Scikit learn Algorithmia demo using the Model Manifest system to tie model data and code together immutably

algorithm.py

from Algorithmia import ADK
import joblib


## This function uses the model manifest `state` or `modelData` class to get model files defined in the model manifest automatically.
## No client work required, just make sure the name in `get_model` matches the name in your model manifest.
def load(state):
    state['model'] = joblib.load(state.get_model("model"))
    state['vectorizer'] = joblib.load(state.get_model("vectorizer"))
    return state



def apply(input, state):
    tokens = state['vectorizer'].transform([input])
    preds = state['model'].predict(tokens)
    if int(preds[0]) == 0:
        formatted_output = {"Sentiment": "Negative", "review": input}
    else:
        formatted_output = {"Sentiment": "Positive", "review": input}
    return formatted_output


algorithm = ADK(apply, load)
algorithm.init("Algorithmia is great.")

model_manifest.json

{
  "required_files": [
      {
        "name":"model",
        "source_uri": "data://demo/sentiment_classifier_imbd/passive_aggressive_model_hash.joblib",
        "metadata": {
          "dataset_md5_checksum": "46a44d32d2c5c07f7f66324bef4c7266"
        }
      },
      {
        "name":"vectorizer",
        "source_uri": "data://demo/sentiment_classifier_imbd/hash_vectorizer.joblib",
        "metadata": {
          "dataset_md5_checksum": "46a44d32d2c5c07f7f66324bef4c7266"
        }
      }
    ],
  "optional_files": []
}

model_manifest.json.freeze

{"required_files": [{"name": "model", "source_uri": "data://demo/sentiment_classifier_imbd/passive_aggressive_model_hash.joblib", "metadata": {"dataset_md5_checksum": "46a44d32d2c5c07f7f66324bef4c7266"}, "md5_checksum": "3ca7d81c325f8a1964bac73edce582cc"}, {"name": "vectorizer", "source_uri": "data://demo/sentiment_classifier_imbd/hash_vectorizer.joblib", "metadata": {"dataset_md5_checksum": "46a44d32d2c5c07f7f66324bef4c7266"}, "md5_checksum": "9c498f1ef9610a714846e843ef26d554"}], "optional_files": [], "timestamp": "1639411596.847047", "lock_checksum": "d6976c758eb949a498d4f0a40a8f250f"}

requirements.txt

algorithmia>=1.0.0,<2.0
scikit-learn==0.23.1
pillow==7.2.0
numpy==1.16.0