feat: Add GPT-OSS support via mcore

.gitmodules

@@ -1,12 +1,12 @@
 [submodule "3rdparty/Megatron-LM"]
 	path = 3rdparty/Megatron-LM-workspace/Megatron-LM
 	url = https://github.com/terrykong/Megatron-LM.git
-	branch = yuya/nemo-rl-use-dev
+	branch = ashors/dev-with-gpt-oss
 	shallow = true
 [submodule "3rdparty/Megatron-Bridge"]
 	path = 3rdparty/Megatron-Bridge-workspace/Megatron-Bridge
 	url = https://github.com/NVIDIA-NeMo/Megatron-Bridge.git
-	branch = main
+	branch = ashors/gpt-oss-tot
 	shallow = true
 [submodule "3rdparty/Automodel-workspace/Automodel"]
 	path = 3rdparty/Automodel-workspace/Automodel

nemo_rl/models/generation/vllm/vllm_worker.py

@@ -132,6 +132,12 @@ def __init__(
         self.cfg = config
 
         self.model_name = self.cfg["model_name"]
+        ## use the bf16 version of the model rather than the quantized version
+        ## megatron --> hf export is done in bf16 so this ensures the vllm
+        ## model is compatible with megatron
+        if "openai/gpt-oss" in self.model_name:
+            size = self.model_name.split("-")[-1]
+            self.model_name = f"unsloth/gpt-oss-{size}-BF16"
         self.tensor_parallel_size = self.cfg["vllm_cfg"]["tensor_parallel_size"]
         self.pipeline_parallel_size = self.cfg["vllm_cfg"]["pipeline_parallel_size"]
         self.expert_parallel_size = self.cfg["vllm_cfg"]["expert_parallel_size"]