Merge branch 'master' into kv_multiheadsize

Author: clee30

.github/workflows/windows_conditional_compilation.yml

@@ -58,8 +58,7 @@ jobs:
     env:
       CMAKE_BUILD_TYPE: 'Release'
       CMAKE_COMPILE_WARNING_AS_ERROR: 'ON'
-      # CMAKE_GENERATOR: 'Ninja Multi-Config'
-      CMAKE_GENERATOR: 'Visual Studio 17 2022'
+      CMAKE_GENERATOR: 'Ninja Multi-Config' # Visual Studio does not seem to work with `ccache`. Ninja is the generator supported by `ccache`
       CMAKE_CXX_COMPILER_LAUNCHER: ccache
       CMAKE_C_COMPILER_LAUNCHER: ccache
       CCACHE_REMOTE_DIR: "C:\\mount\\caches\\ccache\\windows2022_x86_64_itt\\${{ github.base_ref || github.ref_name }}"

docs/articles_en/about-openvino/performance-benchmarks.rst

@@ -26,16 +26,6 @@ implemented in your solutions. Click the buttons below to see the chosen benchma
 .. grid:: 1 1 2 2
    :gutter: 4
 
-   .. grid-item::
-
-      .. button-link:: #
-         :class: ov-toolkit-benchmark-results
-         :color: primary
-         :outline:
-         :expand:
-
-         :material-regular:`bar_chart;1.4em` OpenVINO Benchmark Graphs (general)
-
    .. grid-item::
 
       .. button-link:: #
@@ -66,7 +56,15 @@ implemented in your solutions. Click the buttons below to see the chosen benchma
 
          :material-regular:`bar_chart;1.4em` OVMS for GenAI
 
+   .. grid-item::
 
+      .. button-link:: #
+         :class: ov-toolkit-benchmark-results
+         :color: primary
+         :outline:
+         :expand:
+
+         :material-regular:`bar_chart;1.4em` OpenVINO Benchmark Graphs (general)
 
 
 
@@ -153,11 +151,9 @@ For a listing of all platforms and configurations used for testing, refer to the
 
 **Disclaimers**
 
-* Intel® Distribution of OpenVINO™ toolkit performance results are based on release
-  2025.0, as of February 13, 2025.
+* System configurations used for Intel® Distribution of OpenVINO™ toolkit performance results are based on release 2025.1, as of April 9th, 2025.
 
-* OpenVINO Model Server performance results are based on release
-  2025.0, as of February 13, 2025.
+* OpenVINO Model Server performance results are based on release 2025.0, as of February 13, 2025.
 
 The results may not reflect all publicly available updates. Intel technologies' features and
 benefits depend on system configuration and may require enabled hardware, software, or service

docs/articles_en/about-openvino/performance-benchmarks/model-accuracy-int8-fp32.rst

@@ -27,11 +27,11 @@ the table for more information.
    * - bert-base-cased
      - SST-2_bert_cased_padded
      - spearman@cosine
-     - 2.41%
-     - 2.78%
-     - 2.61%
-     - 2.84%
-   * - mask_rcnn_resnet50_atrous_coco
+     - 2.57%
+     - 2.65%
+     - 2.54%
+     - 2.89%
+   * - Detectron-V2
      - COCO2017_detection_91cl_bkgr
      - coco_orig_precision
      - 
@@ -41,31 +41,31 @@ the table for more information.
    * - mobilenet-v2
      - ImageNet2012
      - accuracy @ top1
+     - -0.93%
+     - -0.91%
      - -1.03%
-     - -1.00%
-     - -1.03%
-     - -1.01%
+     - -0.95%
    * - resnet-50
      - ImageNet2012
      - accuracy @ top1
-     - -0.17%
-     - -0.17%
-     - -0.18%
-     - -0.17%
+     - -0.12%
+     - -0.12%
+     - -0.15%
+     - -0.15%
    * - ssd-resnet34-1200
      - COCO2017_detection_80cl_bkgr
      - map
-     - -0.01%
-     - -0.01%
-     - -0.04%
-     - -0.04%
-   * - yolo_v8n
+     - 0.00%
+     - 0.00%
+     - -0.03%
+     - 0.07%
+   * - yolo_v11
      - COCO2017_detection_80cl
      - map
-     - -0.09%
-     - -0.09%
-     - -0.02%
-     - -0.04%
+     - 
+     - 
+     - 
+     - 
 .. list-table:: Model Accuracy for BF16, FP32 and FP16 (FP16: Arc only. BF16: Xeon® 6972P only)
    :header-rows: 1
 
