config.json

{
  "architectures": [
    "FlamingoModel"
  ],
  "clip_model_type": "openai/clip-vit-large-patch14",
  "dim": 1024,
  "dim_visual": 1024,
  "lm": "facebook/opt-350m",
  "resampler_act": "sqrelu",
  "resampler_depth": 6,
  "resampler_dim_head": 128,
  "resampler_ff_mult": 4,
  "resampler_heads": 16,
  "resampler_num_latents": 64,
  "resampler_num_time_embeds": 4,
  "torch_dtype": "float32",
  "transformers_version": "4.21.0",
  "xattn_act": "sqrelu",
  "xattn_dim_head": 128,
  "xattn_every": 1,
  "xattn_ff_mult": 4,
  "xattn_heads": 16
}