def validate_hyperparameters()

in build_and_train_models/sm-introduction_to_blazingtext_word2vec_text8/validator.py [0:0]

• 57 lines of code
• 24 McCabe index (conditional complexity)

def validate_hyperparameters(cfg):
    warnings = 0

    if "mode" in cfg:
        tmp = cfg["mode"]
        if tmp not in ["skipgram", "cbow", "batch_skipgram"]:
            raise Exception('mode should be one of ["skipgram", "cbow", "batch_skipgram"]')

    if "min_count" in cfg:
        tmp = get_int(cfg["min_count"], "min_count")
        if tmp < 0:
            raise Exception("Parameter 'min_count' should be >= 0.")

    if "sampling_threshold" in cfg:
        tmp = get_float(cfg["sampling_threshold"], "sampling_threshold")
        if tmp <= 0 or tmp >= 1:
            raise Exception("Parameter 'sampling_threshold' should be between (0,1)")

    if "learning_rate" in cfg:  # Default: .05
        tmp = get_float(cfg["learning_rate"], "learning_rate")
        if tmp <= 0:
            raise Exception("Parameter 'learning_rate' should be > 0.")

    ws = 5
    if "window_size" in cfg:  # Default: 5
        ws = get_int(cfg["window_size"], "window_size")
        if ws <= 0:
            raise Exception("Parameter 'window_size' should be > 0.")

    if "vector_dim" in cfg:  # Default: 100
        tmp = get_int(cfg["vector_dim"], "vector_dim")
        if tmp <= 0:
            raise Exception("Parameter 'vector_dim' should be > 0.")
        if tmp > 2048:
            raise Exception("Parameter 'vector_dim' should be <= 2048.")
        if tmp >= 1500:
            warnings += 1
            print(
                "You are using a big vector dimension. Training might take a long time or might fail due to memory "
                "issues."
            )

    if "epochs" in cfg:  # Default: 5
        tmp = get_int(cfg["epochs"], "epochs")
        if tmp <= 0:
            raise Exception("Parameter 'epochs' should be > 0.")

    if "negative_samples" in cfg:  # Default: 5
        tmp = get_int(cfg["negative_samples"], "negative_samples")
        if tmp <= 0:
            raise Exception("Parameter 'negative_samples' should be > 0.")

    if "batch_size" in cfg:  # Default: 11
        tmp = get_int(cfg["batch_size"], "batch_size")
        if tmp <= 0:
            raise Exception("Parameter 'batch_size' should be > 0.")
        if tmp > 32:
            raise Exception("Parameter 'batch_size' should be <= 32.")
        reco = 2 * ws + 1
        if tmp is not reco:
            warnings += 1
            print(
                "It is recommended that you set batch_size as 2*window_size + 1 which is %s in this case."
                % str(reco)
            )
    return warnings