Views
No views yet
waahsvt → waah + svt1請將以下粵語漢字轉寫為 Liujgoj 羅馬字：
2喺江南靠近海邊一條大路之上，有一隊清兵，手執刀槍，押住七架囚車，向住北方行緊。1Hair Gungjnaamx koihzingr hoijbinh yatj tiux daaihlukh zijsoengh,
2yauq yatj deoiraaj cingjbingj, saur zapj doujcoengj,
3aat zvh catj gaar kauxcej, hoeng zvh bouhfongx haangxganr.1請將以下 Liujgoj 羅馬字轉寫為粵語漢字：
2Waahsvt hair Cingjciux ge Hongjheij cojninx, dungjtinj ge mauq yatj yath.
3batjfungj yvx douj, munq deih bingjsoengj.話説喺清初嘅黃海戰爭，當天嘅某日。請將「waahsvt」拆分成音節，並標註音標。waah + svttransformers 載入模型：1import torch
2from transformers import AutoModel, AutoTokenizer
3
4model_id = "Yvthyvq/Liujgoj-Cantonese-Qwen2.5-Omni-7B-SFT"
5
6print("⏳ 載入模型中...")
7
8tokenizer = AutoTokenizer.from_pretrained(
9 model_id,
10 trust_remote_code=True
11)
12
13model = AutoModel.from_pretrained(
14 model_id,
15 torch_dtype=torch.bfloat16,
16 device_map="auto",
17 trust_remote_code=True
18)
19
20messages = [
21 {
22 "role": "user",
23 "content": "請將以下粵語漢字轉寫為 Liujgoj 羅馬字：\n喺江南靠近海邊一條大路之上，有一隊清兵，手執刀槍，押住七架囚車，向住北方行緊。"
24 }
25]
26
27text = tokenizer.apply_chat_template(
28 messages,
29 tokenize=False,
30 add_generation_prompt=True
31)
32
33inputs = tokenizer(
34 [text],
35 return_tensors="pt"
36).to("cuda")
37
38with torch.no_grad():
39 outputs = model.generate(
40 **inputs,
41 max_new_tokens=256,
42 temperature=0.1
43 )
44
45input_len = inputs.input_ids.shape[1]
46
47response = tokenizer.decode(
48 outputs[0][input_len:],
49 skip_special_tokens=True
50)
51
52print("Generated Response:")
53print(response.strip())1Qwen2.5-Omni-7B
2 │
3 ▼
4Liujgoj Cantonese CPT
5 │
6 ▼
7Liujgoj Cantonese SFT
8 │
9 ▼
10 ASR