[Perf] Warm up DeepGEMM MegaMoE kernels before serving (#60468)

Signed-off-by: Simon Veitner <sveitner@redhat.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Roberto L. Castro <38211239+LopezCastroRoberto@users.noreply.github.com>
This commit is contained in:
simon-veitner-redhat
2026-10-13 20:17:48 +08:00
committed by GitHub
co-authored by Claude Roberto L. Castro
parent d9a543c500
commit 253e02b95e
3 changed files with 203 additions and 1 deletions
+1
View File
@@ -566,6 +566,7 @@ class KimiMoE(nn.Module):
)
self.latent_moe_use_norm = config.latent_moe_use_norm
self.tp_size = get_tensor_model_parallel_world_size()
self.use_sequence_parallel = use_sequence_parallel
self.routed_scaling_factor = config.routed_scaling_factor
self.moe_renormalize = moe_renormalize
self.use_grouped_topk = config.use_grouped_topk