RuntimeError: Unsupported compute type Float

使用 decoupleQ 量化了一个 llama 33B的模型，推理时报错
```
Traceback (most recent call last):
File "/mnt/afs/quantization/test/decoupleQ/llama.py", line 476, in
model_output = model.generate(input_token_ids_tensor, max_length=40, do_sample=False)
...
File "/mnt/afs/quantization/test/decoupleQ/decoupleQ/linear_w2a16.py", line 36, in forward
output = dQ_asymm_qw2_gemm(input, self.weight, self.scale, self.zp, self.bias, self.group_size)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: Unsupported compute type Float
```
请问你们是否有遇到过该问题？

谢谢！

Start with the failing call in /mnt/afs/quantization/test/decoupleQ/decoupleQ/linear_w2a16.py and the model generation path in llama.py. Reproduce the Float compute-type error, then trace the supported compute types and document or fix the conditions that trigger it. Done means the 33B inference path no longer fails or the limitation is clearly recorded.

Written by the indexing model from the issue text.