Hugging Face's logo

JH-C-k
/
spatial_reasoning

Safetensors
spatial-reasoning
lora
Model card Files
xet
 Community
JH-C-k commited on
Commit
b9bf6f4
·
verified ·
1 Parent(s): 252bb2d

add vs1/ckpt/S3

Browse files
Files changed (13) hide show
  1. vs1/ckpt/S3/100/heads.pt +3 -0
  2. vs1/ckpt/S3/100/meta.json +74 -0
  3. vs1/ckpt/S3/100/new_rows.pt +3 -0
  4. vs1/ckpt/S3/100/trainable_params.pt +3 -0
  5. vs1/ckpt/S3/50/heads.pt +3 -0
  6. vs1/ckpt/S3/50/meta.json +74 -0
  7. vs1/ckpt/S3/50/new_rows.pt +3 -0
  8. vs1/ckpt/S3/50/trainable_params.pt +3 -0
  9. vs1/ckpt/S3/resume/heads.pt +3 -0
  10. vs1/ckpt/S3/resume/meta.json +74 -0
  11. vs1/ckpt/S3/resume/new_rows.pt +3 -0
  12. vs1/ckpt/S3/resume/optim_state.pt +3 -0
  13. vs1/ckpt/S3/resume/trainable_params.pt +3 -0
vs1/ckpt/S3/100/heads.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ddf6b8b56bb5f73d964c0b5d3d08cc2e7ab61e9c852f831e73f5afd29f765f88
3
+ size 31539355
vs1/ckpt/S3/100/meta.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "variant": "S",
3
+ "step": 625,
4
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
5
+ "lam": 1.0,
6
+ "run_name": "S3",
7
+ "time": "2026-10-20 11:15:02 KST",
8
+ "cfg": {
9
+ "variant": "S",
10
+ "lora_r": 16,
11
+ "lora_alpha": 32,
12
+ "lora_dropout": 0.05,
13
+ "lora_targets": [
14
+ "q_proj",
15
+ "k_proj",
16
+ "v_proj",
17
+ "o_proj",
18
+ "gate_proj",
19
+ "up_proj",
20
+ "down_proj"
21
+ ],
22
+ "lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
23
+ "micro_batch": 1,
24
+ "grad_accum": 8,
25
+ "lr": 0.0001,
26
+ "head_lr": 0.0001,
27
+ "new_token_lr": 0.0001,
28
+ "scheduler": "cosine",
29
+ "warmup_ratio": 0.03,
30
+ "warmup_steps": 19,
31
+ "weight_decay": 0.0,
32
+ "max_grad_norm": 1.0,
33
+ "optimizer": "AdamW",
34
+ "seed": 0,
35
+ "bf16": true,
36
+ "grad_ckpt": true,
37
+ "frame_rule": "D16",
38
+ "n_input_frames": 32,
39
+ "grid_t": 16,
40
+ "frame_hw": [
41
+ 480,
42
+ 640
43
+ ],
44
+ "base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
45
+ "processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
46
+ "lam": 1.0,
47
+ "epochs": 1,
48
+ "n_samples": 5000,
49
+ "n_steps_full": 625,
50
+ "steps_this_run": 625,
51
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
52
+ "data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
53
+ "data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
54
+ "ckpt_name": "S3",
55
+ "git_commit": "35c7b39947ee7440ee4ca57833b525aa44a557b6",
56
+ "workers": 4,
57
+ "new_tokens": [
58
+ "<box>",
59
+ "<cam>",
60
+ "<rel>"
61
+ ],
62
+ "new_token_ids": [
63
+ 151669,
64
+ 151670,
65
+ 151671
66
+ ],
67
+ "heads": [
68
+ "box",
69
+ "cam",
70
+ "rel"
71
+ ],
72
+ "attn": "flash_attention_2"
73
+ }
74
+ }
vs1/ckpt/S3/100/new_rows.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9335fc3b08796a75946f42efc5e438e4b5bfece503347bfc7b8e0ec6b3cd1fe
3
+ size 17072
vs1/ckpt/S3/100/trainable_params.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7730bdf9a7dbee5036e5868cb0870127d62355da7a49aefce88b94e7ce2d255
3
+ size 132339780
vs1/ckpt/S3/50/heads.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a97efc6f55cfd97a7e05f2f40a61d2a50ac79ca0e7146b59c4e4c63cd9f0e51f
3
+ size 31539355
vs1/ckpt/S3/50/meta.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "variant": "S",
3
+ "step": 312,
4
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
5
+ "lam": 1.0,
6
+ "run_name": "S3",
7
+ "time": "2026-10-20 10:13:41 KST",
8
+ "cfg": {
9
+ "variant": "S",
10
+ "lora_r": 16,
11
+ "lora_alpha": 32,
12
+ "lora_dropout": 0.05,
13
+ "lora_targets": [
14
+ "q_proj",
15
+ "k_proj",
16
+ "v_proj",
17
+ "o_proj",
18
+ "gate_proj",
19
+ "up_proj",
20
+ "down_proj"
21
+ ],
22
+ "lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
23
+ "micro_batch": 1,
24
+ "grad_accum": 8,
25
+ "lr": 0.0001,
26
+ "head_lr": 0.0001,
27
+ "new_token_lr": 0.0001,
28
+ "scheduler": "cosine",
29
+ "warmup_ratio": 0.03,
30
+ "warmup_steps": 19,
31
+ "weight_decay": 0.0,
