[QUANT] Add GPTQModel Dynamic Quantization + lm_head Quantization (#3790)

Signed-off-by: ZX-ModelCloud <zx@modelcloud.ai>
Co-authored-by: ZX-ModelCloud <zx@modelcloud.ai>

python/sglang/srt/layers/radix_attention.py

@@ -34,6 +34,7 @@ class RadixAttention(nn.Module):
         v_head_dim: int = -1,
         sliding_window_size: int = -1,
         is_cross_attention: bool = False,
+        prefix: str = "",
     ):
         super().__init__()
         self.tp_q_head_num = num_heads