[Quantization] Support compressed tensors w8a8 static and w8a8 dynamic weight (#4036)

### What this PR does / why we need it?

While using the LLM Compressor quantization tool from the VLLM community
to generate quantized weights, the VLLM Ascend engine needs to be
adapted to support the compressed tensors quantization format.

1. Add AscendCompressedTensorsConfig to replace CompressedTensorsConfig
in vllm.
2. Support CompressedTensorsW8A8 static weight.
- weight: per-channel, int8, symmetric; activation: per-tensor, int8,
symmetric.
4. Support CompressedTensorsW8A8Dynamic weight.
- weight: per-channel, int8, symmetric; activation: per-token, int8,
symmetric, dynamic.
5. Modify the override_quantization_method in AscendQuantConfig.

Co-authored-by: taoqun110 taoqun@huawei.com
Co-authored-by: chenxi-hh chen464822955@163.com

- vLLM version: v0.11.2

---------

Signed-off-by: LHXuuu <scut_xlh@163.com>
Signed-off-by: chenxi-hh <chen464822955@163.com>
Signed-off-by: chenxi-hh <32731611+chenxi-hh@users.noreply.github.com>
Co-authored-by: chenxi-hh <chen464822955@163.com>
Co-authored-by: chenxi-hh <32731611+chenxi-hh@users.noreply.github.com>

vllm_ascend/quantization/w8a8.py

@@ -25,7 +25,8 @@ from vllm.forward_context import get_forward_context
 
 from vllm_ascend.attention.attention_v1 import AscendAttentionState
 from vllm_ascend.ops.fused_moe.experts_selector import select_experts
-from vllm_ascend.utils import (ACL_FORMAT_FRACTAL_NZ, AscendDeviceType,
+from vllm_ascend.utils import (ACL_FORMAT_FRACTAL_NZ,
+                               COMPRESSED_TENSORS_METHOD, AscendDeviceType,
                                get_ascend_device_type, is_enable_nz)
 
 
@@ -149,6 +150,10 @@ class AscendW8A8LinearMethod:
                 )
 
         quant_bias = layer.quant_bias if tp_rank == 0 else None
+        if getattr(layer, "ascend_quant_method",
+                   "") == COMPRESSED_TENSORS_METHOD:
+            quant_bias = bias
+
         if get_ascend_device_type() == AscendDeviceType._310P:
             # On 300I Duo platform, we need transpose again if
             # using nz. This transpose can be skipped in torchair.
@@ -187,6 +192,11 @@ class AscendW8A8LinearMethod:
                 layer.weight.data, ACL_FORMAT_FRACTAL_NZ)
         layer.weight_scale.data = torch.flatten(layer.weight_scale.data)
         layer.weight_offset.data = torch.flatten(layer.weight_offset.data)
+        if getattr(layer, "ascend_quant_method",
+                   "") == COMPRESSED_TENSORS_METHOD:
+            deq_scale = layer.input_scale.data * layer.weight_scale.data
+            layer.deq_scale = torch.nn.Parameter(deq_scale,
+                                                 requires_grad=False)
 
 
 class AscendW8A8FusedMoEMethod: