CI: Optimize autotuning pipeline docs

.github/scripts/op_tune.sh

@@ -18,11 +18,11 @@ testFailedFiles=()
 declare -a tune_jobs=(
   "ck_batched_gemm_a8w8:csrc/ck_batched_gemm_a8w8:op_tests/test_batched_gemm_a8w8.py:python3 csrc/ck_batched_gemm_a8w8/batched_gemm_a8w8_tune.py -i aiter/configs/a8w8_untuned_batched_gemm.csv -o aiter/configs/a8w8_tuned_batched_gemm.csv"
   "ck_batched_gemm_bf16:csrc/ck_batched_gemm_bf16:op_tests/test_batched_gemm_bf16.py:python3 csrc/ck_batched_gemm_bf16/batched_gemm_bf16_tune.py -i aiter/configs/bf16_untuned_batched_gemm.csv -o aiter/configs/bf16_tuned_batched_gemm.csv"
-#  "csrc/ck_gemm_a4w4_blockscale:op_tests/test_gemm_a4w4_blockscale.py:python3 csrc/ck_gemm_a4w4_blockscale/gemm_a4w4_blockscale_tune.py -i aiter/configs/a4w4_blockscale_untuned_gemm.csv -o aiter/configs/a4w4_blockscale_tuned_gemm.csv"
   "ck_gemm_a8w8:csrc/ck_gemm_a8w8:op_tests/test_gemm_a8w8.py:python3 csrc/ck_gemm_a8w8/gemm_a8w8_tune.py -i aiter/configs/a8w8_untuned_gemm.csv -o aiter/configs/a8w8_tuned_gemm.csv"
   "ck_gemm_a8w8_blockscale:csrc/ck_gemm_a8w8_blockscale:op_tests/test_gemm_a8w8_blockscale.py:python3 csrc/ck_gemm_a8w8_blockscale/gemm_a8w8_blockscale_tune.py -i aiter/configs/a8w8_blockscale_untuned_gemm.csv -o aiter/configs/a8w8_blockscale_tuned_gemm.csv"
   "ck_gemm_a8w8_blockscale_bpreshuffle:csrc/ck_gemm_a8w8_blockscale_bpreshuffle:op_tests/test_gemm_a8w8_blockscale.py:python3 csrc/ck_gemm_a8w8_blockscale_bpreshuffle/gemm_a8w8_blockscale_bpreshuffle_tune.py -i aiter/configs/a8w8_blockscale_bpreshuffle_untuned_gemm.csv -o aiter/configs/a8w8_blockscale_bpreshuffle_tuned_gemm.csv"
   "ck_gemm_a8w8_bpreshuffle:csrc/ck_gemm_a8w8_bpreshuffle:op_tests/test_gemm_a8w8.py:python3 csrc/ck_gemm_a8w8_bpreshuffle/gemm_a8w8_bpreshuffle_tune.py -i aiter/configs/a8w8_bpreshuffle_untuned_gemm.csv -o aiter/configs/a8w8_bpreshuffle_tuned_gemm.csv"
+  #"ck_gemm_a4w4_blockscale:csrc/ck_gemm_a4w4_blockscale:op_tests/test_gemm_a4w4_blockscale.py:python3 csrc/ck_gemm_a4w4_blockscale/gemm_a4w4_blockscale_tune.py -i aiter/configs/a4w4_blockscale_untuned_gemm.csv -o aiter/configs/a4w4_blockscale_tuned_gemm.csv"
 )
 
 for job in "${tune_jobs[@]}"; do

.github/workflows/operators-tuning.yaml

@@ -78,7 +78,7 @@ jobs:
           docker exec \
           -w /workspace \
           operators_tuning_test \
-          ./.github/scripts/op_tune.sh test "ck_batched_gemm_a8w8, ck_gemm_a8w8, ck_gemm_a8w8_blockscale"
+          ./.github/scripts/op_tune.sh test "${{ github.event.inputs.shapes }}"
 
       - name: Operators tuning Tests
         run: |
@@ -87,7 +87,7 @@ jobs:
           docker exec \
           -w /workspace \
           operators_tuning_test \
-          ./.github/scripts/op_tune.sh tune "ck_batched_gemm_a8w8, ck_gemm_a8w8, ck_gemm_a8w8_blockscale" "${{ github.event.inputs.arguments }}"
+          ./.github/scripts/op_tune.sh tune "${{ github.event.inputs.shapes }}" "${{ github.event.inputs.arguments }}"
 
       - name: Show the difference after tuning
         run: |
@@ -100,7 +100,7 @@ jobs:
           docker exec \
           -w /workspace \
           operators_tuning_test \
-          ./.github/scripts/op_tune.sh test "ck_batched_gemm_a8w8, ck_gemm_a8w8, ck_gemm_a8w8_blockscale"
+          ./.github/scripts/op_tune.sh test "${{ github.event.inputs.shapes }}"
 
       - name: Upload tuned CSVs
         uses: actions/upload-artifact@v4

docs/autotuning_pipeline.md

@@ -14,4 +14,22 @@ Based on this, we provide CI pipelines to generate and use these tuned CSV files
 
 - [Manual Pipeline](https://github.com/ROCm/aiter/actions/workflows/operators-tuning.yaml): Allows users to select specific shapes to tune and choose whether to upload the results to the Aiter repository.
 
+    1. Navigate to the Autotuning Pipelines GitHub Actions workflow page: https://github.com/ROCm/aiter/actions/workflows/operators-tuning.yaml
+
+    2. To trigger the workflow, click the `Run workflow` button at the top right corner of the Actions page. By default, this will run the tuning process for all shapes available in the `aiter/configs` directory. If you wish to tune only specific shapes, enter a comma-separated list of shape names in the `List of shape names to run` field, for example: `ck_gemm_a8w8, ck_gemm_a8w8_blockscale, ck_gemm_a8w8_blockscale_bpreshuffle, ck_gemm_a8w8_bpreshuffle`. If additional arguments are needed for the tuning script, you can provide them in the `Additional arguments for the tuning script` field. A full list of supported arguments can be found in the [base_tuner.py script](https://github.com/ROCm/aiter/blob/main/aiter/utility/base_tuner.py#L70).
+
+        ![Aiter Autotuning CI Pipeline - 1](https://raw.githubusercontent.com/ROCm/aiter/main/docs/images/autotuning_ci_pipeline_1.jpeg)
+
+    3. During the workflow execution, the following steps will be performed:
+        - Run performance tests before tuning.
+        - Execute the tuning process for the selected operators.
+        - Display the differences in the CSV files after tuning.
+        - Run performance tests again after tuning to compare results.
+        - Upload the tuned CSV files as GitHub workflow artifacts.
+        - You can download the tuned CSV artifacts and upload them to the Aiter repository as needed.
+
+    4. If you wish to upload your own untuned CSV files, please create a new branch and update the relevant untuned CSV files in the `aiter/configs` directory. Then, trigger the workflow on your branch to proceed with tuning.
+
+        ![Aiter Autotuning CI Pipeline - 2](https://raw.githubusercontent.com/ROCm/aiter/main/docs/images/autotuning_ci_pipeline_2.jpeg)
+
 - Scheduled Pipeline: Runs nightly or weekly to generate all tuned CSV files and automatically upload the results to the Aiter repository.