Merge branch 'main' into dvartanians/fix_yolov4_faster_webdemo

tenstorrent · Dec 12, 2024 · ce06d62 · ce06d62
2 parents 0c884bf + 660d249
commit ce06d62
Show file tree

Hide file tree

Showing 99 changed files with 2,682 additions and 643 deletions.
diff --git a/.clang-tidy b/.clang-tidy
@@ -26,11 +26,9 @@ Checks: >
   -bugprone-unhandled-self-assignment,
   -bugprone-unused-raii,
   -cert-env33-c,
-  -cert-err09-cpp,
   -cert-err33-c,
   -cert-err34-c,
   -cert-err58-cpp,
-  -cert-err61-cpp,
   -cert-flp30-c,
   -cert-msc30-c,
   -cert-msc32-c,
@@ -121,7 +119,6 @@ Checks: >
   -misc-no-recursion,
   -misc-non-private-member-variables-in-classes,
   -misc-redundant-expression,
-  -misc-throw-by-value-catch-by-reference,
   -misc-unconventional-assign-operator,
   -misc-uniqueptr-reset-release,
   -misc-unused-parameters,

diff --git a/.github/actions/prepare-metal-run/action.yml b/.github/actions/prepare-metal-run/action.yml
@@ -9,6 +9,10 @@ inputs:
     description: "Whether to load with profiler"
     required: false
     default: 'false'
+  python-version:
+    description: 'Which version of Python to use to run the tests.'
+    required: false
+    default: '3.8'
 
 runs:
   using: "composite"
@@ -25,6 +29,8 @@ runs:
       shell: bash
       run: tar -xvf ttm_${{ inputs.arch }}.tar
     - uses: ./.github/actions/install-python-deps
+      with:
+        python-version: ${{ inputs.python-version }}
     - name: Collect Workflow Telemetry
       if: ${{ !cancelled() }}
       uses: catchpoint/workflow-telemetry-action@v2
diff --git a/.github/workflows/_build-wheels-impl.yaml b/.github/workflows/_build-wheels-impl.yaml
@@ -31,6 +31,17 @@ jobs:
           os: ${{ inputs.os }}
       - name: Clean up dirty files
         run: git clean -f -d
+      - name: Set Python Version
+        id: python-version
+        run: |
+          if [[ "${{ inputs.os }}" == "ubuntu-20.04" ]]; then
+            echo "python-version=3.8" >> $GITHUB_ENV
+          elif [[ "${{ inputs.os }}" == "ubuntu-22.04" ]]; then
+            echo "python-version=3.10" >> $GITHUB_ENV
+          else
+            echo "Unsupported OS version: ${{ inputs.os }}"
+            exit 1
+          fi
       - uses: actions/[email protected]
         with:
           cache: 'pip'
@@ -47,6 +58,7 @@ jobs:
         if: ${{ inputs.from-precompiled }}
         with:
           arch: ${{ inputs.arch }}
+          python-version: ${{ env.python-version }}
       - name: Set precompiled dir for precompile builds
         if: ${{ inputs.from-precompiled }}
         # TT_FROM_PRECOMPILED_DIR env variable allows us to not re-run the full C++ build and instead

diff --git a/.github/workflows/blackhole-post-commit.yaml b/.github/workflows/blackhole-post-commit.yaml
@@ -29,14 +29,14 @@ jobs:
     uses: ./.github/workflows/build-artifact.yaml
     secrets: inherit
     with:
-      os: "ubuntu-20.04-amd64"
+      os: "ubuntu-22.04-amd64"
       arch: '["blackhole"]'
       build-docker: false
   build-wheels:
     needs: build-artifact
     uses: ./.github/workflows/_build-wheels-impl.yaml
     with:
-      os: "ubuntu-20.04"
+      os: "ubuntu-22.04"
       arch: "blackhole"
       from-precompiled: true
 #   build-artifact-profiler:
@@ -58,15 +58,15 @@ jobs:
       arch: blackhole
       runner-label: BH
       timeout: 30
-      os: "ubuntu-20.04"
+      os: "ubuntu-22.04"
   fd-unit-tests:
     needs: build-wheels
     uses: ./.github/workflows/fast-dispatch-build-and-unit-tests.yaml
     secrets: inherit
     with:
       arch: blackhole
       runner-label: BH
-      os: "ubuntu-20.04"
+      os: "ubuntu-22.04"
   # FD C++ Unit Tests
   cpp-unit-tests:
     needs: build-artifact
@@ -76,7 +76,7 @@ jobs:
       arch: blackhole
       runner-label: BH
       timeout: 60
-      os: "ubuntu-20.04"
+      os: "ubuntu-22.04"
 
