Feat/add fi selective state update kernel call (#18070)

Signed-off-by: Shahar Mor <smor@nvidia.com>
2026-02-19 16:56:06 +08:00
parent 0be30d4b0d
commit 82a0bafc1c
8 changed files with 366 additions and 2 deletions
@@ -26,6 +26,19 @@ class TestNvidiaNemotron3Nano30BBF16(LMEvalMixin, DefaultServerBase):
    ] + NEMOTRON_3_NANO_THINKING_ARGS


+class TestNvidiaNemotron3Nano30BBF16FlashInfer(LMEvalMixin, DefaultServerBase):
+    """Test Nemotron-3-Nano-30B BF16 model with lm-eval GSM8K evaluation using flashinfer mamba backend."""
+
+    model = "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16"
+    model_config_name = "lm_eval_configs/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16.yaml"
+    other_args = [
+        "--tp-size",
+        "2",
+        "--mamba-backend",
+        "flashinfer",
+    ] + NEMOTRON_3_NANO_THINKING_ARGS
+
+
 class TestNvidiaNemotron3Nano30BFP8(LMEvalMixin, DefaultServerBase):
    """Test Nemotron-3-Nano-30B FP8 model with lm-eval GSM8K evaluation."""