From ffb58254070bd38a7805d2480effcdb0c5c74129 Mon Sep 17 00:00:00 2001 From: shuiyu1123 Date: Thu, 18 Jun 2026 18:43:47 +0000 Subject: [PATCH] fix: correct vocab_size to max_id+1 (was len(vocab)) --- scripts/train_cnb.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/scripts/train_cnb.py b/scripts/train_cnb.py index 5f13467..e4ca109 100644 --- a/scripts/train_cnb.py +++ b/scripts/train_cnb.py @@ -84,7 +84,7 @@ def main(): delattr(tokenizer_instance, attr) # ====== 模型 ====== - model = LLMModel(vocab_size=len(vocab), d_model=320, n_heads=8, + model = LLMModel(vocab_size=max(vocab.values()) + 1, d_model=320, n_heads=8, n_layers=6, d_ff=1536, max_len=256, dropout=0.1, use_memory=False, use_thinking=False, use_moe=False) total = sum(p.numel() for p in model.parameters())