OVMS requiring manual WEIGHTS_PATH

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. podman run --interactive --tty --publish=1235:8080 --volume=/home/davi/AI/models/:/models/ --device=/dev/dri/renderD128 --user=root docker.io/openvino/model_server:2026.4.0-gpu --rest_port 8080 --model_repository_path /models/ --source_model OpenVINO/Qwen3.6-35B-A3B-int4-ov --target_device GPU --task text_generation --kv_cache_precision u8 --plugin_config '{"OFFLOAD_RATIO": 1}' --max_num_batched_tokens 4000
2. See error

Expected behavior
WEIGHTS_PATH should not be required to be manually added

Logs

davi@davi:~$ podman run --interactive --tty --publish=1235:8080 --volume=/home/davi/AI/models/:/models/ --device=/dev/dri/renderD128 --user=root docker.io/openvino/model_server:2026.4.0-gpu --rest_port 8080 --model_repository_path /models/ --source_model OpenVINO/Qwen3.6-35B-A3B-int4-ov --target_device GPU --task text_generation --kv_cache_precision u8 --plugin_config '{"OFFLOAD_RATIO": 1}' --max_num_batched_tokens 4000
[2026-09-18 02:41:48.761][1][serving][info][server.cpp:115] OpenVINO Model Server 2026.4.0.869b2186a
[2026-09-18 02:41:48.761][1][serving][info][server.cpp:116] OpenVINO backend 2026.4.0-22959-99c81491cc3-releases/2026/4
[2026-09-18 02:41:48.761][1][serving][info][server.cpp:119] OpenVINO GenAI backend 2026.4.0.0-3407-7ea2546852a
[2026-09-18 02:41:48.761][1][serving][info][pythoninterpretermodule.cpp:37] PythonInterpreterModule starting
[2026-09-18 02:41:48.761][1][serving][info][pythoninterpretermodule.cpp:40] Initializing python interpreter
Python version:
3.12.3 (main, Aug 31 2026, 10:18:26) [GCC 13.3.0]
Python sys.path output:
['', '/ovms/lib/python', '/ovms/python_deps', '/lib/python312.zip', '/lib/python3.12', '/lib/python3.12/lib-dynload']
[2026-09-18 02:41:48.843][1][serving][info][pythoninterpretermodule.cpp:59] PythonInterpreterModule started
[2026-09-18 02:41:48.844][8][modelmanager][info][servable_loading_queue.cpp:89] Started servable loading queue thread
[2026-09-18 02:41:48.947][1][modelmanager][info][modelmanager.cpp:252] Available devices for Open VINO: CPU, GPU
[2026-09-18 02:41:48.947][1][serving][info][modelmanager.cpp:243] Running inside Docker container
[2026-09-18 02:41:48.947][1][serving][info][modelmanager.cpp:244] cpu quota: 0, cpu affinity: 16, max_open_files: 524288
[2026-09-18 02:41:48.947][1][serving][info][capimodule.cpp:40] C-APIModule starting
[2026-09-18 02:41:48.947][1][serving][info][capimodule.cpp:42] C-APIModule started
[2026-09-18 02:41:48.947][1][serving][info][grpcservermodule.cpp:106] GRPCServerModule starting
[2026-09-18 02:41:48.947][1][serving][info][grpcservermodule.cpp:110] GRPCServerModule started
[2026-09-18 02:41:48.947][1][serving][info][grpcservermodule.cpp:111] Port was not set. GRPC server will not be started.
[2026-09-18 02:41:48.947][1][serving][info][httpservermodule.cpp:35] HTTPServerModule starting
[2026-09-18 02:41:48.947][1][serving][info][httpservermodule.cpp:39] Will start 16 REST workers
[2026-09-18 02:41:48.948][10][serving][info][drogon_http_server.cpp:157] Binding REST server to address: 0.0.0.0:8080
[2026-09-18 02:41:48.998][1][serving][info][drogon_http_server.cpp:184] REST server listening on port 8080 with 16 unary threads and 16 streaming threads
[2026-09-18 02:41:48.998][1][serving][info][http_server.cpp:242] API key not provided via --api_key_file or API_KEY environment variable. Authentication will be disabled.
[2026-09-18 02:41:48.998][1][serving][info][httpservermodule.cpp:52] HTTPServerModule started
[2026-09-18 02:41:48.998][1][serving][info][httpservermodule.cpp:53] Started REST server at 0.0.0.0:8080
[INFO] LFS resume: attempts=5 interval=10 s
[2026-09-18 02:41:49.049][1][serving][info][libgit2.cpp:1342] Path "/models/OpenVINO/Qwen3.6-35B-A3B-int4-ov" exists but is not a git repository. Skipping download and using existing files.
Model: OpenVINO/Qwen3.6-35B-A3B-int4-ov downloaded to: /models/OpenVINO/Qwen3.6-35B-A3B-int4-ov
Graph: graph.pbtxt created in: /models/OpenVINO/Qwen3.6-35B-A3B-int4-ov
[2026-09-18 02:41:49.051][1][serving][info][servablemanagermodule.cpp:51] ServableManagerModule starting
[2026-09-18 02:41:49.052][8][serving][info][mediapipegraphdefinition.cpp:643] MediapipeGraphDefinition initializing graph nodes
[2026-09-18 02:41:49.052][8][modelmanager][info][servable_initializer.cpp:688] Initializing Visual Language Model Continuous Batching servable
[2026-09-18 02:41:56.687][8][serving][error][servable_initializer.cpp:284] Error during llm node initialization for models_path: /models/OpenVINO/Qwen3.6-35B-A3B-int4-ov/./ exception: Exception from src/inference/src/cpp/core.cpp:117:
Exception from src/inference/src/dev/plugin.cpp:54:
Check '!weights_path.empty()' failed at src/plugins/intel_gpu/src/plugin/ops/moe.cpp:55:
ov::weights_path property is not set. OTD requires a valid path to the model .bin file. Please set ov::weights_path when compiling the model.