 #   profiler-regression:
 #     needs: build-artifact-profiler

diff --git a/.github/workflows/build-and-unit-tests.yaml b/.github/workflows/build-and-unit-tests.yaml
@@ -70,7 +70,8 @@ jobs:
             -e TT_METAL_SLOW_DISPATCH_MODE=1
             -e LD_LIBRARY_PATH=${{ github.workspace }}/build/lib
           run_args: |
-            python3 -m pip install -r $(pwd)/tt_metal/python_env/requirements-dev.txt
+            pip install --force-reinstall pip==21.2.4
+            pip install -r tt_metal/python_env/requirements-dev.txt
             pip install -e .
             ./tests/scripts/run_tests.sh --tt-arch ${{ inputs.arch }} --pipeline-type post_commit --dispatch-mode slow
       - uses: ./.github/actions/slack-report

diff --git a/.github/workflows/cpp-post-commit.yaml b/.github/workflows/cpp-post-commit.yaml
@@ -84,8 +84,9 @@ jobs:
             -e ARCH_NAME=${{ inputs.arch }}
             -e LD_LIBRARY_PATH=${{ github.workspace }}/build/lib
           run_args: |
-            python3 -m pip install -r $(pwd)/tt_metal/python_env/requirements-dev.txt
-            python3 -m pip install -e .
+            pip install --force-reinstall pip==21.2.4
+            pip install -r tt_metal/python_env/requirements-dev.txt
+            pip install -e .
             ${{ matrix.test-group.cmd }}
       - uses: ./.github/actions/slack-report
         if: ${{ failure() }}

diff --git a/.github/workflows/perf-device-models-impl.yaml b/.github/workflows/perf-device-models-impl.yaml
@@ -12,7 +12,7 @@ jobs:
       matrix:
         test-info: [
           {name: "GS", arch: grayskull, runs-on: ["perf-no-reset-grayskull", "bare-metal", "in-service"], machine-type: "bare_metal", timeout: 40},
-          {name: "N300 WH B0", arch: wormhole_b0, runs-on: ["N300", "pipeline-perf", "bare-metal", "in-service"], machine-type: "bare_metal", timeout: 30},
+          {name: "N300 WH B0", arch: wormhole_b0, runs-on: ["N300", "pipeline-perf", "bare-metal", "in-service"], machine-type: "bare_metal", timeout: 40},
         ]
     name: "${{ matrix.test-info.name }} device perf"
     env:

diff --git a/docs/source/tt-metalium/tools/kernel_print.rst b/docs/source/tt-metalium/tools/kernel_print.rst
@@ -21,12 +21,13 @@ Note that the core coordinates are logical coordinates, so worker cores and ethe
 
 .. code-block::
 
-    export TT_METAL_DPRINT_CORES=0,0           # required, x,y OR (x1,y1),(x2,y2),(x3,y3) OR (x1,y1)-(x2,y2) OR all OR worker OR dispatch
-    export TT_METAL_DPRINT_ETH_CORES=0,0       # optional, x,y OR (x1,y1),(x2,y2),(x3,y3) OR (x1,y1)-(x2,y2) OR all OR worker OR dispatch
-    export TT_METAL_DPRINT_CHIPS=0             # optional, comma separated list of chips
-    export TT_METAL_DPRINT_RISCVS=BR           # optional, default is all RISCs.  Use a subset of BR,NC,TR0,TR1,TR2
-    export TT_METAL_DPRINT_FILE=log.txt        # optional, default is to print to the screen
-    export TT_METAL_DPRINT_ONE_FILE_PER_RISC=1 # optional, splits DPRINT data on a per-RISC basis into files under $TT_METAL_HOME/generated/dprint/. Overrides TT_METAL_DPRINT_FILE.
+    export TT_METAL_DPRINT_CORES=0,0                    # required, x,y OR (x1,y1),(x2,y2),(x3,y3) OR (x1,y1)-(x2,y2) OR all OR worker OR dispatch
+    export TT_METAL_DPRINT_ETH_CORES=0,0                # optional, x,y OR (x1,y1),(x2,y2),(x3,y3) OR (x1,y1)-(x2,y2) OR all OR worker OR dispatch
+    export TT_METAL_DPRINT_CHIPS=0                      # optional, comma separated list of chips
+    export TT_METAL_DPRINT_RISCVS=BR                    # optional, default is all RISCs. Use a subset of BR,NC,TR0,TR1,TR2
+    export TT_METAL_DPRINT_FILE=log.txt                 # optional, default is to print to the screen
+    export TT_METAL_DPRINT_PREPEND_DEVICE_CORE_RISC=0   # optional, enabled by default. Prepends prints with <device id>:(<core x>, <core y>):<RISC>:.
+    export TT_METAL_DPRINT_ONE_FILE_PER_RISC=1          # optional, splits DPRINT data on a per-RISC basis into files under $TT_METAL_HOME/generated/dprint/. Overrides TT_METAL_DPRINT_FILE and disables TT_METAL_DPRINT_PREPEND_DEVICE_CORE_RISC.
 
