tests: skip non SM100/103 for grouped deepgemm (#1767)

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

## 📌 Description

skip test_fp8_groupwise_group_deepgemm and
test_fp8_groupwise_group_deepgemm where SM not 100 and 103, add relevant
checks in library

## 🔍 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.
-->

---------

Co-authored-by: jimmzhou <[email protected]>

Author: jimmyzho

flashinfer/gemm.py

@@ -3078,6 +3078,11 @@ def group_deepgemm_fp8_nt_groupwise(
     """
     from flashinfer.deep_gemm import m_grouped_fp8_gemm_nt_contiguous
 
+    if not _match_sm_version(a.device, ["100", "103"]):
+        raise ValueError(
+            "m_grouped_fp8_gemm_nt_contiguous is only supported on SM100, SM100, SM103."
+        )
+
     if out is None:
         out_dtype = out_dtype or torch.bfloat16
         out = torch.empty(a.shape[0], b.shape[1], dtype=out_dtype, device=a.device)
@@ -3206,6 +3211,11 @@ def batch_deepgemm_fp8_nt_groupwise(
     """
     from flashinfer.deep_gemm import m_grouped_fp8_gemm_nt_masked
 
+    if not _match_sm_version(a.device, ["100", "103"]):
+        raise ValueError(
+            "m_grouped_fp8_gemm_nt_masked is only supported on SM100, SM103."
+        )
+
     if out is None:
         out_dtype = out_dtype or torch.bfloat16
         out = torch.empty(

tests/test_groupwise_scaled_gemm_fp8.py

@@ -202,6 +202,11 @@ def test_fp8_groupwise_group_deepgemm(
     group_size,
     out_dtype,
 ):
+    compute_capability = get_compute_capability(torch.device(device="cuda"))
+    if compute_capability[0] != 10:
+        pytest.skip(
+            "group_deepgemm_fp8_nt_groupwise is only supported on SM100, SM103 in trtllm backend."
+        )
     torch.random.manual_seed(0)
     m_per_group = m // group_size
     if m_per_group < 128:
@@ -245,6 +250,11 @@ def test_fp8_groupwise_batch_deepgemm_masked(
     group_size,
     out_dtype,
 ):
+    compute_capability = get_compute_capability(torch.device(device="cuda"))
+    if compute_capability[0] != 10:
+        pytest.skip(
+            "batch_deepgemm_fp8_nt_groupwise is only supported on SM100, SM103."
+        )
     torch.random.manual_seed(0)
     n, k = nk
     a = torch.randn((group_size, m, k), device="cuda", dtype=torch.float32)