DownloadAndLoadFlorence2Model downloads all kinds of files

[felixniemeyer]

When running the following workflow using ComfyUI-Florence2 the first time, many non-model files are being downloaded into comfyUI/models/LLM/.... This does not seem right. Below is the workflow and also the comfyUI log.

{
  "10": {
    "inputs": {
      "model": "gokaygokay/Florence-2-Flux-Large",
      "precision": "fp16",
      "attention": "sdpa"
    },
    "class_type": "DownloadAndLoadFlorence2Model",
    "_meta": {
      "title": "DownloadAndLoadFlorence2Model"
    }
  },
  "11": {
    "inputs": {
      "text_input": "",
      "task": "caption",
      "fill_mask": true,
      "keep_model_loaded": false,
      "max_new_tokens": 1024,
      "num_beams": 3,
      "do_sample": true,
      "output_mask_select": "",
      "seed": 667037814261723,
      "image": [
        "12",
        0
      ],
      "florence2_model": [
        "10",
        0
      ]
    },
    "class_type": "Florence2Run",
    "_meta": {
      "title": "Florence2Run"
    }
  },
  "12": {
    "inputs": {
      "image": "ss2.png",
      "upload": "image"
    },
    "class_type": "LoadImage",
    "_meta": {
      "title": "Load Image"
    }
  },
  "14": {
    "inputs": {
      "mode": "raw value",
      "input": [
        "11",
        2
      ]
    },
    "class_type": "DisplayAny",
    "_meta": {
      "title": "🔧 Display Any"
    }
  }
}

Downloading Florence2 model to: /home/felix/davdev/ai/ComfyUI/models/LLM/Florence-2-Flux-Large
/home/felix/davdev/ai/ComfyUI/venv/lib/python3.12/site-packages/huggingface_hub/file_download.py:1204: UserWarning: `local_dir_use_symlinks` parameter is deprecated and will be ignored. The process to download files to a local folder has been updated and do not rely on symlinks anymore. You only need to pass a destination folder as`local_dir`.
For more details, check out https://huggingface.co/docs/huggingface_hub/main/en/guides/download#download-files-to-local-folder.
  warnings.warn(
Fetching 15 files:  20%|██        | 3/15 [00:00<00:03,  3.21it/s]

Here are all the files already present after 20%:

ls models/LLM/Florence-2-Flux-Large/
added_tokens.json           merges.txt                README.md                vocab.json
config.json                 modeling_florence2.py     special_tokens_map.json
configuration_florence2.py  preprocessor_config.json  tokenizer_config.json
generation_config.json      processing_florence2.py   tokenizer.json

[kijai]

Yes... that's how diffusers models are, they are folders with lots of files, including confings, model weights and in the case of Florence2 and some other LLMs, part of the inference code.

[felixniemeyer]

Ok, thanks for the explanantion. I was used to downloading only single model files (checkpoints, loras, controlnets, upscalers ...).