[2026-09-18 02:41:56.687][8][modelmanager][error][servable_initializer.cpp:693] Error during LLM node resources initialization: The LLM Node resource initialization failed
[2026-09-18 02:41:56.687][8][serving][error][llm_node_initializer.cpp:60] Failed to process LLM node graph OpenVINO/Qwen3.6-35B-A3B-int4-ov
[2026-09-18 02:41:56.687][8][modelmanager][info][pipelinedefinitionstatus.hpp:60] Mediapipe: OpenVINO/Qwen3.6-35B-A3B-int4-ov state changed to: LOADING_PRECONDITION_FAILED after handling: ValidationFailedEvent: 
[2026-09-18 02:41:56.687][1][modelmanager][error][modelmanager.cpp:286] Couldn't start model manager
[2026-09-18 02:41:56.687][1][serving][error][servablemanagermodule.cpp:58] ovms::ModelManager::Start() Error: The LLM Node resource initialization failed
[2026-09-18 02:41:56.687][1][serving][info][grpcservermodule.cpp:195] GRPCServerModule shutting down
[2026-09-18 02:41:56.687][1][serving][info][grpcservermodule.cpp:205] GRPCServerModule shutdown
[2026-09-18 02:41:56.687][1][serving][info][httpservermodule.cpp:59] HTTPServerModule shutting down
[2026-09-18 02:41:56.688][1][serving][info][httpservermodule.cpp:64] Shutdown HTTP server
[2026-09-18 02:41:56.688][1][serving][info][servablemanagermodule.cpp:65] ServableManagerModule shutting down
[2026-09-18 02:41:56.688][8][modelmanager][info][servable_loading_queue.cpp:113] Stopped servable loading queue thread
[2026-09-18 02:41:56.693][1][serving][info][servablemanagermodule.cpp:71] ServableManagerModule shutdown
[2026-09-18 02:41:56.693][1][serving][info][pythoninterpretermodule.cpp:73] PythonInterpreterModule shutting down
[2026-09-18 02:41:56.693][1][serving][info][pythoninterpretermodule.cpp:77] PythonInterpreterModule shutdown
[2026-09-18 02:41:56.701][1][serving][info][capimodule.cpp:50] C-APIModule shutting down
[2026-09-18 02:41:56.701][1][serving][info][capimodule.cpp:52] C-APIModule shutdown

Configuration
Ryzen 7 5700X3D, B580

Reproduce the failure with the Podman command in the issue, then trace initialization through servable_initializer.cpp and llm_node_initializer.cpp to the weights_path check reported in intel_gpu/src/plugin/ops/moe.cpp. Determine why this model reaches compilation without a weights path; done means the same command loads the model without manually adding WEIGHTS_PATH and the relevant initialization test or reproduction succeeds.