feat: patch sm103 for 3xfp4 moe generation (#2082)

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

## 📌 Description

Patch sm103 for 3xfp4 moe generation

## 🔍 Related Issues

Following up of #2020 #1925 

## 🚀 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

- [x] I have installed `pre-commit` by running `pip install pre-commit`
(or used your preferred method).
- [x] I have installed the hooks with `pre-commit install`.
- [x] 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

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

## Reviewer Notes

```
$ ls csrc/nv_internal/tensorrt_llm/cutlass_instantiations/103/gemm_grouped
100  103  80

$ pytest tests/moe/test_trtllm_cutlass_fused_moe.py
22 passed, 3 skipped, 1 warning in 771.89s (0:12:51)
```


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

## Summary by CodeRabbit

* **New Features**
* Added support for Blackwell (SM103) GPU architecture in MOE (Mixture
of Experts) operations with specialized CUTLASS-optimized modules.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Author: aleozlx

flashinfer/aot.py

@@ -43,6 +43,7 @@
 from .jit.fp8_quantization import gen_mxfp8_quantization_sm100_module
 from .jit.fused_moe import (
     gen_cutlass_fused_moe_sm120_module,
+    gen_cutlass_fused_moe_sm103_module,
     gen_cutlass_fused_moe_sm100_module,
     gen_cutlass_fused_moe_sm90_module,
     gen_trtllm_gen_fused_moe_sm100_module,
@@ -495,6 +496,7 @@ def gen_all_modules(
             jit_specs.append(gen_tgv_gemm_sm10x_module(torch.float16, use_sm_100f=True))
         if has_sm103:
             jit_specs.append(gen_fp4_quantization_sm103_module())
+            jit_specs.append(gen_cutlass_fused_moe_sm103_module())
         if has_sm110:
             jit_specs.append(gen_fp4_quantization_sm110_module())
         if has_sm120:

flashinfer/fused_moe/__init__.py

@@ -21,6 +21,7 @@
     convert_to_block_layout,
     cutlass_fused_moe,
     gen_cutlass_fused_moe_sm120_module,
+    gen_cutlass_fused_moe_sm103_module,
     gen_cutlass_fused_moe_sm100_module,
     gen_cutlass_fused_moe_sm90_module,
     gen_trtllm_gen_fused_moe_sm100_module,
@@ -39,6 +40,7 @@
     "convert_to_block_layout",
     "cutlass_fused_moe",
     "gen_cutlass_fused_moe_sm120_module",
+    "gen_cutlass_fused_moe_sm103_module",
     "gen_cutlass_fused_moe_sm100_module",
     "gen_cutlass_fused_moe_sm90_module",
     "gen_trtllm_gen_fused_moe_sm100_module",

flashinfer/fused_moe/core.py

@@ -34,6 +34,7 @@
 )
 from ..jit.fused_moe import (
     gen_cutlass_fused_moe_sm120_module,
+    gen_cutlass_fused_moe_sm103_module,
     gen_cutlass_fused_moe_sm100_module,
     gen_cutlass_fused_moe_sm90_module,
     gen_cutlass_fused_moe_sm89_module,
@@ -315,7 +316,9 @@ def convert_to_block_layout(input_tensor: torch.Tensor, blockK: int) -> torch.Te
 def get_cutlass_fused_moe_module(backend: str = "100", use_fast_build: bool = False):
     if backend in ("120", "121"):
         module = gen_cutlass_fused_moe_sm120_module(use_fast_build).build_and_load()
-    elif backend in ("100", "103", "110"):
+    elif backend == "103":
+        module = gen_cutlass_fused_moe_sm103_module(use_fast_build).build_and_load()
+    elif backend in ("100", "110"):
         module = gen_cutlass_fused_moe_sm100_module(use_fast_build).build_and_load()
     elif backend == "90":
         module = gen_cutlass_fused_moe_sm90_module(use_fast_build).build_and_load()

flashinfer/jit/fused_moe.py

@@ -47,6 +47,24 @@ def gen_cutlass_fused_moe_sm120_module(use_fast_build: bool = False) -> JitSpec:
     return gen_cutlass_fused_moe_module(nvcc_flags, "120", use_fast_build)
 
 
+def gen_cutlass_fused_moe_sm103_module(use_fast_build: bool = False) -> JitSpec:
+    nvcc_flags = [
+        "-DCOMPILE_BLACKWELL_TMA_GEMMS",
+        "-DCOMPILE_BLACKWELL_TMA_GROUPED_GEMMS",
+        "-DENABLE_BF16",
+        "-DENABLE_FP8",
+        "-DENABLE_FP4",
+        "-DUSING_OSS_CUTLASS_MOE_GEMM",
+        "-DCOMPILE_BLACKWELL_SM103_TMA_GROUPED_GEMMS",
+    ]
+
+    nvcc_flags += current_compilation_context.get_nvcc_flags_list(
+        supported_major_versions=[10]
+    )
+
+    return gen_cutlass_fused_moe_module(nvcc_flags, "103", use_fast_build)
+
+
 def gen_cutlass_fused_moe_sm100_module(use_fast_build: bool = False) -> JitSpec:
     nvcc_flags = [
         "-DCOMPILE_BLACKWELL_TMA_GEMMS",