fix: graniteMoE does support PP (but requires enforce-eager)

Signed-off-by: Travis Johnson <[email protected]>

Author: tjohnson31415

docs/source/models/supported_models.md

@@ -297,12 +297,12 @@ See [this page](#generative-models) for more information on how to use generativ
   * Granite 3.0 MoE, PowerMoE
   * `ibm-granite/granite-3.0-1b-a400m-base`, `ibm-granite/granite-3.0-3b-a800m-instruct`, `ibm/PowerMoE-3b`, etc.
   * ✅︎
-  *
+  * ✅︎
 - * `GraniteMoeSharedForCausalLM`
   * Granite MoE Shared
   * `ibm-research/moe-7b-1b-active-shared-experts` (test model)
   * ✅︎
-  *
+  * ✅︎
 - * `GritLM`
   * GritLM
   * `parasail-ai/GritLM-7B-vllm`.

vllm/model_executor/models/granitemoeshared.py

@@ -28,7 +28,7 @@
 
 from . import mixtral
 from .granitemoe import GraniteMoeAttention, GraniteMoeMoE
-from .interfaces import SupportsLoRA
+from .interfaces import SupportsLoRA, SupportsPP
 from .utils import make_layers, maybe_prefix
 
 
@@ -208,7 +208,7 @@ def forward(
         return hidden_states
 
 
-class GraniteMoeSharedForCausalLM(nn.Module, SupportsLoRA):
+class GraniteMoeSharedForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
     fall_back_to_pt_during_load = False
 
     packed_modules_mapping = {