Update base for Update on "Fix Cuda out of memory issue for eager runner"

This PR updates the eager runner to disable grad and save memory usage.

It also update the prompt format to not include bos.

Differential Revision: [D65962743](https://our.internmc.facebook.com/intern/diff/D65962743/)

[ghstack-poisoned]

Author: helunwencser

.ci/scripts/gather_test_models.py

@@ -25,7 +25,7 @@
         "resnet50": "linux.12xlarge",
         "llava": "linux.12xlarge",
         "llama3_2_vision_encoder": "linux.12xlarge",
-        "llama3_2_text_decoder": "linux.12xlarge",
+        # "llama3_2_text_decoder": "linux.12xlarge",  # TODO: re-enable test when Huy's change is in / model gets smaller.
         # This one causes timeout on smaller runner, the root cause is unclear (T161064121)
         "dl3": "linux.12xlarge",
         "emformer_join": "linux.12xlarge",

backends/arm/TARGETS

@@ -9,6 +9,7 @@ python_library(
     typing = True,
     deps = [
         ":arm_backend",
+        "//executorch/backends/arm/operator_support:operator_support",
         "//executorch/backends/arm/_passes:passes",
         "//executorch/exir:lib",
     ],

backends/arm/operator_support/TARGETS

@@ -0,0 +1,12 @@
+load("@fbcode_macros//build_defs:python_library.bzl", "python_library")
+
+python_library(
+    name = "operator_support",
+    srcs = glob(["*.py"]),
+    typing = True,
+    deps = [
+        "//executorch/backends/xnnpack/_passes:xnnpack_passes",
+        "//executorch/exir:lib",
+        "//executorch/backends/arm:tosa_specification"
+    ],
+)

backends/arm/operator_support/tosa_supported_operators.py

@@ -6,6 +6,7 @@
 # pyre-unsafe
 
 import operator
+from typing import Type
 
 import torch.fx as fx
 from executorch.backends.arm.tosa_specification import TosaSpecification
@@ -31,7 +32,9 @@ def is_node_supported(self, node: fx.Node, tosa_spec: TosaSpecification) -> bool
 
 
 # container for all SupportedTosaOperatorCheck classes
-_tosa_spec_dicts: dict[TosaSpecification, dict[str, SupportedTOSAOperatorCheck]] = {
+_tosa_spec_dicts: dict[
+    TosaSpecification, dict[str, Type[SupportedTOSAOperatorCheck]]
+] = {
     TosaSpecification.create_from_string("TOSA-0.80.0+BI"): {},
     TosaSpecification.create_from_string("TOSA-0.80.0+MI"): {},
 }