add vs1/ckpt/Gclean
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- vs1/ckpt/Gclean/step_100/heads.pt +3 -0
- vs1/ckpt/Gclean/step_100/meta.json +97 -0
- vs1/ckpt/Gclean/step_100/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_100/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_100/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_125/heads.pt +3 -0
- vs1/ckpt/Gclean/step_125/meta.json +97 -0
- vs1/ckpt/Gclean/step_125/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_125/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_125/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_150/heads.pt +3 -0
- vs1/ckpt/Gclean/step_150/meta.json +97 -0
- vs1/ckpt/Gclean/step_150/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_150/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_150/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_175/heads.pt +3 -0
- vs1/ckpt/Gclean/step_175/meta.json +97 -0
- vs1/ckpt/Gclean/step_175/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_175/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_175/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_200/heads.pt +3 -0
- vs1/ckpt/Gclean/step_200/meta.json +97 -0
- vs1/ckpt/Gclean/step_200/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_200/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_200/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_225/heads.pt +3 -0
- vs1/ckpt/Gclean/step_225/meta.json +97 -0
- vs1/ckpt/Gclean/step_225/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_225/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_225/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_25/heads.pt +3 -0
- vs1/ckpt/Gclean/step_25/meta.json +97 -0
- vs1/ckpt/Gclean/step_25/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_25/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_25/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_250/heads.pt +3 -0
- vs1/ckpt/Gclean/step_250/meta.json +97 -0
- vs1/ckpt/Gclean/step_250/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_250/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_250/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_275/heads.pt +3 -0
- vs1/ckpt/Gclean/step_275/meta.json +97 -0
- vs1/ckpt/Gclean/step_275/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_275/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_275/trainable_params.pt +3 -0
- vs1/ckpt/Gclean/step_300/heads.pt +3 -0
- vs1/ckpt/Gclean/step_300/meta.json +97 -0
- vs1/ckpt/Gclean/step_300/new_rows.pt +3 -0
- vs1/ckpt/Gclean/step_300/optim_state.pt +3 -0
- vs1/ckpt/Gclean/step_300/trainable_params.pt +3 -0
vs1/ckpt/Gclean/step_100/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b557e9a99606685e0122cb3d35469288e2eff9cf0e569f2d0f4c47ae7c83a61
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_100/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 100,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 02:04:07 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_100/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a565e803e26fe5e9b265bb982932e14c23fba79d34f7dbc749a6ba4ffd8b8eb3
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_100/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bed5e4fa3a42af68db7f6ca0842a2c1b1c66d29a35a5aaa6f3a975b5e6552cb6
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_100/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:52a38a231d824827b4aa4a603938e8a6e4864d773872bf1b9335afd029837d0c
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_125/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0e52e9c4c57c89cb0f7d8d3541b5048064b79461a9f21167e98d9053eb70fc0
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_125/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 125,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 02:15:21 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_125/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e28146c625a2ecad051a92632b05db8a1940c50e2a95f80496108e4429eaca82
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_125/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5312f55a0fda24608d1bb9ed9d9bb5aacd25fa07365fdf7c2097cf672b4e4603
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_125/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22351a5a351b986212820f60930ea5f355f28643393b32d0b4cb15fd4fe08c97
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_150/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c05455427ff0104c32b20786a30814185c57913aca50716e9eb2a330f5297ac
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_150/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 150,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 02:26:13 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_150/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1779386416c91f34e70e65d49077f9aacd438535ac66ac61dcd355acac6b2b5
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_150/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d29f6da0ed568ba22484dd0dfa05588fb12cc2e5c72b45c4cb4f5587ffd6e99c
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_150/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b160b00a909f545a0364c83211935420420c0088ad2334b44a053118796ebc08
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_175/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:54377bfe5cdb67bcd9ae0ba39aea4084b1a234dce163c210de843809ee8b2ab2
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_175/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 175,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 02:37:19 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_175/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b152146c526f18074949419cd17b98dd7c62214d70c4bf4dad3514bfed8b5181
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_175/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:92e60474de047aeea4615864cd7bf8d6c3ce11c9b846f87e02f397dacecd3c96
