[None][test] AutoDeploy: enable gpt-oss registry accuracy tests

- Uncomment gpt-oss-{20b,120b} in models.yaml; registry-driven accuracy
  tests raised ValueError ("not found in model registry") without them.
- gpt-oss world sizes: 20b=1, 120b=2; test_mxfp4_gsm8k now requires 2+ Blackwell GPUs.
- gpt_oss.yaml: vocab-parallel lm_head (shard_layers+=lm_head) + AUTO allreduce for TP.
- Gotcha: re-exposes these models to the external AutoDeploy dashboard.

Signed-off-by: Taylor Yeonbok Lee <249374542+taylor-yb-lee@users.noreply.github.com>

Author: taylor-yb-lee

examples/auto_deploy/model_registry/configs/gpt_oss.yaml

@@ -30,7 +30,10 @@ transforms:
   apply_sharding_hints:
     enabled: true
     requires_shape_prop: true
-    shard_layers: ["mha", "moe"]
+    shard_layers: ["mha", "moe", "lm_head"]  # V3: vocab-parallel lm_head (colwise + all_gather)
+    # TP2 trial: AUTO -> tunable_allreduce picks oneshot-lamport fused AR.
+    # NCCL has no fused residual+rmsnorm kernel -> RING_LL + separate rmsnorm.
+    allreduce_strategy: AUTO
   quantize_mxfp4_moe:
     backend: trtllm
     trtllm_quant_act: mxfp8

examples/auto_deploy/model_registry/models.yaml

@@ -168,10 +168,9 @@ models:
 - name: nvidia/Mistral-NeMo-Minitron-8B-Base
   config_id: default_ws_2
   yaml_extra: ['dashboard_default.yaml', 'world_size_2.yaml']
-# OOM during AutoDeploy run.
-# - name: openai/gpt-oss-20b
-#   config_id: gpt_oss
-#   yaml_extra: ['gpt_oss.yaml', 'world_size_1.yaml']
+- name: openai/gpt-oss-20b
+  config_id: gpt_oss
+  yaml_extra: ['gpt_oss.yaml', 'world_size_1.yaml']
 - name: ibm-granite/granite-3.0-8b-instruct
   config_id: default_ws_1
   yaml_extra: ['dashboard_default.yaml', 'world_size_1.yaml']
@@ -333,10 +332,9 @@ models:
 # - name: meta-llama/Llama-3.2-90B-Vision-Instruct
 #   config_id: multimodal
 #   yaml_extra: ['dashboard_default.yaml', 'world_size_8.yaml', 'multimodal.yaml']
-# torch.distributed.DistStoreError: Timed out after 601 seconds waiting for clients. 1/4 clients joined.
-# - name: openai/gpt-oss-120b
-#   config_id: gpt_oss
-#   yaml_extra: ['gpt_oss.yaml', 'world_size_1.yaml']
+- name: openai/gpt-oss-120b
+  config_id: gpt_oss
+  yaml_extra: ['gpt_oss.yaml', 'world_size_2.yaml']
 # [RANK 3] Error querying confidential compute state: Function Not Found
 # - name: meta-llama/Llama-4-Scout-17B-16E-Instruct
 #   config_id: multimodal__llama4_scout

tests/integration/defs/accuracy/test_llm_api_autodeploy.py

@@ -1276,7 +1276,6 @@ class TestGPTOSS(LlmapiAccuracyTestHarness):
             "openai/gpt-oss-20b",
             None,
             None,
-            marks=pytest.mark.skip_less_device(2),
             id="20b",
         ),
         pytest.param(
@@ -1291,20 +1290,19 @@ class TestGPTOSS(LlmapiAccuracyTestHarness):
             "openai/gpt-oss-120b",
             2,
             "tp",
-            marks=pytest.mark.skip_less_device(2),
             id="120b-tp2",
         ),
         pytest.param(
             "120b",
             "openai/gpt-oss-120b",
             2,
             "ep",
-            marks=pytest.mark.skip_less_device(2),
             id="120b-ep2",
         ),
     ]
 
     @skip_pre_blackwell
+    @pytest.mark.skip_less_device(2)
     @pytest.mark.parametrize(
         "model_id,model_name,world_size_override,moe_topology", MODEL_PARAMS)
     def test_mxfp4_gsm8k(self, model_id, model_name, world_size_override,