 To generate kernel debug prints on the device, include the ``debug/dprint.h`` header and use the APIs defined there.
 An example with the different features available is shown below:
@@ -122,4 +123,4 @@ formats for printing from CBs are ``DataFormat::Float32``, ``DataFormat::Float16
     }
 
 .. note::
-    Note that the DPRINT buffer for a RISC is flushed when ``ENDL()`` is called, a ``\n`` character is read, or the device that the RISC belongs to is closed.
+    The DPRINT buffer for a RISC is only flushed when ``ENDL()`` is called, a ``\n`` character is read, or the device that the RISC belongs to is closed.
diff --git a/docs/source/ttnn/ttnn/get_started.rst b/docs/source/ttnn/ttnn/get_started.rst
@@ -26,7 +26,7 @@ Get started with the Falcon 7B demo. Navigate to the `Falcon 7B demo folder
 for details.
 
 You can also check our demos for
-`ResNet <https://github.com/tenstorrent/tt-metal/tree/main/models/demos/resnet>`_,
+`ResNet <https://github.com/tenstorrent/tt-metal/tree/main/models/demos/ttnn_resnet>`_,
 `BERT <https://github.com/tenstorrent/tt-metal/tree/main/models/demos/metal_BERT_large_11>`_,
 `Mistral 7B <https://github.com/tenstorrent/tt-metal/tree/main/models/demos/wormhole/mistral7b>`_,
 and

diff --git a/models/demos/wormhole/stable_diffusion/README.md b/models/demos/wormhole/stable_diffusion/README.md
@@ -19,22 +19,36 @@ Inputs by default are provided from `input_data.json`. If you wish to change the
 >
 > If you are using Wormhole, you must set the `WH_ARCH_YAML` environment variable.
 >
-> ```
+> ```sh
 > export WH_ARCH_YAML=wormhole_b0_80_arch_eth_dispatch.yaml
 > ```
 
 To run the demo, make sure to build the project, activate the environment, and set the appropriate environment variables.
 For more information, refer [installation and build guide](https://github.com/tenstorrent/tt-metal/blob/main/INSTALLING.md).
 
-Use `pytest --disable-warnings --input-path="models/demos/wormhole/stable_diffusion/demo/input_data.json" models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo` to run the demo.
-
-If you wish to run the demo with a different input use `pytest --disable-warnings --input-path="<address_to_your_json_file.json>" models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo`
-
-If you would like to run an interactive demo which will prompt you for the input, use `pytest models/demos/wormhole/stable_diffusion/demo/demo.py::test_interactive_demo`
-
-Our second demo is designed to run poloclub/diffusiondb dataset, run this with `pytest --disable-warnings models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo_diffusiondb`.
-
-If you wish to run for `num_prompts` samples and `num_inference_steps` denoising steps, use `pytest --disable-warnings models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo_diffusiondb[<num_prompts>-<num_inference_steps>]`
+```sh
+pytest --disable-warnings --input-path="models/demos/wormhole/stable_diffusion/demo/input_data.json" models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo
+```
+
+If you wish to run the demo with a different input:
+```sh
+pytest --disable-warnings --input-path="<address_to_your_json_file.json>" models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo
+
+```
+If you would like to run an interactive demo which will prompt you for the input:
+```sh
+pytest models/demos/wormhole/stable_diffusion/demo/demo.py::test_interactive_demo
+```
+
+Our second demo is designed to run poloclub/diffusiondb dataset, run this with:
+```sh
+pytest --disable-warnings models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo_diffusiondb
+```
+
+If you wish to run for `num_prompts` samples and `num_inference_steps` denoising steps:
+```sh
+pytest --disable-warnings models/demos/wormhole/stable_diffusion/demo/demo.py::test_demo_diffusiondb[<num_prompts>-<num_inference_steps>]
+```
 
 Note: ttnn stable diffusion utilizes `PNDMScheduler` and requires `num_inference_steps to be greater than or equal to 4`. [Reference](https://arxiv.org/pdf/2202.09778)