wip clean up

Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com>

Author: nrghosh

doc/source/serve/doc_code/cross_node_parallelism_example.py

@@ -124,7 +124,7 @@
         max_model_len=8192,
     ),
     placement_group_config=dict(
-        bundles=[{"GPU": 1, "CPU": 2}] * 2,
+        bundles=[{"GPU": 1}] * 2,
         strategy="PACK",
     ),
 )
@@ -158,7 +158,7 @@
         max_model_len=8192,
     ),
     placement_group_config=dict(
-        bundles=[{"GPU": 1, "CPU": 2}] * 4,
+        bundles=[{"GPU": 1}] * 4,
         strategy="SPREAD",
     ),
 )
@@ -192,7 +192,7 @@
         max_model_len=8192,
     ),
     placement_group_config=dict(
-        bundles=[{"GPU": 1, "CPU": 2}] * 2,
+        bundles=[{"GPU": 1}] * 2,
         strategy="STRICT_PACK",
     ),
 )
@@ -201,61 +201,3 @@
 app = build_openai_app({"llm_configs": [llm_config]})
 serve.run(app, blocking=True)
 # __custom_placement_group_strict_pack_example_end__
-
-# __yaml_cross_node_tp_pp_example_start__
-# config.yaml
-# applications:
-# - args:
-#     llm_configs:
-#       - model_loading_config:
-#           model_id: llama-3.1-8b
-#           model_source: meta-llama/Llama-3.1-8B-Instruct
-#         accelerator_type: L4
-#         deployment_config:
-#           autoscaling_config:
-#             min_replicas: 1
-#             max_replicas: 1
-#         engine_kwargs:
-#           tensor_parallel_size: 2
-#           pipeline_parallel_size: 2
-#           distributed_executor_backend: ray
-#           max_model_len: 8192
-#           enable_chunked_prefill: true
-#           max_num_batched_tokens: 4096
-#   import_path: ray.serve.llm:build_openai_app
-#   name: llm_app
-#   route_prefix: "/"
-# __yaml_cross_node_tp_pp_example_end__
-
-# __yaml_custom_placement_group_example_start__
-# config.yaml
-# applications:
-# - args:
-#     llm_configs:
-#       - model_loading_config:
-#           model_id: llama-3.1-8b
-#           model_source: meta-llama/Llama-3.1-8B-Instruct
-#         accelerator_type: L4
-#         deployment_config:
-#           autoscaling_config:
-#             min_replicas: 1
-#             max_replicas: 1
-#         engine_kwargs:
-#           tensor_parallel_size: 4
-#           distributed_executor_backend: ray
-#           max_model_len: 8192
-#         placement_group_config:
-#           bundles:
-#             - GPU: 1
-#               CPU: 2
-#             - GPU: 1
-#               CPU: 2
-#             - GPU: 1
-#               CPU: 2
-#             - GPU: 1
-#               CPU: 2
-#           strategy: SPREAD
-#   import_path: ray.serve.llm:build_openai_app
-#   name: llm_app
-#   route_prefix: "/"
-# __yaml_custom_placement_group_example_end__

doc/source/serve/llm/index.md

@@ -11,8 +11,6 @@ Ray Serve LLM APIs allow users to deploy multiple LLM models together with a fam
 - 🔌 OpenAI compatible
 - 🔄 Multi-LoRA support with shared base models
 - 🚀 Engine agnostic architecture (i.e. vLLM, SGLang, etc)
-- 🔗 Cross-node tensor and pipeline parallelism
-- ⚙️ Custom :ref:`placement group strategies <pgroup-strategy>` for fine-grained resource control
 
 ## Requirements
 
@@ -50,10 +48,9 @@ The LLMConfig class specifies model details such as:
 
 - Model loading sources (HuggingFace or cloud storage)
 - Hardware requirements (accelerator type)
-- Engine arguments (e.g. vLLM engine kwargs, tensor/pipeline parallelism)
+- Engine arguments (e.g. vLLM engine kwargs)
 - LoRA multiplexing configuration
 - Serve auto-scaling parameters
-- Placement group configuration for multi-node deployments
 
 ```{toctree}
 :hidden: