Create the Model

Run generate-huggingface-resource-scripts.py with the following arguments filled in:

  • hf_name: The model name. Can use the “copy model name to clipboard” option.

  • hf_revision: The most recent full commit hash (or an older commit if preferred).

  • model_size: Click on “Files and Versions” and the model size should be listed next to the branch dropdown.

Example generation command

python generate-huggingface-resource-scripts.py \
    --hf_name=allenai/olmo-2-0325-32b-instruct \
    --hf_revision=b96024342a77a69aa0dda815c3454a671f477463 \
    --model_size=70Gi \
    model

Run the generated script

python models/allenai--olmo-2-0325-32b-instruct.py

Note how the script is generated in the models/ subdirectory.

This step will take some time while the model is pulled from HuggingFace into Dyff.

You can view the model’s progress at:

Once the model is fully ready you’ll see a green check mark on the left-hand side.

Example output

{
  "name": "allenai/olmo-2-0325-32b-instruct",
  "artifact": {
    "kind": "HuggingFaceCache",
    "huggingFaceCache": {
      "repoID": "allenai/olmo-2-0325-32b-instruct",
      "revision": "0000000000000000000000000000000000000000"
    }
  },
  "storage": {
    "medium": "FUSEVolume"
  },
  "source": {
    "kind": "HuggingFaceHub",
    "huggingFaceHub": {
      "repoID": "allenai/olmo-2-0325-32b-instruct",
      "revision": "0000000000000000000000000000000000000000",
      "allowPatterns": [
        "*.safetensors",
        ".gitattributes",
        "*.json",
        "*.md",
        "*.model",
        "*.py",
        "*.txt"
      ],
      "ignorePatterns": [
        "*/*"
      ]
    }
  },
  "resources": {
    "storage": "70Gi"
  },
  "id": "00000000000000000000000000000000",
  "account": "public",
  "schemaVersion": "0.1",
  "labels": {},
  "status": "Created",
  "kind": "Model",
  "metadata": {
    "documentation": {}
  },
  "annotations": [],
  "creationTime": "2025-07-08T04:12:34.546788+00:00"
}