5 exercises on pronouncing machine learning library and model names.
0 / 5 completed
1 / 5
How is "PyTorch" pronounced?
PyTorch is pronounced "PIE-torch" /ˈpaɪtɔːrtʃ/ — "Py" said like "pie" (the same /paɪ/ as in Python) followed by "torch" (the everyday word). Stress on the first part. So "train it in PIE-torch", "a PIE-torch model." The "ch" is /tʃ/ as in "church," not /ʃ/. Do not spell out "pee-why" or shorten it to "pit-orch." The "Py" prefix signals its Python roots, matching Python's "pie" sound.
2 / 5
How is "NumPy" pronounced?
NumPy is pronounced "NUM-pie" /ˈnʌmpaɪ/ — "num" (from "numerical," short /ʌ/ vowel as in "number") plus "pie" /paɪ/ (the "Py" from Python). Stress on the first syllable. So "import NUM-pie", "a NUM-pie array." Do not say "num-PEE" (the "Py" is "pie," not "pee") or "noom-pee." This follows the Python-ecosystem convention where "Py" is pronounced "pie": NumPy, SciPy ("SIGH-pie"), PyTorch ("PIE-torch").
3 / 5
How is "SciPy" pronounced?
SciPy is pronounced "SIGH-pie" /ˈsaɪpaɪ/ — "Sci" said like "sigh" (from "scientific") plus "Py" said like "pie." Stress on the first syllable. So "use SIGH-pie for the stats", "a SIGH-pie function." Do not say "skippy" or "see-pee." This matches the ecosystem pattern: the "Py" in scientific Python libraries is always "pie," and "Sci" is "sigh," giving "SIGH-pie." Compare NumPy ("NUM-pie").
4 / 5
How is "Keras" (the deep learning API) pronounced?
Keras is pronounced "CARE-ass" /ˈkɛrəs/ — two syllables, stress on the first, like "care" plus a weak "-ass." The name comes from the Greek word for "horn" (κέρας). So "build it in CARE-ass", "a CARE-ass model." The creator Francois Chollet uses this pronunciation. Do not stress the second syllable ("kuh-RASS") or use a long "ee" ("keer-as"). The first vowel is the /ɛ/ of "bed."
5 / 5
How is "GPT" (as in GPT models) read aloud?
GPT is an initialism, always spelled out: "G-P-T" /dʒiː piː tiː/. So "a G-P-T model", "fine-tune G-P-T", "G-P-T four." Each letter individually: "jee, pee, tee." It stands for Generative Pre-trained Transformer. Never blend it into a word like "gippt" or "jept." Similarly "LLM" is spelled out "L-L-M" /ɛl ɛl ɛm/, and "NLP" is "N-L-P." These AI initialisms are read as letters, not words.