Got it! Converting your command-line FastText supervised training setup into a Python script is straightforward—you just pass each of those command-line flags as keyword arguments to the fasttext.supervised() function.

Here's the exact Python code that replicates your original command:

import fasttext

# Train the classifier with your specified parameters
classifier = fasttext.supervised(
    input='FT_Race_data.txt',
    output='race_model',
    lr=0.4,
    epoch=30,
    loss='hs'
)

Let me break down how each command-line parameter maps to the Python function:

• -input FT_Race_data.txt → input='FT_Race_data.txt' (path to your training dataset)
• -output race_model → output='race_model' (prefix for saved model files; you’ll get race_model.bin and race_model.vec after training completes)
• -lr 0.4 → lr=0.4 (sets the learning rate)
• -epoch 30 → epoch=30 (number of full passes over the training data)
• -loss hs → loss='hs' (uses hierarchical softmax as the loss function)

If you want to tweak other parameters later (like adding -dim 100 for embedding size, or -wordNgrams 2 for bigram features), you just add those as additional keyword arguments to the function call—for example, dim=100 or wordNgrams=2.

内容的提问来源于stack exchange，提问作者Raady