Hugging Face's logo

Prisma-Multimodal
/
sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-10-hook_resid_post-l1-1e-05

Feature Extraction
Transformers
English
torch
clip
vision
interpretability
sparse autoencoder
sae
mechanistic interpretability
Model card Files Community
sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-10-hook_resid_post-l1-1e-05
File size: 2,254 Bytes 
e7e2129
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
 
{
    "model_class_name": "HookedViT",
    "model_name": "open-clip:laion/CLIP-ViT-B-32-DataComp.XL-s13B-b90K",
    "hook_point_layer": 10,
    "layer_subtype": "hook_resid_post",
    "hook_point_head_index": null,
    "context_size": 50,
    "use_cached_activations": false,
    "use_patches_only": false,
    "cached_activations_path": "activations/_network_scratch_s_sonia.joseph_datasets_kaggle_datasets/open-clip:laion_CLIP-ViT-B-32-DataComp.XL-s13B-b90K/blocks.9.hook_mlp_out",
    "d_in": 768,
    "activation_fn_str": "relu",
    "activation_fn_kwargs": {},
    "cls_token_only": false,
    "max_grad_norm": 1.0,
    "initialization_method": "encoder_transpose_decoder",
    "normalize_activations": null,
    "n_batches_in_buffer": 20,
    "store_batch_size": 32,
    "num_workers": 16,
    "num_epochs": 10,
    "total_training_images": 13000000,
    "total_training_tokens": 650000000,
    "image_size": 224,
    "device": {
        "__type__": "torch.device",
        "value": "cuda"
    },
    "seed": 42,
    "dtype": {
        "__type__": "torch.dtype",
        "value": "torch.float32"
    },
    "architecture": "standard",
    "verbose": false,
    "b_dec_init_method": "geometric_median",
    "expansion_factor": 64,
    "from_pretrained_path": null,
    "d_sae": 49152,
    "l1_coefficient": 1e-05,
    "lp_norm": 1,
    "lr": 0.0004,
    "lr_scheduler_name": "cosineannealingwarmup",
    "lr_warm_up_steps": 200,
    "train_batch_size": 4096,
    "dataset_name": "imagenet1k",
    "dataset_path": "/network/scratch/s/sonia.joseph/datasets/kaggle_datasets",
    "dataset_train_path": "/network/scratch/s/sonia.joseph/datasets/kaggle_datasets/ILSVRC/Data/CLS-LOC/train",
    "dataset_val_path": "/network/scratch/s/sonia.joseph/datasets/kaggle_datasets/ILSVRC/Data/CLS-LOC/val",
    "use_ghost_grads": true,
    "feature_sampling_window": 1000,
    "dead_feature_window": 5000,
    "dead_feature_threshold": 1e-08,
    "log_to_wandb": true,
    "wandb_project": "clip_b_mlp_out_sae_hyperparam_sweep",
    "wandb_entity": null,
    "wandb_log_frequency": 100,
    "n_validation_runs": 10,
    "n_checkpoints": 10,
    "checkpoint_path": "/network/scratch/s/sonia.joseph/checkpoints/clip-b/77257359-clip_b_mlp_out_sae_hyperparam_sweep"
}