blip2-action-model / config.json
reneepr's picture
Upload BLIP-2 vision encoder + classifier head + processor
ae4b8ab verified
raw
history blame contribute delete
672 Bytes
{
"architectures": [
"Blip2VisionModel"
],
"attention_dropout": 0.0,
"dropout": 0.0,
"hidden_act": "gelu",
"hidden_size": 1408,
"id2label": {
"0": "idle_time",
"1": "unplugging_connector"
},
"image_size": 224,
"initializer_factor": 1.0,
"initializer_range": 1e-10,
"intermediate_size": 6144,
"label2id": {
"idle_time": 0,
"unplugging_connector": 1
},
"layer_norm_eps": 1e-06,
"model_type": "blip_2_vision_model",
"num_attention_heads": 16,
"num_channels": 3,
"num_hidden_layers": 39,
"patch_size": 14,
"projection_dim": 512,
"qkv_bias": true,
"torch_dtype": "float32",
"transformers_version": "4.51.3"
}