Update base for Update on "[ET Device Support] Schema changes: device info on Tensor and buffer-level device array"

This diff adds device placement information to the ExecuTorch schema to support representing tensor-level device type information, which will be the basic requirement for the following tensor_parser updates.

This is part of the Phase 1 implementation to make ET device type work E2E without user-specified device placement.

Design doc: https://docs.google.com/document/d/1lwd9BlohmwkN5EEvRulO_b-XnZBwv1nMb5l2K3jfuwA/edit?tab=t.0#heading=h.o6anuvkix4bu

Differential Revision: [D93635657](https://our.internmc.facebook.com/intern/diff/D93635657/)

[ghstack-poisoned]

Author: Gasoonjia

.ci/docker/ci_commit_pins/pytorch.txt

@@ -1 +1 @@
-659af3c353e49b35c191cdd2dba3b3c79d0e6822
+release/2.11

.ci/docker/common/install_pytorch.sh

@@ -32,9 +32,9 @@ install_pytorch_and_domains() {
   pip_install "$(echo dist/*.whl)"
 
   # Grab the pinned audio and vision commits from PyTorch
-  TORCHAUDIO_VERSION=$(cat .github/ci_commit_pins/audio.txt)
+  TORCHAUDIO_VERSION=release/2.11
   export TORCHAUDIO_VERSION
-  TORCHVISION_VERSION=$(cat .github/ci_commit_pins/vision.txt)
+  TORCHVISION_VERSION=release/0.26
   export TORCHVISION_VERSION
 
   install_domains

.ci/scripts/export_model_artifact.sh

@@ -334,6 +334,7 @@ if [ "$MODEL_NAME" = "voxtral_realtime" ]; then
     VR_QUANT_ARGS="--qlinear-encoder 8da4w --qlinear 8da4w --qlinear-group-size 32 --qembedding 8w"
   elif [ "$QUANT_NAME" = "quantized-int4-metal" ]; then
     VR_QUANT_ARGS="--qlinear-encoder fpa4w --qlinear fpa4w"
+    VR_DTYPE_ARGS="--dtype bf16"
   elif [ "$QUANT_NAME" = "quantized-int4-tile-packed" ]; then
     VR_QUANT_ARGS="--qlinear-encoder 4w --qlinear-encoder-packing-format tile_packed_to_4d --qlinear 4w --qlinear-packing-format tile_packed_to_4d --qembedding 8w"
     VR_DTYPE_ARGS="--dtype bf16"
@@ -373,6 +374,7 @@ if [ "$MODEL_NAME" = "voxtral_realtime" ]; then
   fi
   # Copy tokenizer from downloaded model weights
   cp "$LOCAL_MODEL_DIR/tekken.json" "${OUTPUT_DIR}/tekken.json"
+  rm -rf "$LOCAL_MODEL_DIR"
   ls -al "${OUTPUT_DIR}"
   echo "::endgroup::"
   exit 0

.ci/scripts/test_lora.sh

@@ -139,7 +139,8 @@ Okay, so I need to calculate 15% of 80."
 EXPECTED_QUANT_LORA_PREFIX="
 <|im_start|>user Calculate 15% of 80?<|im_end|><|im_start|>assistant
 To calculate 15% of 80, we can multiply 80 by 15/100.
-So, 15% of 80 is equal to (80 * 15) / 100 = 1200 / 100 = 12.
+80 * 15/100 = 12.
+So, 15% of 80 is 12.
 #### 12
 The answer is: 12<|im_end|>"
 

.ci/scripts/test_model_e2e.sh

@@ -248,7 +248,7 @@ if [ "$AUDIO_URL" != "" ]; then
 elif [[ "$MODEL_NAME" == *whisper* ]] || [ "$MODEL_NAME" = "voxtral_realtime" ]; then
   conda install -y -c conda-forge "ffmpeg<8"
   pip install datasets soundfile
-  pip install torchcodec==0.11.0.dev20260217 --extra-index-url https://download.pytorch.org/whl/nightly/cpu
+  pip install torchcodec==0.11.0 --extra-index-url https://download.pytorch.org/whl/test/cpu
   python -c "from datasets import load_dataset;import soundfile as sf;sample = load_dataset('distil-whisper/librispeech_long', 'clean', split='validation')[0]['audio'];sf.write('${MODEL_DIR}/$AUDIO_FILE', sample['array'][:sample['sampling_rate']*30], sample['sampling_rate'])"
 fi
 

.ci/scripts/test_wheel_package_qnn.sh

@@ -158,17 +158,17 @@ print(module_vars["TORCH_VERSION"])
 PY
 )
 
