Qwen2.5使用moba之后显示sliding window attention相关问题

[xvlincaigou]

我对test/llama.py做了一个修改，把--model改为了Qwen2.5-1.5B-Instruct，其config文件如下：
{
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151645,
"hidden_act": "silu",
"hidden_size": 1536,
"initializer_range": 0.02,
"intermediate_size": 8960,
"max_position_embeddings": 32768,
"max_window_layers": 28,
"model_type": "qwen2",
"num_attention_heads": 12,
"num_hidden_layers": 28,
"num_key_value_heads": 2,
"rms_norm_eps": 1e-06,
"rope_theta": 1000000.0,
"sliding_window": 32768,
"tie_word_embeddings": true,
"torch_dtype": "bfloat16",
"transformers_version": "4.44.0",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
运行之后输出了Sliding Window Attention is enabled but not implemented for moba; unexpected results may be encountered.
想请教一下这是怎么回事？

[derby-ding]

遇到了相同问题，"use_sliding_window": false，已经设为false了，请问是不需要理会么，还是改为使用其他无此参数的模型。

[derby-ding]

更换到llama3.1，不再提示问题。