Instructions to use nvidia/DAM-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Describe Anything
How to use nvidia/DAM-3B with Describe Anything:
# pip install git+https://github.com/NVlabs/describe-anything from huggingface_hub import snapshot_download from dam import DescribeAnythingModel snapshot_download(nvidia/DAM-3B, local_dir="checkpoints") dam = DescribeAnythingModel( model_path="checkpoints", conv_mode="v1", prompt_mode="focal_prompt", )
- Notebooks
- Google Colab
- Kaggle
File size: 989 Bytes
aaa2533 | 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 | {
"_name_or_path": "./checkpoints/describe_anything_v0.7.2_1ep_e2a_with_words_paug_samv2_lvisv4_mapillary_cocostuff_openimages_context_crop_constrained_blur0.5/context_provider",
"architectures": [
"ContextProvider"
],
"attention_dropout": 0.0,
"context_image_as_queries": false,
"context_provider_layer_indices": null,
"context_provider_type": "cross_attn_end_to_all",
"crop_embedding_mode": "add",
"crop_position_single_embedding": false,
"hidden_act": "gelu_pytorch_tanh",
"hidden_size": 1152,
"image_size": 384,
"intermediate_size": 4304,
"layer_norm_eps": 1e-06,
"masked_cross_attn": false,
"model_type": "context_provider",
"num_attention_heads": 16,
"num_channels": 3,
"num_hidden_layers": 27,
"num_mask_channels": 0,
"patch_size": 14,
"residual_dropout": 0.0,
"torch_dtype": "bfloat16",
"trainable_crop_position_embedding": true,
"transformers_version": "4.36.2",
"treat_image_as_cimage": false,
"zero_init_output": true
}
|