JH-C-k commited on
Commit
15b910e
·
verified ·
1 Parent(s): 931723a

add vs1/ckpt/G

Browse files
vs1/ckpt/G/100/heads.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1eaa1ec346959ec2ce295e6414d204db3e7b953f59ce126b8b5a62567181af0a
3
+ size 46299566
vs1/ckpt/G/100/meta.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "variant": "G",
3
+ "step": 625,
4
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
5
+ "lam": 1.0,
6
+ "run_name": "G",
7
+ "time": "2026-10-21 06:27:30 KST",
8
+ "cfg": {
9
+ "variant": "G",
10
+ "base_variant": "S3",
11
+ "lora_r": 16,
12
+ "lora_alpha": 32,
13
+ "lora_dropout": 0.05,
14
+ "lora_targets": [
15
+ "q_proj",
16
+ "k_proj",
17
+ "v_proj",
18
+ "o_proj",
19
+ "gate_proj",
20
+ "up_proj",
21
+ "down_proj"
22
+ ],
23
+ "lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
24
+ "micro_batch": 1,
25
+ "grad_accum": 8,
26
+ "lr": 0.0001,
27
+ "head_lr": 0.0001,
28
+ "new_token_lr": 0.0001,
29
+ "scheduler": "cosine",
30
+ "warmup_ratio": 0.03,
31
+ "warmup_steps": 19,
32
+ "weight_decay": 0.0,
33
+ "max_grad_norm": 1.0,
34
+ "optimizer": "AdamW",
35
+ "seed": 0,
36
+ "bf16": true,
37
+ "grad_ckpt": true,
38
+ "frame_rule": "D16",
39
+ "n_input_frames": 32,
40
+ "grid_t": 16,
41
+ "frame_hw": [
42
+ 480,
43
+ 640
44
+ ],
45
+ "base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
46
+ "processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
47
+ "lam": 1.0,
48
+ "epochs": 1,
49
+ "n_samples": 5000,
50
+ "n_steps_full": 625,
51
+ "steps_this_run": 625,
52
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
53
+ "data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
54
+ "data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
55
+ "ckpt_name": "G",
56
+ "git_commit": "b0457f9481f00c0deb596e0fe2caf37e278c01e2",
57
+ "workers": 4,
58
+ "new_tokens": [
59
+ "<box>",
60
+ "<cam>",
61
+ "<rel>"
62
+ ],
63
+ "new_token_ids": [
64
+ 151669,
65
+ 151670,
66
+ 151671
67
+ ],
68
+ "heads": [
69
+ "box",
70
+ "cam(GeoVR CameraDec)",
71
+ "rel"
72
+ ],
73
+ "cam_init": "GeoVR model.camera_token.weight[1]",
74
+ "cam_head": "GeoVR camera_head (CameraDec 2560, fp32, ckpt)",
75
+ "cam_target": "gconv.cam_g: first-frame w2c [s·t, quat xyzw w>=0], s=VGGT scale",
76
+ "cam_loss": "|dt|.clamp(100).mean + |dq|.mean, no FoV, no 2.5x",
77
+ "scale_s1": false,
78
+ "scale_file_sha256": "9e2f0a5abe5f1dbc4f4f6a074fd5584f668c209302f997b778e373bfea42bb95",
79
+ "cam_mask": {
80
+ "n_pair": 4000,
81
+ "n_valid": 4000,
82
+ "frame0_invalid": 0,
83
+ "scale_none": 0,
84
+ "scale_missing_scene": 0,
85
+ "frames_mismatch": 0,
86
+ "s1_used": false,
87
+ "n_masked": 0
88
+ },
89
+ "cam_mode": "untied",
90
+ "deviation_from_order": "지시서와 다른 곳: <cam> 입력 임베딩 분리(B안), 이유 = camera_token[1] 노름 51.05 vs 어휘 중앙값 1.15, tie 로 CE 23.5→31.4 (A안 스모크 VS1/K/logs/smoke_G_smoke_A.json)",
91
+ "curve_every": 25,
92
+ "curve_ids": "/workspace/code/spatial_reasoning/VS1/K/data/g_curve_ids.json",
93
+ "attn": "flash_attention_2"
94
+ }
95
+ }
vs1/ckpt/G/100/new_rows.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:113bcfce8ddeca995efe8d66aebf7f82c83a08b89a6bc09be056463b955a78ba
3
+ size 17072
vs1/ckpt/G/100/trainable_params.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:977f2a0195cd0eedb52da3bb349d7a07abac24dc4ff4e21e5c52b5d653b7f850
3
+ size 132339780