32
+ "max_grad_norm": 1.0,
33
+ "optimizer": "AdamW",
34
+ "seed": 0,
35
+ "bf16": true,
36
+ "grad_ckpt": true,
37
+ "frame_rule": "D16",
38
+ "n_input_frames": 32,
39
+ "grid_t": 16,
40
+ "frame_hw": [
41
+ 480,
42
+ 640
43
+ ],
44
+ "base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
45
+ "processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
46
+ "lam": 1.0,
47
+ "epochs": 1,
48
+ "n_samples": 5000,
49
+ "n_steps_full": 625,
50
+ "steps_this_run": 625,
51
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
52
+ "data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
53
+ "data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
54
+ "ckpt_name": "S3",
55
+ "git_commit": "35c7b39947ee7440ee4ca57833b525aa44a557b6",
56
+ "workers": 4,
57
+ "new_tokens": [
58
+ "<box>",
59
+ "<cam>",
60
+ "<rel>"
61
+ ],
62
+ "new_token_ids": [
63
+ 151669,
64
+ 151670,
65
+ 151671
66
+ ],
67
+ "heads": [
68
+ "box",
69
+ "cam",
70
+ "rel"
71
+ ],
72
+ "attn": "flash_attention_2"
73
+ }
74
+ }
vs1/ckpt/S3/50/new_rows.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb4dfd8d5dee3637d566f904ed9b0af22194289248925521dc1fccde9bcacc3b
3
+ size 17072
vs1/ckpt/S3/50/trainable_params.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:791ae66f6c005e59870ebeb292287090fa0850a4e228b384301c47ad8bdea489
3
+ size 132339780
vs1/ckpt/S3/resume/heads.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ddf6b8b56bb5f73d964c0b5d3d08cc2e7ab61e9c852f831e73f5afd29f765f88
3
+ size 31539355
vs1/ckpt/S3/resume/meta.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "variant": "S",
3
+ "step": 625,
4
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
5
+ "lam": 1.0,
6
+ "run_name": "S3",
7
+ "time": "2026-10-20 11:15:02 KST",
8
+ "cfg": {
9
+ "variant": "S",
10
+ "lora_r": 16,
11
+ "lora_alpha": 32,
12
+ "lora_dropout": 0.05,
13
+ "lora_targets": [
14
+ "q_proj",
15
+ "k_proj",
16
+ "v_proj",
17
+ "o_proj",
18
+ "gate_proj",
19
+ "up_proj",
20
+ "down_proj"
21
+ ],
22
+ "lora_regex": "model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.(q_proj|k_proj|v_proj|o_proj)|mlp\\.(gate_proj|up_proj|down_proj))",
23
+ "micro_batch": 1,
24
+ "grad_accum": 8,
25
+ "lr": 0.0001,
26
+ "head_lr": 0.0001,
27
+ "new_token_lr": 0.0001,
28
+ "scheduler": "cosine",
29
+ "warmup_ratio": 0.03,
30
+ "warmup_steps": 19,
31
+ "weight_decay": 0.0,
32
+ "max_grad_norm": 1.0,
33
+ "optimizer": "AdamW",
34
+ "seed": 0,
35
+ "bf16": true,
36
+ "grad_ckpt": true,
37
+ "frame_rule": "D16",
38
+ "n_input_frames": 32,
39
+ "grid_t": 16,
40
+ "frame_hw": [
41
+ 480,
42
+ 640
43
+ ],
44
+ "base_model_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B-qwen3vl",
45
+ "processor_dir": "/workspace/code/spatial_reasoning/repro_baseline22/weights/GeoVR-VGGT-Qwen3-VL-4B",
46
+ "lam": 1.0,
47
+ "epochs": 1,
48
+ "n_samples": 5000,
49
+ "n_steps_full": 625,
50
+ "steps_this_run": 625,
51
+ "order_sha1": "d6a78f37e8dad34cfe01e38cc4458a84c1bb8985",
52
+ "data": "/workspace/code/spatial_reasoning/VS1/data/train_v2.jsonl",
53
+ "data_sha256": "5f03730d4cfd2310c740ccf15d25779d9fabdf68b90f182b8f770e656b916438",
54
+ "ckpt_name": "S3",
55
+ "git_commit": "35c7b39947ee7440ee4ca57833b525aa44a557b6",
56
+ "workers": 4,
57
+ "new_tokens": [
58
+ "<box>",
59
+ "<cam>",
60
+ "<rel>"
61
+ ],
62
+ "new_token_ids": [
63
+ 151669,
64
+ 151670,
65
+ 151671
66
+ ],
67
+ "heads": [
68
+ "box",
69
+ "cam",
70
+ "rel"
71
+ ],
72
+ "attn": "flash_attention_2"
73
+ }
74
+ }
vs1/ckpt/S3/resume/new_rows.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9335fc3b08796a75946f42efc5e438e4b5bfece503347bfc7b8e0ec6b3cd1fe
3
+ size 17072
vs1/ckpt/S3/resume/optim_state.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01e92a942ae9670d227fff8b4af5754f4e0579c03ded5d6c7165a7f93cf7e7e1
3
+ size 327818018
vs1/ckpt/S3/resume/trainable_params.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7730bdf9a7dbee5036e5868cb0870127d62355da7a49aefce88b94e7ce2d255
3
+ size 132339780