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_175/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:26d83e3cfa5b00c73768267e67b74089a1ea61cd86ce61b18e650e5fc2704b7f
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_200/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5618f8cd21af5e75275fff8cb792f504a2a483b366ec177413762f32a91a6876
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_200/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 200,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 02:48:24 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_200/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d2b398c253e9f66c34443543dc4402e504341cfd368973761c9124944c6307ba
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_200/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:016d856e8369058d60358287baa95377ad179c68db6b600f0ecaae299783e845
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_200/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff9ed42533b7dbf57be186633fb9fa0dabc686d78bbf7685fbf4e39e41f75e8c
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_225/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de2b59ac89abef7bb0e5a2f1e47eb0dfce0080f94a97e5f3fd1f7dfc834e0462
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_225/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 225,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 02:59:28 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_225/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:65bd084dbf72f7ea4ae73cae1120aef53689b2bdaf7a0bf13940d41bcd315d24
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_225/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5706e57c5903000b09e33e9c1cf603cd23253c95d865504684ee1644781c812
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_225/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b18a3f3d407ada41e88ba972cb41b6347a6b63bd8c548058703e3a82b9b3b632
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_25/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:105bae569398534a00f1ca330705fc0738288f0060b7ea755787e61842c4a5d3
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_25/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 25,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 01:41:31 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_25/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b78e560f72031c03fb3fd769f9bafeef416ddb22407ef34620c4bbc103efb56
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_25/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:99d4e622f2480ebf85e69e362c88e198c1f24fb06430a2b1dab1cfc63c6c3506
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_25/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2c7f43f8d957cdc6321119af8c10e4b1139f9c5a26a91b82a49beec4d0e5c8e6
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_250/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ba4e01747650abfb96e8509beaecb7fed2e59eb391700e4ea88d256d9353f00
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_250/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 250,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 03:11:08 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_250/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:241087e9bb1fcca189e7dd12079922d6eb2d59e46b8384a328c70048431c27d2
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_250/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa1312f14dc636fadef8e9d1e3d1e1aa9abce2b25c85a010e117680c04541f1c
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_250/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b608561f23770bdc8f8dee399a390de34ba9b760cce04fce8160f96bcffecec4
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_275/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95ddab7c6a01970eebf886b5f00a6f27a8ead44e75059d2cea7da78a0cb1224e
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_275/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 275,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 03:22:35 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_275/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b8676f11d16b307f66a56eeb6bb0df213ab17c1085881f49a72b6bbbe2e75a04
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_275/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e77347ab526d2d92971595835cc3dbdd1db420aace0381648effcec7f07dda84
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_275/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d2806e88328d5aed126f14a2cc45e4fa393f79f0ce1b766f53eb039fc0e18b3e
|
| 3 |
+
size 132339780
|
vs1/ckpt/Gclean/step_300/heads.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e094264358f6eeb737c0f8cae8bd2bd659f0411fc37f45604f161f50f240f5ad
|
| 3 |
+
size 46299566
|
vs1/ckpt/Gclean/step_300/meta.json
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "Gclean",
|
| 3 |
+
"step": 300,
|
| 4 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 5 |
+
"lam": 1.0,
|
| 6 |
+
"p1": false,
|
| 7 |
+
"run_name": "Gclean",
|
| 8 |
+
"time": "2026-10-23 03:33:53 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": 625,
|
| 54 |
+
"order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
|
| 55 |
+
"data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
|
| 56 |
+
"data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
|
| 57 |
+
"ckpt_name": "Gclean",
|
| 58 |
+
"git_commit": "205619c04e5ad473afc7df31a8e3bea59efb9858",
|
| 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": 25,
|
| 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/step_300/new_rows.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:888c55711a0c0d511939813e14976bc47e22c892e57f23795bd0162a17bf64f2
|
| 3 |
+
size 17072
|
vs1/ckpt/Gclean/step_300/optim_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:149f3f8182752d8b7d4c8944e45f7b445cab98a13b4c5ba6203afb4adbebe737
|
| 3 |
+
size 357349235
|
vs1/ckpt/Gclean/step_300/trainable_params.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:85f9a5b455eaed216f1c550aa2b713d2b7df59933d2d43da47ce4c40cefbaf69
|
| 3 |
+
size 132339780
|