use_model.py

"""Yonetix-V1 快速使用脚本"""
import json
import torch
from model import YonetixTransformer, ModelConfig

# 加载模型
with open("config.json") as f:
    config = ModelConfig(**json.load(f))
model = YonetixTransformer(config)
model.load_state_dict(torch.load("model.pt", map_location="cpu", weights_only=True))
model.eval()
print(f"✅ 模型加载完成，参数量: {sum(p.numel() for p in model.parameters()) / 1e6:.2f}M")

# 加载词表
vocab = {}
with open("vocab.txt") as f:
    for line in f:
        idx, token = line.strip().split("\t")
        vocab[int(idx)] = token

PAD, UNK, BOS, EOS = 0, 1, 2, 3

def encode(text, max_len=128):
    ids = [BOS]
    for ch in text:
        found = False
        for idx, token in vocab.items():
            if idx < 4:
                continue
            if token == ch:
                ids.append(idx)
                found = True
                break
        if not found:
            ids.append(UNK)
    ids.append(EOS)
    return torch.tensor([ids[:max_len]])

def decode(ids):
    return "".join(vocab.get(int(i), "") for i in ids if int(i) > 3)

def chat(prompt, max_new_tokens=128):
    input_ids = encode(prompt)
    output_ids = model.generate(input_ids, max_new_tokens=max_new_tokens)
    reply = output_ids[0][input_ids.size(1):].tolist()
    return decode(reply)

# 测试
print("\n🤖 Yonetix-V1 测试对话")
print("=" * 40)
tests = [
    "你好，你是谁？",
    "介绍一下YONETIX",
    "你会做什么？",
]
for t in tests:
    print(f"\n👤 > {t}")
    print(f"🤖 > {chat(t)}")