vs1/ckpt/G/50/heads.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5411dafa3dc870616654b2a5140f6df7f44b275645ffffdeddba608d62628c8a
3
+ size 46299566
vs1/ckpt/G/50/meta.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "variant": "G",
3
+ "step": 312,
4
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
5
+ "lam": 1.0,
6
+ "run_name": "G",
7
+ "time": "2026-10-21 04:56:19 KST",
8
+ "cfg": {
9
+ "variant": "G",
10
+ "base_variant": "S3",
11
+ "lora_r": 16,
12
+ "lora_alpha": 32,
13
+ "lora_dropout": 0.05,
14
+ "lora_targets": [
15
+ "q_proj",
16
+ "k_proj",
17
+ "v_proj",
18
+ "o_proj",
19
+ "gate_proj",
20
+ "up_proj",
21
+ "down_proj"
22
+ ],
23
+ "lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
24
+ "micro_batch": 1,
25
+ "grad_accum": 8,
26
+ "lr": 0.0001,
27
+ "head_lr": 0.0001,
28
+ "new_token_lr": 0.0001,
29
+ "scheduler": "cosine",
30
+ "warmup_ratio": 0.03,
31
+ "warmup_steps": 19,
32
+ "weight_decay": 0.0,
33
+ "max_grad_norm": 1.0,
34
+ "optimizer": "AdamW",
35
+ "seed": 0,
36
+ "bf16": true,
37
+ "grad_ckpt": true,
38
+ "frame_rule": "D16",
39
+ "n_input_frames": 32,
40
+ "grid_t": 16,
41
+ "frame_hw": [
42
+ 480,
43
+ 640
44
+ ],
45
+ "base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
46
+ "processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
47
+ "lam": 1.0,
48
+ "epochs": 1,
49
+ "n_samples": 5000,
50
+ "n_steps_full": 625,
51
+ "steps_this_run": 625,
52
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
53
+ "data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
54
+ "data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
55
+ "ckpt_name": "G",
56
+ "git_commit": "b0457f9481f00c0deb596e0fe2caf37e278c01e2",
57
+ "workers": 4,
58
+ "new_tokens": [
59
+ "<box>",
60
+ "<cam>",
61
+ "<rel>"
62
+ ],
63
+ "new_token_ids": [
64
+ 151669,
65
+ 151670,
66
+ 151671
67
+ ],
68
+ "heads": [
69
+ "box",
70
+ "cam(GeoVR CameraDec)",
71
+ "rel"
72
+ ],
73
+ "cam_init": "GeoVR model.camera_token.weight[1]",
74
+ "cam_head": "GeoVR camera_head (CameraDec 2560, fp32, ckpt)",
75
+ "cam_target": "gconv.cam_g: first-frame w2c [s·t, quat xyzw w>=0], s=VGGT scale",
76
+ "cam_loss": "|dt|.clamp(100).mean + |dq|.mean, no FoV, no 2.5x",
77
+ "scale_s1": false,
78
+ "scale_file_sha256": "9e2f0a5abe5f1dbc4f4f6a074fd5584f668c209302f997b778e373bfea42bb95",
79
+ "cam_mask": {
80
+ "n_pair": 4000,
81
+ "n_valid": 4000,
82
+ "frame0_invalid": 0,
83
+ "scale_none": 0,
84
+ "scale_missing_scene": 0,
85
+ "frames_mismatch": 0,
86
+ "s1_used": false,
87
+ "n_masked": 0
88
+ },
89
+ "cam_mode": "untied",
90
+ "deviation_from_order": "지시서와 다른 곳: <cam> 입력 임베딩 분리(B안), 이유 = camera_token[1] 노름 51.05 vs 어휘 중앙값 1.15, tie 로 CE 23.5→31.4 (A안 스모크 VS1/K/logs/smoke_G_smoke_A.json)",
91
+ "curve_every": 25,
92
+ "curve_ids": "/workspace/code/spatial_reasoning/VS1/K/data/g_curve_ids.json",
93
+ "attn": "flash_attention_2"
94
+ }
95
+ }
vs1/ckpt/G/50/new_rows.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a44b3ac057510ebcf660fac2ab5c7104a8e166573ff73630bf74885ea1ba6da
3
+ size 17072
vs1/ckpt/G/50/trainable_params.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:605d9a8de52c42a530afc7750a55f1332933961de101cd118838d0aefd6853af
3
+ size 132339780
vs1/ckpt/G/resume/heads.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1eaa1ec346959ec2ce295e6414d204db3e7b953f59ce126b8b5a62567181af0a
