[refact] unified soc_version code (#4359)

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

Currently, there are two paths to judge the chip type in code,
`get_ascend_soc_version` use `get_soc_version` api in torch_npu, and
`is_310p` `use _build_info.__soc_version__`, which generate when
install. We need to unify the two paths.

We need to unify these codes based on the following points:

1. We need to ensure consistency in chip type judgment between compiling
and running states;
2. In compiling state, we need chip type to complete op's compilation,
but in running state, we only need device
type(910B/910_93/310P/910_95/etc) to make code branch judgement;
3. In compiling state, torch_npu may not have been installed yet, so we
can't use torch_npu's api.

Based on the above points, we have made the following changes:

1. When user set env `SOC_VERSION`, use it; when not set, query
soc_version by `npu-smi`;
2. generate device_type based on soc_version when compiling, and write
`__device_type__` instead of `__soc_version__` in `_build_info.py`;
3. In running state, use `__device_type__` to judge code branch.

### Does this PR introduce _any_ user-facing change?

When not set env `SOC_VERSION`, it will not be `ASCEND910B1` by default,
we will query soc_version by `npu-smi`. And env `SOC_VERSION` must be in
the list `soc_to_device` in `setup.py`.

- vLLM version: v0.11.0
- vLLM main:
2918c1b49c

Signed-off-by: zzzzwwjj <1183291235@qq.com>

vllm_ascend/torchair/ops/torchair_activation.py

@@ -28,9 +28,9 @@ def torchair_silu_and_mul_forward_oot(self, x: torch.Tensor) -> torch.Tensor:
 
     import torch_npu
 
-    from vllm_ascend.utils import is_310p
+    from vllm_ascend.utils import AscendDeviceType, get_ascend_device_type
 
-    if is_310p():
+    if get_ascend_device_type() == AscendDeviceType._310P:
         out = torch_npu.npu_swiglu(x.to(torch.float32)).to(torch.float16)
     else:
         out = torch_npu.npu_swiglu(x)

vllm_ascend/torchair/ops/torchair_fused_moe.py

@@ -51,8 +51,8 @@ from vllm_ascend.torchair.utils import (get_all_reduce_merge_state,
                                         get_rm_router_logits_state,
                                         npu_stream_switch, npu_wait_tensor,
                                         super_kernel)
-from vllm_ascend.utils import (AscendSocVersion, dispose_tensor,
-                               get_ascend_soc_version, is_310p,
+from vllm_ascend.utils import (AscendDeviceType, dispose_tensor,
+                               get_ascend_device_type,
                                is_hierarchical_communication_enabled)
 
 
@@ -75,11 +75,11 @@ def torchair_fused_experts_with_mc2(
     ep_world_size = moe_parallel_config.ep_size
 
     # NOTE: Currently, when in A3 or in torchair graph, we need to pass in some extra param into dispatch & combine
-    need_extra_args = (get_ascend_soc_version() == AscendSocVersion.A3
+    need_extra_args = (get_ascend_device_type() == AscendDeviceType._910_93
                        or is_torchair)
 
     # NOTE: Currently, when in A3, we need to pass in some extra param into dispatch & combine
-    a3_need_extra_args = get_ascend_soc_version() == AscendSocVersion.A3
+    a3_need_extra_args = get_ascend_device_type() == AscendDeviceType._910_93
     # NOTE: When in A2, setting the environment variables HCCL_INTRA_PCIE_ENABLE=1 and
     # HCCL_INTRA_ROCE_ENABLE=0 can reduce cross-machine communication traffic and significantly
     # improve communication performance.
@@ -467,7 +467,7 @@ def torchair_fused_experts_moge(
         group_list=group_list,
     )[0]
 
-    if is_310p():
+    if get_ascend_device_type() == AscendDeviceType._310P:
         gate_up_out = torch_npu.npu_swiglu(gate_up_out.to(torch.float32)).to(
             torch.float16)
     else:

vllm_ascend/torchair/ops/torchair_layernorm.py

@@ -57,9 +57,9 @@ def torchair_rmsnorm_forward_oot(
 
     import torch_npu
 
-    from vllm_ascend.utils import is_310p
+    from vllm_ascend.utils import AscendDeviceType, get_ascend_device_type
     if residual is not None:
-        if is_310p():
+        if get_ascend_device_type() == AscendDeviceType._310P:
             orig_dtype = residual.dtype
             x = x + residual.to(x.dtype)
             residual = x.to(orig_dtype)

vllm_ascend/torchair/ops/torchair_rotary_embedding.py

@@ -25,7 +25,8 @@ from vllm.model_executor.layers.rotary_embedding import (
     DeepseekScalingRotaryEmbedding, RotaryEmbedding)
 
 from vllm_ascend.ascend_config import get_ascend_config
-from vllm_ascend.utils import enable_custom_op, is_310p
+from vllm_ascend.utils import (AscendDeviceType, enable_custom_op,
+                               get_ascend_device_type)
 
 
 def custom_rotary_embedding_enabled(query, neox_style, head_size):
@@ -60,8 +61,9 @@ def rope_forward_oot(
     if is_neox_style_override is not None:
         neox_style = is_neox_style_override
     # adopt custom kernel path for rotary_embedding
-    if custom_rotary_embedding_enabled(query, neox_style,
-                                       self.head_size) and not is_310p():
+    if custom_rotary_embedding_enabled(
+            query, neox_style, self.head_size) and get_ascend_device_type(
+            ) != AscendDeviceType._310P:
         query, key = torch.ops._C_ascend.rotary_embedding(
             positions,
             query,