Hugging Face's logo

AbstractPhil
/
rose-geoclip

TensorBoard
Model card Files
xet
Metrics Community
AbstractPhil commited on
Commit
c40165e
·
verified ·
1 Parent(s): d4504b7

[CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d] epoch=1 step=98 is_best=True

Browse files
Files changed (7) hide show
  1. CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/checkpoint_best.meta.json +115 -0
  2. CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/checkpoint_best.safetensors +3 -0
  3. CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/encoder_snapshot.meta.json +115 -0
  4. CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/encoder_snapshot.safetensors +3 -0
  5. CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/head_best.meta.json +115 -0
  6. CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/head_best.safetensors +3 -0
  7. CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/training_latest.json +7 -0
CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/checkpoint_best.meta.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1,
3
+ "global_step": 98,
4
+ "avg_loss": 2.7008541645431516,
5
+ "best_metric": 0.8,
6
+ "timestamp": "2025-08-29T04:14:08.608991",
7
+ "run_name": "CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d",
8
+ "group": "checkpoint",
9
+ "metrics": {
10
+ "epoch": 1,
11
+ "train_loss": 2.7008541645431516,
12
+ "val_loss": 0.9892485931396484,
13
+ "top1": 0.8,
14
+ "top3": 0.9327999994277955,
15
+ "rose_weight": 0.15,
16
+ "rose_margin": 0.25,
17
+ "rose_temperature": 1.0
18
+ },
19
+ "config_snapshot": {
20
+ "device": "cuda",
21
+ "seed": 42,
22
+ "precision": "fp32",
23
+ "hf_clip_id": "laion/CLIP-ViT-B-32-laion2B-s34B-b79K",
24
+ "freeze_encoder": true,
25
+ "normalize_output": true,
26
+ "dataset": "cifar100",
27
+ "data_root": "./data",
28
+ "image_size": 224,
29
+ "clip_mean": [
30
+ 0.48145466,
31
+ 0.4578275,
32
+ 0.40821073
33
+ ],
34
+ "clip_std": [
35
+ 0.26862954,
36
+ 0.26130258,
37
+ 0.27577711
38
+ ],
39
+ "train_aug": true,
40
+ "train_crop_scale": [
41
+ 0.9,
42
+ 1.0
43
+ ],
44
+ "hflip_p": 0.5,
45
+ "vocab_repo": "AbstractPhil/geometric-vocab-768d",
46
+ "vocab_split": "unicode",
47
+ "vocab_dim_override": 768,
48
+ "crystal_dims": 768,
49
+ "role_index": {
50
+ "anchor": 0,
51
+ "support": 1,
52
+ "contrast": 2,
53
+ "purpose": 3,
54
+ "observer": 4
55
+ },
56
+ "rose_loss_weight": 0.15,
57
+ "rose_margin": 0.25,
58
+ "rose_temperature": 1.0,
59
+ "rose_role_weights": {
60
+ "anchor": 1.0,
61
+ "need": -0.5,
62
+ "relation": 0.75,
63
+ "purpose": 0.75,
64
+ "observer": -0.5
65
+ },
66
+ "classes": 100,
67
+ "classes_per_penta": 2,
68
+ "penta_branches": 50,
69
+ "pentachora_blocks": 4,
70
+ "temperature": 0.07,
71
+ "batch_size": 512,
72
+ "epochs": 20,
73
+ "lr": 0.0001,
74
+ "weight_decay": 0.0001,
75
+ "label_smoothing": 0.0,
76
+ "use_cosine_sched": true,
77
+ "warmup_steps": 0,
78
+ "max_steps": null,
79
+ "amp": true,
80
+ "grad_clip_norm": null,
81
+ "log_every": 25,
82
+ "auto_train": false,
83
+ "num_workers": 2,
84
+ "pin_memory": true,
85
+ "persistent_workers": true,
86
+ "ckpt_dir": "./checkpoints",
87
+ "run_name": "CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d",
88
+ "save_latest": true,
89
+ "save_best": true,
90
+ "resume": false,
91
+ "resume_which": "best",
92
+ "resume_source": "auto",
93
+ "hf_enable": true,
94
+ "hf_repo_id": "AbstractPhil/rose-geoclip",
95
+ "hf_branch": "main",
96
+ "hf_private": true,
97
+ "hf_push_latest": false,
98
+ "hf_push_best": true,
99
+ "hf_create": true,
100
+ "hf_token_env": [
101
+ "HF_TOKEN",
102
+ "HUGGINGFACE_HUB_TOKEN"
103
+ ],
104
+ "features_enable": true,
105
+ "features_local_root": "./features_cache",
106
+ "features_batch_size": 1024,
107
+ "features_push_to_hub": true,
108
+ "features_offload_after_prepare": false,
109
+ "features_repo_id": "AbstractPhil/CLIP-ViT-B-32-laion2B-s34B-b79K_cifar100_extracted_features",
110
+ "features_local_dir": "./features_cache/CLIP-ViT-B-32-laion2B-s34B-b79K-cifar100",
111
+ "features_embed_dim": 512,
112
+ "features_ready": true,
113
+ "use_precomputed_features": true
114
+ }
115
+ }
CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/checkpoint_best.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:194d60bfbbe61f36695fab20487cf829c34888565f1c3e6879bc2cbd47008a74
3
+ size 3938232
CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/encoder_snapshot.meta.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1,
3
+ "global_step": 98,
4
+ "avg_loss": 2.7008541645431516,
5
+ "best_metric": 0.8,
6
+ "timestamp": "2025-08-29T04:14:08.362065",
7
+ "run_name": "CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d",
8
+ "group": "encoder",
9
+ "metrics": {
10
+ "epoch": 1,
11
+ "train_loss": 2.7008541645431516,
12
+ "val_loss": 0.9892485931396484,
13
+ "top1": 0.8,
14
+ "top3": 0.9327999994277955,
15
+ "rose_weight": 0.15,
16
+ "rose_margin": 0.25,
17
+ "rose_temperature": 1.0
18
+ },
19
+ "config_snapshot": {
20
+ "device": "cuda",
21
+ "seed": 42,
22
+ "precision": "fp32",
23
+ "hf_clip_id": "laion/CLIP-ViT-B-32-laion2B-s34B-b79K",
24
+ "freeze_encoder": true,
25
+ "normalize_output": true,
26
+ "dataset": "cifar100",
27
+ "data_root": "./data",
28
+ "image_size": 224,
29
+ "clip_mean": [
30
+ 0.48145466,
31
+ 0.4578275,
32
+ 0.40821073
33
+ ],
34
+ "clip_std": [
35
+ 0.26862954,
36
+ 0.26130258,
37
+ 0.27577711
38
+ ],
39
+ "train_aug": true,
40
+ "train_crop_scale": [
41
+ 0.9,
42
+ 1.0
43
+ ],
44
+ "hflip_p": 0.5,
45
+ "vocab_repo": "AbstractPhil/geometric-vocab-768d",
46
+ "vocab_split": "unicode",
47
+ "vocab_dim_override": 768,
48
+ "crystal_dims": 768,
49
+ "role_index": {
50
+ "anchor": 0,
51
+ "support": 1,
52
+ "contrast": 2,
53
+ "purpose": 3,
54
+ "observer": 4
55
+ },
56
+ "rose_loss_weight": 0.15,
57
+ "rose_margin": 0.25,
58
+ "rose_temperature": 1.0,
59
+ "rose_role_weights": {
60
+ "anchor": 1.0,
61
+ "need": -0.5,
62
+ "relation": 0.75,
63
+ "purpose": 0.75,
64
+ "observer": -0.5
65
+ },
66
+ "classes": 100,
67
+ "classes_per_penta": 2,
68
+ "penta_branches": 50,
69
+ "pentachora_blocks": 4,
70
+ "temperature": 0.07,
71
+ "batch_size": 512,
72
+ "epochs": 20,
73
+ "lr": 0.0001,
74
+ "weight_decay": 0.0001,
75
+ "label_smoothing": 0.0,
76
+ "use_cosine_sched": true,
77
+ "warmup_steps": 0,
78
+ "max_steps": null,
79
+ "amp": true,
80
+ "grad_clip_norm": null,
81
+ "log_every": 25,
82
+ "auto_train": false,
83
+ "num_workers": 2,
84
+ "pin_memory": true,
85
+ "persistent_workers": true,
86
+ "ckpt_dir": "./checkpoints",
87
+ "run_name": "CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d",
88
+ "save_latest": true,
89
+ "save_best": true,
90
+ "resume": false,
91
+ "resume_which": "best",
92
+ "resume_source": "auto",
93
+ "hf_enable": true,
94
+ "hf_repo_id": "AbstractPhil/rose-geoclip",
95
+ "hf_branch": "main",
96
+ "hf_private": true,
97
+ "hf_push_latest": false,
98
+ "hf_push_best": true,
99
+ "hf_create": true,
100
+ "hf_token_env": [
101
+ "HF_TOKEN",
102
+ "HUGGINGFACE_HUB_TOKEN"
103
+ ],
104
+ "features_enable": true,
105
+ "features_local_root": "./features_cache",
106
+ "features_batch_size": 1024,
107
+ "features_push_to_hub": true,
108
+ "features_offload_after_prepare": false,
109
+ "features_repo_id": "AbstractPhil/CLIP-ViT-B-32-laion2B-s34B-b79K_cifar100_extracted_features",
110
+ "features_local_dir": "./features_cache/CLIP-ViT-B-32-laion2B-s34B-b79K-cifar100",
111
+ "features_embed_dim": 512,
112
+ "features_ready": true,
113
+ "use_precomputed_features": true
114
+ }
115
+ }
CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/encoder_snapshot.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:591af02d797904d5c60ab0fb9e47d4c8a59a1a673113020522a53dc68f7b3a7b
3
+ size 2736
CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/head_best.meta.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1,
3
+ "global_step": 98,
4
+ "avg_loss": 2.7008541645431516,
5
+ "best_metric": 0.8,
6
+ "timestamp": "2025-08-29T04:14:08.335155",
7
+ "run_name": "CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d",
8
+ "group": "head",
9
+ "metrics": {
10
+ "epoch": 1,
11
+ "train_loss": 2.7008541645431516,
12
+ "val_loss": 0.9892485931396484,
13
+ "top1": 0.8,
14
+ "top3": 0.9327999994277955,
15
+ "rose_weight": 0.15,
16
+ "rose_margin": 0.25,
17
+ "rose_temperature": 1.0
18
+ },
19
+ "config_snapshot": {
20
+ "device": "cuda",
21
+ "seed": 42,
22
+ "precision": "fp32",
23
+ "hf_clip_id": "laion/CLIP-ViT-B-32-laion2B-s34B-b79K",
24
+ "freeze_encoder": true,
25
+ "normalize_output": true,
26
+ "dataset": "cifar100",
27
+ "data_root": "./data",
28
+ "image_size": 224,
29
+ "clip_mean": [
30
+ 0.48145466,
31
+ 0.4578275,
32
+ 0.40821073
33
+ ],
34
+ "clip_std": [
35
+ 0.26862954,
36
+ 0.26130258,
37
+ 0.27577711
38
+ ],
39
+ "train_aug": true,
40
+ "train_crop_scale": [
41
+ 0.9,
42
+ 1.0
43
+ ],
44
+ "hflip_p": 0.5,
45
+ "vocab_repo": "AbstractPhil/geometric-vocab-768d",
46
+ "vocab_split": "unicode",
47
+ "vocab_dim_override": 768,
48
+ "crystal_dims": 768,
49
+ "role_index": {
50
+ "anchor": 0,
51
+ "support": 1,
52
+ "contrast": 2,
53
+ "purpose": 3,
54
+ "observer": 4
55
+ },
56
+ "rose_loss_weight": 0.15,
57
+ "rose_margin": 0.25,
58
+ "rose_temperature": 1.0,
59
+ "rose_role_weights": {
60
+ "anchor": 1.0,
61
+ "need": -0.5,
62
+ "relation": 0.75,
63
+ "purpose": 0.75,
64
+ "observer": -0.5
65
+ },
66
+ "classes": 100,
67
+ "classes_per_penta": 2,
68
+ "penta_branches": 50,
69
+ "pentachora_blocks": 4,
70
+ "temperature": 0.07,
71
+ "batch_size": 512,
72
+ "epochs": 20,
73
+ "lr": 0.0001,
74
+ "weight_decay": 0.0001,
75
+ "label_smoothing": 0.0,
76
+ "use_cosine_sched": true,
77
+ "warmup_steps": 0,
78
+ "max_steps": null,
79
+ "amp": true,
80
+ "grad_clip_norm": null,
81
+ "log_every": 25,
82
+ "auto_train": false,
83
+ "num_workers": 2,
84
+ "pin_memory": true,
85
+ "persistent_workers": true,
86
+ "ckpt_dir": "./checkpoints",
87
+ "run_name": "CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d",
88
+ "save_latest": true,
89
+ "save_best": true,
90
+ "resume": false,
91
+ "resume_which": "best",
92
+ "resume_source": "auto",
93
+ "hf_enable": true,
94
+ "hf_repo_id": "AbstractPhil/rose-geoclip",
95
+ "hf_branch": "main",
96
+ "hf_private": true,
97
+ "hf_push_latest": false,
98
+ "hf_push_best": true,
99
+ "hf_create": true,
100
+ "hf_token_env": [
101
+ "HF_TOKEN",
102
+ "HUGGINGFACE_HUB_TOKEN"
103
+ ],
104
+ "features_enable": true,
105
+ "features_local_root": "./features_cache",
106
+ "features_batch_size": 1024,
107
+ "features_push_to_hub": true,
108
+ "features_offload_after_prepare": false,
109
+ "features_repo_id": "AbstractPhil/CLIP-ViT-B-32-laion2B-s34B-b79K_cifar100_extracted_features",
110
+ "features_local_dir": "./features_cache/CLIP-ViT-B-32-laion2B-s34B-b79K-cifar100",
111
+ "features_embed_dim": 512,
112
+ "features_ready": true,
113
+ "use_precomputed_features": true
114
+ }
115
+ }
CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/head_best.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38ec79d6ecedb4202342ccd36f69cd3ac81e08f248c31c3b79cf05ed9b961305
3
+ size 3938224
CLIP-ViT-B-32-laion2B-s34B-b79K-uni-768dx768d/training_latest.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1,
3
+ "step": 98,
4
+ "avg_loss": 2.7008541645431516,
5
+ "best_metric": 0.8,
6
+ "ts": "2025-08-29 04:14:08"
7
+ }