{ "architectures": [ "FlavaImageCodebook" ], "dtype": "float32", "freeze": true, "hidden_size": 32, "initializer_range": 0.02, "input_channels": 3, "num_blocks_per_group": 2, "num_groups": 2, "transformers_version": "5.15.0.dev0", "vocab_size": 1024 }