Merge branch 'main' into tuner-candidate-spec-gen

.github/workflows/ci-llama-large-tests.yaml

@@ -53,11 +53,11 @@ jobs:
         run: |
           source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
@@ -66,6 +66,8 @@ jobs:
             iree-base-runtime \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
+
           pip freeze
 
       - name: Run llama tests

.github/workflows/ci-llama-quick-tests.yaml

@@ -53,11 +53,11 @@ jobs:
         run: |
           source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
@@ -66,6 +66,7 @@ jobs:
             iree-base-runtime \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
           pip freeze
 
       - name: Run llama 8b f16 decomposed test

.github/workflows/ci-sglang-integration-tests.yml

@@ -50,11 +50,11 @@ jobs:
         run: |
           source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -e sharktank/ shortfin/
 
           # Use newest possible releases to be able to track commits that may
           # cause errors.
@@ -64,6 +64,8 @@ jobs:
             iree-turbine \
             "numpy<2.0"
 
+          pip install --no-compile -r requirements.txt -e sharktank/ shortfin/
+
           # Install SGLang and sentence_transformers
           pip install "git+https://github.com/nod-ai/sglang.git#subdirectory=python"
           pip install sentence_transformers

.github/workflows/ci-shark-ai.yml

@@ -49,11 +49,11 @@ jobs:
         run: |
           source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -e sharktank/ shortfin/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
@@ -62,6 +62,8 @@ jobs:
             iree-base-runtime \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -e sharktank/ shortfin/
+
           pip freeze
 
       - name: Run LLM Integration Tests

.github/workflows/ci-sharktank.yml

@@ -63,11 +63,11 @@ jobs:
       - name: Install pip deps
         run: |
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile --pre --index-url https://download.pytorch.org/whl/test/cpu torch==${{matrix.torch-version}}+cpu
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
@@ -76,6 +76,8 @@ jobs:
             iree-base-runtime \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
+
       - name: Run sharktank tests
         if: ${{ !cancelled() }}
         run: |
@@ -116,7 +118,6 @@ jobs:
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
@@ -125,6 +126,8 @@ jobs:
             iree-base-runtime \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
+
       - name: Run tests
       # TODO: unify with-t5-data and with-clip-data flags into a single flag
       # and make it possible to run only tests that require data.
@@ -162,18 +165,21 @@ jobs:
       - name: Install pip deps
         run: |
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
-          pip install -f https://iree.dev/pip-release-links.html --pre --upgrade \
+          pip install -f https://iree.dev/pip-release-links.html --pre \
             iree-base-compiler \
             iree-base-runtime \
             iree-turbine
+
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
+
       - name: Run punet tests
         run: |
           pytest -v sharktank/ -m punet_quick \

.github/workflows/ci_eval.yaml

@@ -51,19 +51,20 @@ jobs:
         run: |
           source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
-          pip install -f https://iree.dev/pip-release-links.html --pre \
+          pip install -f https://iree.dev/pip-release-links.html --pre --upgrade \
             iree-base-compiler \
             iree-base-runtime \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
           pip freeze
 
       - name: Run perplexity test with IREE
@@ -109,17 +110,19 @@ jobs:
         run: |
           source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly iree-turbine.
           # We could also pin to a known working or stable version.
           pip install -f https://iree.dev/pip-release-links.html --pre --upgrade \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
+
       - name: Run perplexity test with Torch
         run: |
           source ${VENV_DIR}/bin/activate

.github/workflows/ci_eval_short.yaml

@@ -50,11 +50,11 @@ jobs:
         run: |
           source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
+
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
           # wheels saves multiple minutes and a lot of bandwidth on runner setup.
           pip install --no-compile -r pytorch-cpu-requirements.txt
-          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
 
           # Install nightly IREE packages.
           # We could also pin to a known working or stable version.
@@ -63,6 +63,7 @@ jobs:
             iree-base-runtime \
             iree-turbine
 
+          pip install --no-compile -r requirements.txt -r sharktank/requirements-tests.txt -e sharktank/
           pip freeze
 
       - name: Run perplexity test with vmfb