{"id":495832,"date":"2026-09-24T02:30:18","date_gmt":"2026-09-24T02:30:18","guid":{"rendered":"https:\/\/savepearlharbor.com\/?p=495832"},"modified":"-0001-11-30T00:00:00","modified_gmt":"-0001-11-29T21:00:00","slug":"","status":"publish","type":"post","link":"https:\/\/savepearlharbor.com\/?p=495832","title":{"rendered":"Jev \u0437\u0430 25 \u0441\u0442\u0440\u043e\u043a \u043d\u0430 Python"},"content":{"rendered":"
\n
\"My<\/p>\n
My name is Jev<\/figcaption><\/div>\n<\/figure>\n

\u0412\u0441\u0435 \u043a\u043e\u043c\u0443 \u043d\u0435 \u043b\u0435\u043d\u044c \u0433\u043e\u0432\u043e\u0440\u044f\u0442 \u043e Jev<\/a>. Jev \u0442\u0443\u0442, Jev \u0442\u0430\u043c. \u0412\u0441\u0435 \u0432 Twitter \u0442\u043e\u043b\u044c\u043a\u043e \u0438 \u0433\u043e\u0432\u043e\u0440\u044f\u0442 \u043e Jev, \u0447\u0442\u043e \u044d\u0442\u043e \u0441\u043b\u0435\u0434\u0443\u044e\u0449\u0435\u0435 \u0434\u043e\u0441\u0442\u0438\u0436\u0435\u043d\u0438\u0435 \u0431\u043e\u043b\u044c\u0448\u0438\u0445 \u044f\u0437\u044b\u043a\u043e\u0432\u044b\u0445 \u043c\u043e\u0434\u0435\u043b\u0435\u0439 \u0438 \u043f\u0430\u0440\u0430\u0434\u0438\u0433\u043c\u044b \u0418\u0418. \u041c\u044b \u0432 \u044d\u0442\u043e\u043c \u043d\u0435 \u0443\u0432\u0435\u0440\u0435\u043d\u044b. \u041f\u043e\u044d\u0442\u043e\u043c\u0443 \u0432\u043e\u0442 Jev \u0437\u0430 25 \u0441\u0442\u0440\u043e\u043a \u043d\u0430 Python.<\/p>\n

\u0417\u0430\u0433\u0440\u0443\u0437\u0438\u043c \u043c\u043e\u0434\u0435\u043b\u044c.<\/p>\n

# \/\/\/ script# requires-python = \">=3.12\"# dependencies = [\"huggingface-hub\", \"llama-cpp-python\", \"numpy\"]# \/\/\/import numpyfrom llama_cpp import Llama# Really, you can use any GGUF model from https:\/\/huggingface.co\/models?library=ggufmodel = Llama.from_pretrained(    repo_id=\"Qwen\/Qwen3-0.6B-GGUF\",    filename=\"Qwen3-0.6B-Q8_0.gguf\",    n_ctx=512,    logits_all=True,    verbose=False,)<\/code>
<\/a><\/div><\/pre>\n
\u041f\u043e\u0434\u0433\u043e\u0442\u043e\u0432\u0438\u043c \u043f\u0440\u043e\u043c\u043f\u0442 \u0438 \u0443\u043a\u0430\u0436\u0435\u043c \u0432\u0430\u0440\u0438\u0430\u043d\u0442\u044b.<\/p>\n
labels = [\"A\", \"B\", \"C\"]choices = [\"Legitimate\", \"Spam\", \"Phishing\"]email = \"Payroll asks for your password on a non-company sign-in page.\"options = \"\\n\".join(    f\"{label}. {choice}\" for label, choice in zip(labels, choices, strict=True))prompt = f\"\"\"<|im_start|>systemChoose one option.<|im_end|><|im_start|>userEmail: {email}\\n\\n{options}<|im_end|><|im_start|>assistant<think>\\n\\n<\/think>\\n\\n\"\"\"model.eval(tokens=model.tokenize(text=prompt.encode(), add_bos=False, special=True))<\/code>
<\/a><\/div><\/pre>\n
\u0412\u044b\u0436\u043c\u0435\u043c \u0438\u0437 \u043b\u043e\u0433\u0438\u0442\u043e\u0432 \u0432\u0435\u0440\u043e\u044f\u0442\u043d\u043e\u0441\u0442\u0438. (\u043f\u0440\u0438\u043c. \u043f\u0435\u0440\u0435\u0432.: \u043b\u043e\u0433\u0438\u0442\u044b \u2014 \u0441\u044b\u0440\u044b\u0435, \u043d\u0435\u043d\u043e\u0440\u043c\u0438\u0440\u043e\u0432\u0430\u043d\u043d\u044b\u0435 \u043e\u0446\u0435\u043d\u043a\u0438 \u0434\u043b\u044f \u043a\u0430\u0436\u0434\u043e\u0433\u043e \u0432\u0430\u0440\u0438\u0430\u043d\u0442\u0430 \u043e\u0442\u0432\u0435\u0442\u0430).<\/em><\/p>\n
logits = model.scores[model.n_tokens - 1]token_ids = [model.tokenize(text=label.encode(), add_bos=False)[0] for label in labels]choice_logits = numpy.asarray([logits[token_id] for token_id in token_ids])logprobs = choice_logits - numpy.logaddexp.reduce(choice_logits)probabilities = numpy.exp(logprobs)for name, scores in (    (\"Logits\", choice_logits),    (\"Log probabilities\", logprobs),    (\"Probabilities\", probabilities),):    values = numpy.round(scores.astype(float), 3).tolist()    print(f\"{name}:\", dict(zip(choices, values, strict=True)))# Logits: {'Legitimate': 26.254, 'Spam': 27.262, 'Phishing': 29.614}# Log probabilities: {'Legitimate': -3.482, 'Spam': -2.474, 'Phishing': -0.122}# Probabilities: {'Legitimate': 0.031, 'Spam': 0.084, 'Phishing': 0.885}<\/code>
<\/a><\/div><\/pre>\n
\u0412\u043e\u0442. \u042d\u0442\u043e \u0438 \u0435\u0441\u0442\u044c Jev.<\/p>\n
\u041d\u043e \u043d\u0435\u0442, \u0432\u044b \u043d\u0435 \u043f\u043e\u043d\u0438\u043c\u0430\u0435\u0442\u0435 Jev!<\/h3>\n
\u0414\u0430, \u043c\u044b \u0432 \u043a\u0443\u0440\u0441\u0435.<\/p>\n