fix(qwen3): adapt serving to latest PyPTO runtime

Remove the obsolete pto_isa_commit CLI and executor plumbing now that PyPTO resolves managed PTO-ISA revisions from its runtime pin.

Validation: pre-commit passed; 12 focused batching tests passed; deterministic 48-token A8W8 generation preserved the expected token IDs at 42.7 ms/token.

Author: vegetabledoww

examples/model/qwen3_14b/npu_generate.py

@@ -392,11 +392,6 @@ def build_parser() -> argparse.ArgumentParser:
     parser.add_argument("--top-k", type=int, default=None)
     parser.add_argument("--stream", action="store_true", default=False)
     parser.add_argument("--save-kernels-dir", default=None)
-    parser.add_argument(
-        "--pto-isa-commit",
-        default=None,
-        help="For qwen3-a8w8 only: pin PyPTO compile/assemble to the installed runtime's pto-isa revision.",
-    )
     parser.add_argument(
         "--num-layers-override",
         type=int,
@@ -488,7 +483,6 @@ def main() -> None:
         "save_kernels_dir": args.save_kernels_dir,
     }
     if args.model_format == _QWEN3_A8W8_FORMAT:
-        executor_kwargs["pto_isa_commit"] = args.pto_isa_commit
         executor_kwargs["l3_trace"] = args.profile_verbose
     executor = executor_cls(
         kv_cache_manager,

pypto_serving/model/common/executor/pypto_executor.py

@@ -42,7 +42,6 @@ def __init__(
         platform: str = "a2a3sim",
         device_ids: Sequence[int] = (0,),
         save_kernels_dir: str | None = None,
-        pto_isa_commit: str | None = None,
     ) -> None:
         """Initialize common PyPTO runtime options and model registries."""
         super().__init__(kv_cache_manager)
@@ -51,7 +50,6 @@ def __init__(
         if not self._device_ids:
             raise ValueError("device_ids must contain at least one device id")
         self._save_kernels_dir = save_kernels_dir
-        self._pto_isa_commit = pto_isa_commit
         self._runners: dict[str, ModelRunner] = {}
         self._compiled: dict[str, object] = {}
 
@@ -127,7 +125,6 @@ def _run_config(self, *, codegen_only: bool):
             codegen_only=codegen_only,
             save_kernels=self._save_kernels_dir is not None,
             save_kernels_dir=self._save_kernels_dir,
-            pto_isa_commit=self._pto_isa_commit,
         )
 
     @abstractmethod

pypto_serving/model/qwen/npu_executor_a8w8.py

@@ -136,7 +136,6 @@ def __init__(
         platform: str = "a2a3sim",
         device_ids: Sequence[int] = (0,),
         save_kernels_dir: str | None = None,
-        pto_isa_commit: str | None = None,
         l3_trace: bool = False,
         pypto_root: str | None = None,
     ) -> None:
@@ -145,7 +144,6 @@ def __init__(
             platform=platform,
             device_ids=device_ids,
             save_kernels_dir=save_kernels_dir,
-            pto_isa_commit=pto_isa_commit,
         )
         self._l3_trace = l3_trace
         self._pypto_root = pypto_root
@@ -460,7 +458,6 @@ def _compile_l3_jit_fwd_callable(
             save_kernels=config.save_kernels,
             save_kernels_dir=config.save_kernels_dir,
             codegen_only=True,
-            pto_isa_commit=config.pto_isa_commit,
             diagnostic_phase=config.diagnostic_phase,
             disabled_diagnostics=config.disabled_diagnostics,
             compile_profiling=config.compile_profiling,