masked": 0
|
| 90 |
+
},
|
| 91 |
+
"cam_mode": "untied",
|
| 92 |
+
"ckpt_every": 2,
|
| 93 |
+
"curve_every": 25,
|
| 94 |
+
"curve_ids": "/workspace/code/spatial_reasoning/VS1/K/data/g_curve_ids.json",
|
| 95 |
+
"attn": "flash_attention_2"
|
| 96 |
+
}
|
| 97 |
+
}
|
vs1/ckpt/Gclean_smoke/step_2/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4932917ea05135c3c5d2748217c490eb0be23bd3c8b07a42bc1000edfc3de56b
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean_smoke/step_2/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:46ba76a29987ccd82c269df86e9d3b17ffca699fd4f6af4c2cd886f8e3f29ef9
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean_smoke/step_2/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:222de2bcdaf75a2abc2b6612c792eed4d62716edeeae849f93fa3db8057f08f3
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean_smoke/step_3/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ecaa93324fc763e0d7a68309a1f3feaffc4f05e6c6f5c0e07c9c0f1fa188c4e4
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean_smoke/step_3/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 3,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 00:19:00 KST",
|
| 9 |
+
"cfg": {
|
| 10 |
+
"variant": "Gclean",
|
| 11 |
+
"base_variant": "G",
|
| 12 |
+
"fmt": "clean (지시서 53 §3, 49 §4.3)",
|
| 13 |
+
"p1": false,
|
| 14 |
+
"p1_ids": null,
|
| 15 |
+
"lora_r": 16,
|
| 16 |
+
"lora_alpha": 32,
|
| 17 |
+
"lora_dropout": 0.05,
|
| 18 |
+
"lora_targets": [
|
| 19 |
+
"q_proj",
|
| 20 |
+
"k_proj",
|
| 21 |
+
"v_proj",
|
| 22 |
+
"o_proj",
|
| 23 |
+
"gate_proj",
|
| 24 |
+
"up_proj",
|
| 25 |
+
"down_proj"
|
| 26 |
+
],
|
| 27 |
+
"lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
|
| 28 |
+
"micro_batch": 1,
|
| 29 |
+
"grad_accum": 8,
|
| 30 |
+
"lr": 0.0001,
|
| 31 |
+
"head_lr": 0.0001,
|
| 32 |
+
"new_token_lr": 0.0001,
|
| 33 |
+
"scheduler": "constant, linear warmup 10 steps (lr_k = 1e-4·min(1,k/10))",
|
| 34 |
+
"warmup_steps": 10,
|
| 35 |
+
"weight_decay": 0.0,
|
| 36 |
+
"max_grad_norm": 1.0,
|
| 37 |
+
"optimizer": "AdamW",
|
| 38 |
+
"seed": 0,
|
| 39 |
+
"bf16": true,
|
| 40 |
+
"grad_ckpt": true,
|
| 41 |
+
"frame_rule": "D16",
|
| 42 |
+
"n_input_frames": 32,
|
| 43 |
+
"grid_t": 16,
|
| 44 |
+
"frame_hw": [
|
| 45 |
+
480,
|
| 46 |
+
640
|
| 47 |
+
],
|
| 48 |
+
"base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
|
| 49 |
+
"processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
|
| 50 |
+
"lam": 1.0,
|
| 51 |
+
"n_samples": 5000,
|
| 52 |
+
"n_steps_full": 625,
|
| 53 |
+
"steps_this_run": 3,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean_smoke",
|
| 58 |
+
"git_commit": "2440a377907a57e6edb492efc284b87e0d8a369f",
|
| 59 |
+
"workers": 4,
|
| 60 |
+
"new_tokens": [
|
| 61 |
+
"<box>",
|
| 62 |
+
"<cam>",
|
| 63 |
+
"<rel>"
|
| 64 |
+
],
|
| 65 |
+
"new_token_ids": [
|
| 66 |
+
151669,
|
| 67 |
+
151670,
|
| 68 |
+
151671
|
| 69 |
+
],
|
| 70 |
+
"heads": [
|
| 71 |
+
"box",
|
| 72 |
+
"cam(GeoVR CameraDec)",
|
| 73 |
+
"rel"
|
| 74 |
+
],
|
| 75 |
+
"cam_init": "GeoVR model.camera_token.weight[1]",
|
| 76 |
+
"cam_head": "GeoVR camera_head (CameraDec 2560, fp32, ckpt)",
|
| 77 |
+
"cam_target": "gconv.cam_g: first-frame w2c [s·t, quat xyzw w>=0], s=VGGT scale",
|
| 78 |
+
"cam_loss": "|dt|.clamp(100).mean + min(|q̂-q|,|q̂+q|).mean, no FoV, no 2.5x",
|
| 79 |
+
"scale_s1": false,
|
| 80 |
+
"scale_file_sha256": "9e2f0a5abe5f1dbc4f4f6a074fd5584f668c209302f997b778e373bfea42bb95",
|
| 81 |
+
"cam_mask": {
|
| 82 |
+
"n_pair": 4000,
|
| 83 |
+
"n_valid": 4000,
|
| 84 |
+
"frame0_invalid": 0,
|
| 85 |
+
"scale_none": 0,
|
| 86 |
+
"scale_missing_scene": 0,
|
| 87 |
+
"frames_mismatch": 0,
|
| 88 |
+
"s1_used": false,
|
| 89 |
+
"n_masked": 0
|
| 90 |
+
},
|
| 91 |
+
"cam_mode": "untied",
|
| 92 |
+
"ckpt_every": 2,
|
| 93 |
+
"curve_every": 25,
|
| 94 |
+
"curve_ids": "/workspace/code/spatial_reasoning/VS1/K/data/g_curve_ids.json",
|
| 95 |
+
"attn": "flash_attention_2"
|
| 96 |
+
}
|
| 97 |
+
}
|
vs1/ckpt/Gclean_smoke/step_3/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d57cfb30ac7c0749b396b82c1cacde2780d08ae450683374ad66be01c0b45130
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean_smoke/step_3/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b2f25ae9cf2a7ebb50d21f3a21ae61c9ab7a0ccec226e4f7a6aaab7a75479b8
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean_smoke/step_3/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b374f01905caea70155e98e19ef182a6997b9a46ea34c09cae971e21eef61752
|
| 3 |
+
size 132339780
|