3
+ size 46299566
vs1/ckpt/G/resume/meta.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "variant": "G",
3
+ "step": 625,
4
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
5
+ "lam": 1.0,
6
+ "run_name": "G",
7
+ "time": "2026-10-21 06:27:30 KST",
8
+ "cfg": {
9
+ "variant": "G",
10
+ "base_variant": "S3",
11
+ "lora_r": 16,
12
+ "lora_alpha": 32,
13
+ "lora_dropout": 0.05,
14
+ "lora_targets": [
15
+ "q_proj",
16
+ "k_proj",
17
+ "v_proj",
18
+ "o_proj",
19
+ "gate_proj",
20
+ "up_proj",
21
+ "down_proj"
22
+ ],
23
+ "lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
24
+ "micro_batch": 1,
25
+ "grad_accum": 8,
26
+ "lr": 0.0001,
27
+ "head_lr": 0.0001,
28
+ "new_token_lr": 0.0001,
29
+ "scheduler": "cosine",
30
+ "warmup_ratio": 0.03,
31
+ "warmup_steps": 19,
32
+ "weight_decay": 0.0,
33
+ "max_grad_norm": 1.0,
34
+ "optimizer": "AdamW",
35
+ "seed": 0,
36
+ "bf16": true,
37
+ "grad_ckpt": true,
38
+ "frame_rule": "D16",
39
+ "n_input_frames": 32,
40
+ "grid_t": 16,
41
+ "frame_hw": [
42
+ 480,
43
+ 640
44
+ ],
45
+ "base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
46
+ "processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
47
+ "lam": 1.0,
48
+ "epochs": 1,
49
+ "n_samples": 5000,
50
+ "n_steps_full": 625,
51
+ "steps_this_run": 625,
52
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
53
+ "data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
54
+ "data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
55
+ "ckpt_name": "G",
56
+ "git_commit": "b0457f9481f00c0deb596e0fe2caf37e278c01e2",
57
+ "workers": 4,
58
+ "new_tokens": [
59
+ "<box>",
60
+ "<cam>",
61
+ "<rel>"
62
+ ],
63
+ "new_token_ids": [
64
+ 151669,
65
+ 151670,
66
+ 151671
67
+ ],
68
+ "heads": [
69
+ "box",
70
+ "cam(GeoVR CameraDec)",
71
+ "rel"
72
+ ],
73
+ "cam_init": "GeoVR model.camera_token.weight[1]",
74
+ "cam_head": "GeoVR camera_head (CameraDec 2560, fp32, ckpt)",
75
+ "cam_target": "gconv.cam_g: first-frame w2c [s·t, quat xyzw w>=0], s=VGGT scale",
76
+ "cam_loss": "|dt|.clamp(100).mean + |dq|.mean, no FoV, no 2.5x",
77
+ "scale_s1": false,
78
+ "scale_file_sha256": "9e2f0a5abe5f1dbc4f4f6a074fd5584f668c209302f997b778e373bfea42bb95",
79
+ "cam_mask": {
80
+ "n_pair": 4000,
81
+ "n_valid": 4000,
82
+ "frame0_invalid": 0,
83
+ "scale_none": 0,
84
+ "scale_missing_scene": 0,
85
+ "frames_mismatch": 0,
86
+ "s1_used": false,
87
+ "n_masked": 0
88
+ },
89
+ "cam_mode": "untied",
90
+ "deviation_from_order": "지시서와 다른 곳: <cam> 입력 임베딩 분리(B안), 이유 = camera_token[1] 노름 51.05 vs 어휘 중앙값 1.15, tie 로 CE 23.5→31.4 (A안 스모크 VS1/K/logs/smoke_G_smoke_A.json)",
91
+ "curve_every": 25,
92
+ "curve_ids": "/workspace/code/spatial_reasoning/VS1/K/data/g_curve_ids.json",
93
+ "attn": "flash_attention_2"
94
+ }
95
+ }
vs1/ckpt/G/resume/new_rows.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:113bcfce8ddeca995efe8d66aebf7f82c83a08b89a6bc09be056463b955a78ba
3
+ size 17072
vs1/ckpt/G/resume/optim_state.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca94c4fb79af642e78676aac876cde8edd1ac10655b4b8e2bbaa17915c36a8d1
3
+ size 357341427
vs1/ckpt/G/resume/trainable_params.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:977f2a0195cd0eedb52da3bb349d7a07abac24dc4ff4e21e5c52b5d653b7f850
3
+ size 132339780