[Model Loader] Apply EP weight filtering in InstantTensor (#61167)

Signed-off-by: Itay Etelis <92247226+Etelis@users.noreply.github.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: Misha Goin <mgoin64@gmail.com>

requirements/cuda.txt

@@ -18,7 +18,7 @@ nvidia-cudnn-frontend>=1.19.1
 nvtx==0.2.15
 # Required for faster safetensors model loading
 fastsafetensors >= 0.3.3
-instanttensor >= 0.1.9
+instanttensor >= 0.2.2
 
 # QuACK and Cutlass DSL for FA4 (cute-DSL implementation)
 nvidia-cutlass-dsl[cu13]==4.8.0

requirements/test/cpu.txt

@@ -406,7 +406,7 @@ inflect==5.6.2
     # via datamodel-code-generator
 iniconfig==2.0.0
     # via pytest
-instanttensor==0.1.9
+instanttensor==0.2.2
     # via -r requirements/test/cuda.in
 interegular==0.3.3
     # via lm-format-enforcer

requirements/test/cuda.in

@@ -57,7 +57,7 @@ arctic-inference == 0.1.1; platform_machine == "x86_64" # Required for suffix de
 numba == 0.65.0 # Required for N-gram speculative decoding
 runai-model-streamer[s3,gcs,azure]==0.15.7
 fastsafetensors>=0.3.3
-instanttensor>=0.1.9; platform_machine == "x86_64"
+instanttensor>=0.2.2; platform_machine == "x86_64"
 decord==0.6.0; platform_machine == "x86_64"
 # terratorch is temporarily disabled while PyPI has the `lightning` package
 # in `quarantined` status (every published terratorch version transitively

requirements/test/cuda.txt

@@ -426,7 +426,7 @@ inflect==5.6.2
     # via datamodel-code-generator
 iniconfig==2.0.0
     # via pytest
-instanttensor==0.1.9
+instanttensor==0.2.2
     # via
     #   -c requirements/cuda.txt
     #   -r requirements/test/cuda.in

requirements/test/nightly-torch.txt

@@ -42,5 +42,5 @@ numba == 0.65.0 # Required for N-gram speculative decoding
 numpy
 runai-model-streamer[s3,gcs,azure]==0.15.7
 fastsafetensors>=0.3.3
-instanttensor>=0.1.9
+instanttensor>=0.2.2
 pydantic>=2.12 # 2.11 leads to error on python 3.13