[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/attention/attention_v1.py

@@ -42,9 +42,9 @@ from vllm_ascend.attention.utils import (AscendCommonAttentionMetadata,
 from vllm_ascend.compilation.acl_graph import (get_graph_params,
                                                update_graph_params_workspaces)
 from vllm_ascend.ops.attention import vanilla_chunked_prefill
-from vllm_ascend.utils import (ACL_FORMAT_FRACTAL_NZ, aligned_16, is_310p,
-                               nd_to_nz_2d, nd_to_nz_spec,
-                               prefill_context_parallel_enable,
+from vllm_ascend.utils import (ACL_FORMAT_FRACTAL_NZ, AscendDeviceType,
+                               aligned_16, get_ascend_device_type, nd_to_nz_2d,
+                               nd_to_nz_spec, prefill_context_parallel_enable,
                                weak_ref_tensors)
 
 # isort: off
@@ -83,7 +83,7 @@ class AscendAttentionBackend(AttentionBackend):
         num_kv_heads: int,
         head_size: int,
     ) -> Tuple[int, ...]:
-        if is_310p():
+        if get_ascend_device_type() == AscendDeviceType._310P:
             return (2, num_blocks, num_kv_heads * head_size // 16, block_size,
                     16)
         return (2, num_blocks, block_size, num_kv_heads, head_size)
@@ -351,7 +351,7 @@ class AscendAttentionMetadataBuilder:
         query_start_loc = query_start_loc_cpu.to(self.device,
                                                  non_blocking=True)
 
-        if is_310p():
+        if get_ascend_device_type() == AscendDeviceType._310P:
             if attn_state == AscendAttentionState.PrefillNoCache:
                 mask_nz = nd_to_nz_2d(attn_mask)
                 attn_mask = torch_npu.npu_format_cast(mask_nz.contiguous(),
@@ -702,7 +702,7 @@ class AscendAttentionBackendImpl(AttentionImpl):
 
         mask = attn_metadata.attn_mask
 
-        if is_310p():
+        if get_ascend_device_type() == AscendDeviceType._310P:
             # align q k v output tensors
             query = aligned_16(query)
             key = aligned_16(key)
@@ -783,7 +783,7 @@ class AscendAttentionBackendImpl(AttentionImpl):
         attn_metadata: AscendMetadata,
         output: Optional[torch.Tensor] = None,
     ) -> torch.Tensor:
-        if is_310p():
+        if get_ascend_device_type() == AscendDeviceType._310P:
             # seq_lens_tensor needs to be transferred to the device for 310P.
             attn_metadata.seq_lens = \
                 attn_metadata.seq_lens.to(device=query.device)
@@ -857,7 +857,7 @@ class AscendAttentionBackendImpl(AttentionImpl):
         assert attn_metadata is not None
         assert attn_metadata.attn_mask is not None
 
-        if is_310p():
+        if get_ascend_device_type() == AscendDeviceType._310P:
             # Do reformat in case of broadcasted tensors.
             attn_metadata.attn_mask = \
                 torch_npu.npu_format_cast(attn_metadata.attn_mask.contiguous(),