fix: compile flags for trtllm fmha_v2 (#2175)

<!-- .github/pull_request_template.md -->

## 📌 Description

<!-- What does this PR do? Briefly describe the changes and why they’re
needed. -->

## 🔍 Related Issues

<!-- Link any related issues here -->

## 🚀 Pull Request Checklist

Thank you for contributing to FlashInfer! Before we review your pull
request, please make sure the following items are complete.

### ✅ Pre-commit Checks

- [ ] I have installed `pre-commit` by running `pip install pre-commit`
(or used your preferred method).
- [ ] I have installed the hooks with `pre-commit install`.
- [ ] I have run the hooks manually with `pre-commit run --all-files`
and fixed any reported issues.

> If you are unsure about how to set up `pre-commit`, see [the
pre-commit documentation](https://pre-commit.com/).

## 🧪 Tests

- [ ] Tests have been added or updated as needed.
- [ ] All tests are passing (`unittest`, etc.).

## Reviewer Notes

<!-- Optional: anything you'd like reviewers to focus on, concerns, etc.
-->


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Chores**
  * Removed noisy runtime console prints during build/generation.
* Updated CUDA compiler requirements to target CUDA 12 and added a new
compiler flag for compatibility.

* **Bug Fixes**
* Added an early check that raises a clear error on unsupported GPU
devices (SM120a), preventing misruns.

* **Tests**
* Test now skips automatically when the required SM120a GPU support is
not present.

<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Author: jimmyzho

.pre-commit-config.yaml

@@ -3711,10 +3711,10 @@ def generate_files(specs_names):
     ]
     if "CUDA_PATH" in os.environ:
         cmd[0] = os.environ["CUDA_PATH"] + "/bin/" + cmd[0]
-    print('Running command "{}" to build "bin/print_traits.exe":'.format(" ".join(cmd)))
+    # print('Running command "{}" to build "bin/print_traits.exe":'.format(" ".join(cmd)))
     process = subprocess.Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE)
     output, error = process.communicate()
-    print('Running "bin/print_traits.exe":')
+    # print('Running "bin/print_traits.exe":')
     process = subprocess.Popen(
         "bin/print_traits.exe", stdin=subprocess.PIPE, stdout=subprocess.PIPE
     )

flashinfer/jit/attention/fmha_v2/generator_utils.py

@@ -1901,9 +1901,10 @@ def gen_trtllm_fmha_v2_module() -> JitSpec:
     source_paths = kernel_paths + [binding_source_path]
 
     nvcc_flags = current_compilation_context.get_nvcc_flags_list(
-        supported_major_versions=[10, 11, 12]
+        supported_major_versions=[12]
     )
     nvcc_flags.append(f"-I{jit_env.FLASHINFER_CSRC_DIR / 'fmha_v2'}")
+    nvcc_flags.append("-Wno-deprecated-gpu-targets")
 
     return gen_jit_spec(
         uri,

flashinfer/jit/attention/modules.py

@@ -3603,6 +3603,8 @@ def fmha_v2_prefill_deepseek(
         If return_lse is True, the output will be a tuple of two tensors, the first is the output tensor, the second is the lse tensor.
         If return_lse is False, the output will be a single tensor.
     """
+    if not is_sm120a_supported(query.device):
+        raise ValueError("fmha_v2_prefill_deepseek is only supported on SM120 GPUs.")
     assert query.shape[3] == 192 and key.shape[3] == 192 and value.shape[3] == 128, (
         "currently only support deepseek r1 192 query and 128 value"
     )

flashinfer/prefill.py

@@ -5,6 +5,7 @@
 
 from flashinfer.prefill import fmha_v2_prefill_deepseek
 from tests.utils_fp8 import to_float8
+from flashinfer.utils import is_sm120a_supported
 
 
 def attention_ref(
@@ -56,6 +57,8 @@ def attention_ref(
 def test_fmha_v2_prefill_deepseek(
     batch_size, num_heads, head_dim_qk, head_dim_v, seq_len, qkv_dtype, o_dtype
 ):
+    if not is_sm120a_supported(torch.device("cuda")):
+        pytest.skip("fmha_v2_prefill_deepseek is only supported on SM120 GPUs.")
     torch.manual_seed(42)
 
     def initialize_tensors(batch_size, num_heads, head_dim_qk, head_dim_v, seq_len):