-  NIGHTLY_VERSION=$(
-  "$PYBIN" - <<'PY'
-import runpy
-module_vars = runpy.run_path("torch_pin.py")
-print(module_vars["NIGHTLY_VERSION"])
-PY
-)
-  echo "=== [$LABEL] Install torch==${TORCH_VERSION}.${NIGHTLY_VERSION} ==="
-
-  # Install torchao based on the pinned PyTorch version
-  "$PIPBIN" install torch=="${TORCH_VERSION}.${NIGHTLY_VERSION}" --index-url "https://download.pytorch.org/whl/nightly/cpu"
+#   NIGHTLY_VERSION=$(
+#   "$PYBIN" - <<'PY'
+# import runpy
+# module_vars = runpy.run_path("torch_pin.py")
+# print(module_vars["NIGHTLY_VERSION"])
+# PY
+# )
+  echo "=== [$LABEL] Install torch==${TORCH_VERSION} ==="
+
+  # Install torch based on the pinned PyTorch version, preferring the PyTorch test index
+  "$PIPBIN" install torch=="${TORCH_VERSION}" --extra-index-url "https://download.pytorch.org/whl/test"
   "$PIPBIN" install wheel
 
   # Install torchao based on the pinned commit from third-party/ao submodule

.ci/scripts/utils.sh

@@ -53,7 +53,7 @@ dedupe_macos_loader_path_rpaths() {
   pushd ..
   torch_lib_dir=$(python -c "import importlib.util; print(importlib.util.find_spec('torch').submodule_search_locations[0])")/lib
   popd
-  
+
   if [[ -z "${torch_lib_dir}" || ! -d "${torch_lib_dir}" ]]; then
     return
   fi
@@ -141,9 +141,9 @@ install_pytorch_and_domains() {
 
   dedupe_macos_loader_path_rpaths
   # Grab the pinned audio and vision commits from PyTorch
-  TORCHAUDIO_VERSION=$(cat .github/ci_commit_pins/audio.txt)
+  TORCHAUDIO_VERSION=release/2.11
   export TORCHAUDIO_VERSION
-  TORCHVISION_VERSION=$(cat .github/ci_commit_pins/vision.txt)
+  TORCHVISION_VERSION=release/0.26
   export TORCHVISION_VERSION
 
   install_domains

.ci/scripts/wheel/pre_build_script.sh

@@ -26,7 +26,12 @@ fi
 # we should update the core job in test-infra to enable long paths before
 # checkout to avoid needing to do this.
 pushd extension/llm/tokenizers
-git submodule update --init
+UNAME_S=$(uname -s)
+if [[ $UNAME_S == *"MINGW"* || $UNAME_S == *"MSYS"* ]]; then
+  git -c http.sslBackend=openssl submodule update --init
+else
+  git submodule update --init
+fi
 popd
 
 # On Windows, enable symlinks and re-checkout the current revision to create

.github/pytorch-probot.yml

@@ -3,6 +3,8 @@ tracking_issue: 7679
 ciflow_push_tags:
 - ciflow/android
 - ciflow/apple
+- ciflow/cuda
+- ciflow/metal
 - ciflow/nightly
 - ciflow/trunk
 - ciflow/binaries

.github/workflows/add-unanswered-to-project.yml

@@ -50,10 +50,10 @@ jobs:
               "Jerry-Ge", "gggekov", "fumchin", "wwwind", "benkli01", "Tessil", "maddun01", "Michiel-Olieslagers", "armwaheed", "agrima1304", 
               "emmakujala", "annietllnd", "MatthiasHertel80", "AlexTawseArm", "jmahbs", "morgolock", "Christoffer-JL", "ArmRyan", "xingguo01", 
               "tgonzalezorlandoarm", "chizkiyahu", "sarah-blades", "itsMarco-G", "usamahz", "haowhsu-quic", "shewu-quic", "winskuo-quic", 
-              "chunit-quic", "DannyYuyang-quic", "chuntl", "thchenqti", "jethroqti", "chenweng-quic", "cymbalrush", "DenisVieriu97", "billmguo", 
-              "StrycekSimon", "jirioc", "robert-kalmar", "skywall", "MartinPavella", "roman-janik-nxp", "novak-vaclav", "neuropilot-captain", 
-              "dijopaul", "cad-rlc", "cad-audio", "ynimmaga", "daniil-lyakhov", "emmanuel-ferdman", "cavusmustafa", "anzr299", "suryasidd", 
-              "Jiseong-oh", "alexdean08",
+              "chunit-quic", "DannyYuyang-quic", "chuntl", "thchenqti", "jethroqti", "chenweng-quic", "qti-horodnic", "qti-mmadhava", "quic-boyuc",
+              "cymbalrush", "DenisVieriu97", "billmguo", "StrycekSimon", "jirioc", "robert-kalmar", "skywall", "MartinPavella", "roman-janik-nxp", 
+              "novak-vaclav", "neuropilot-captain", "dijopaul", "cad-rlc", "cad-audio", "ynimmaga", "daniil-lyakhov", "emmanuel-ferdman", 
+              "cavusmustafa", "anzr299", "suryasidd", "Jiseong-oh", "alexdean08",
               // explicitly include the dependabot bot login seen in PRs
               "dependabot[bot]"
             ]);