Remove assertions about per group quant fp8 (#8717)

python/sglang/srt/layers/quantization/fp8_kernel.py

@@ -354,10 +354,6 @@ def sglang_per_token_group_quant_fp8(
     ), "the last dimension of `x` cannot be divisible by `group_size`"
     assert x.is_contiguous(), "`x` is not contiguous"
 
-    if scale_ue8m0:
-        # TODO: handle this case by fixing the (token=4, dim=256, group_size=128) UT case
-        assert x.shape[-1] % (group_size * 4) == 0
-
     x_q = torch.empty_like(x, device=x.device, dtype=fp8_dtype)
     x_s = create_per_token_group_quant_fp8_output_scale(
         x_shape=x.shape,