fix: update ray executor import for vLLM 0.20

vLLM 0.20 renamed the executor module from ray_distributed_executor to
ray_executor; the class name RayDistributedExecutor and its
collective_rpc method are unchanged. Without this fix, any FP8
generation path that goes through monkey_patch_vllm_ray_executor() with
TP*PP > 1 (i.e. any multi-GPU vllm_cfg.precision='fp8' run) fails at
worker init with:

  ModuleNotFoundError: No module named 'vllm.v1.executor.ray_distributed_executor'

Signed-off-by: Aurelien Chartier <2567591+achartier@users.noreply.github.com>

Author: achartier

nemo_rl/models/generation/vllm/quantization/fp8.py

@@ -82,7 +82,7 @@ def monkey_patch_vllm_ray_executor(fp8_config):
     if fp8_config.model_parallel_size > 1:
         # we patch vllm's collective_rpc so that before vllm initalizes the model on each rank, we execute
         # a ray remote that patches each worker with the required fp8 vllm patches
-        from vllm.v1.executor.ray_distributed_executor import RayDistributedExecutor
+        from vllm.v1.executor.ray_executor import RayDistributedExecutor
 
         original_run_workers = RayDistributedExecutor.collective_rpc