tencent/HY-MT1.5-1.8B を Q5_K_M (5.78 BPW) で量子化した GGUF ファイルです。llama-quantize from llama.cpp b9093)Q5_K_M quantization by DokodemoTranslate maintainer, derived fromtencent/HY-MT1.5-1.8B(FP16 safetensors). Conversion pipeline:convert_hf_to_gguf.py --outtype f16→llama-quantize ... Q5_K_M. No model architecture, weights training, or vocabulary modification.
Tencent HY is licensed under the Tencent HY Community License Agreement, Copyright © 2025 Tencent. All Rights Reserved. The trademark rights of "Tencent HY" are owned by Tencent or its affiliate.
"Territory" shall mean the worldwide territory, excluding the territory of the European Union, United Kingdom and South Korea.You must not use, reproduce, modify, distribute, or display the Tencent HY Works ... outside the Territory.
1{
2 "top_k": 20,
3 "top_p": 0.6,
4 "repetition_penalty": 1.05,
5 "temperature": 0.7
6}将以下文本翻译为{target},注意只需要输出翻译后的结果,不要额外解释:
{source_text}Translate the following segment into {target}, without additional explanation.
{source_text}hunyuan-dense アーキテクチャがサポートされます。1./llama-cli -m hy-mt1.5-1.8b-q5_k_m.gguf \
2 -p "<|hy_begin▁of▁sentence|><|hy_User|>Translate the following segment into Japanese, without additional explanation.\n\nHello, world.<|hy_Assistant|>" \
3 --temp 0.7 --top-k 20 --top-p 0.6 --repeat-penalty 1.05 \
4 --no-context-shift -n 256convert_hf_to_gguf.py が tokenizer.ggml.eos_token_id を 誤って 3 ('$') として書き込みます (本来は 120020 = <|hy_place▁holder▁no▁2|>)。'$' を含むテキストで生成が即停止する潜在バグになります。1// llama_vocab_is_eog は信用せず、専用トークンを tokenize して逆引き
2let stopTokens: Set<llama_token> = [
3 tokenize("<|hy_place▁holder▁no▁2|>", special: true),
4 tokenize("<|hy_User|>", special: true),
5]