@@ -75,56 +75,49 @@ the table for more information.
      - A, FP32
      - B, FP32
      - C, FP32
-     - C, BF16
      - D, FP16
    * - bert-base-cased
      - SST-2_bert_cased_padded
      - spearman@cosine
      - 0.00%
      - 0.00%
      - 0.00%
-     - -0.01%
      - 0.02%
-   * - mask_rcnn_resnet50_atrous_coco
+   * - Detectron-V2
      - COCO2017_detection_91cl_bkgr
      - coco_orig_precision
      - 
      - 
      - 
      - 
-     - 
    * - mobilenet-v2
      - ImageNet2012
      - accuracy @ top1
      - 0.00%
      - 0.00%
      - 0.00%
-     - -0.23%
-     - -0.03%
+     - 0.01%
    * - resnet-50
      - ImageNet2012
      - accuracy @ top1
      - 0.00%
      - 0.00%
      - 0.00%
-     - 0.06%
      - 0.01%
    * - ssd-resnet34-1200
      - COCO2017_detection_80cl_bkgr
      - map
      - 0.02%
      - 0.02%
      - 0.01%
-     - 0.02%
-     - 0.06%
-   * - yolo_v8n
+     - -0.06%
+   * - yolo_v11
      - COCO2017_detection_80cl
      - map
-     - 0.01%
-     - 0.01%
-     - 0.01%
+     - -2.70%
+     - 
+     - 
      - 
-     - -0.03%
 .. list-table:: Model Accuracy for AMX-FP16, AMX-INT4, Arc-FP16 and Arc-INT4 (Arc™ B-series)
    :header-rows: 1
 
@@ -138,101 +131,101 @@ the table for more information.
    * - DeepSeek-R1-Distill-Llama-8B
      - Data Default WWB
      - Similarity
-     - 10.3%
-     - 21.4%
-     - 0.21%
-     - 23.5%
+     - 9.71%
+     - 21.25%
+     - 
+     - 21.04%
    * - DeepSeek-R1-Distill-Qwen-1.5B
      - Data Default WWB
      - Similarity
-     - 16.1%
+     - 8.45%
      - 34.5%
-     - 2.48%
-     - 36.4%
+     - 22.10%
+     - 32.02%
    * - DeepSeek-R1-Distill-Qwen-7B
      - Data Default WWB
      - Similarity
      - 25.5%
      - 35.6%
      - 3.9%
      - 37.2%
-   * - GLM4-9B-Chat
+   * - Gemma-2-9B-it
      - Data Default WWB
      - Similarity
-     - 6.9%
-     - 3.8%
-     - 6.3%
-     - 15.1%
-   * - Qwen-2.5-7B-instruct
+     - 0.89%
+     - 3.99%
+     - %
+     - 4.04%
+   * - GLM4-9B-Chat
      - Data Default WWB
      - Similarity
-     - 7.97%
-     - 25.12%
-     - 0.09%
-     - 23.87%
-   * - Gemma-2-9B
+     - 2.52%
+     - 8.48%
+     - 8.38%
+     - 
+   * - Qwen-2.5-7B-instruct
      - Data Default WWB
      - Similarity
-     - 4.81%
-     - 10.25%
-     - 1.73%
-     - 10.24%
+     - 1.51%
+     - 8.3%
+     - 
+     - 8.237%
    * - Llama-2-7b-chat
      - Data Default WWB
      - Similarity
-     - 1.80%
-     - 22.31%
-     - 0.13%
-     - 21.54%
-   * - Llama-3-8b
-     - Data Default WWB
-     - Similarity
-     - 2.26%
-     - 23.00%
-     - 0.12%
-     - 23.59%
+     - 1.43%
+     - 7.46%
+     - 
+     - 7.18%
    * - Llama-3.2-3b-instruct
      - Data Default WWB
      - Similarity
-     - 2.40%
-     - 11.25%
-     - 0.00%
-     - 12.32%
-   * - Mistral-7b-instruct-V0.2
+     - 2.75%
+     - 12.05%
+     - 0.52%
+     - 11.95%
+   * - Mistral-7b-instruct-V0.3
      - Data Default WWB
      - Similarity
-     - 2.94%
-     - 9.08%
-     - 0.37%
-     - 9.53%
+     - 2.46%
+     - 8.93%
+     - 3.17%
+     - 7.90%
    * - Phi3-mini-4k-instruct
      - Data Default WWB
      - Similarity
-     - 8.08%
-     - 7.93%
-     - 0.00%
-     - 8.30%
-   * - Qwen-2-7B
+     - 4.55%
+     - 7.23%
+     - 1.39%
+     - 8.47%
+   * - Phi4-mini-instruct
      - Data Default WWB
      - Similarity
-     - 4.97%
-     - 18.97%
-     - 0.00%
-     - 22.38%
+     - 6.59%
+     - 12.17%
+     - 1.91%
+     - 12.03%
+   * - Qwen2-VL-7B
+     - Data Default WWB
+     - Similarity
+     - 1.29%
+     - 8.71%
+     - 4.22%
+     - 9.43%
    * - Flux.1-schnell
      - Data Default WWB
      - Similarity
-     - 4.60%
-     - 4.20%
-     - 5.00%
-     - 3.30%
+     - 4.80%
+     - 3.80%
+     - 
+     - 2.80%
    * - Stable-Diffusion-V1-5
      - Data Default WWB
      - Similarity
-     - 2.50%
-     - 1.90%
-     - 2.10%
-     - 0.10%
+     - 3.00%
+     - 4.30%
+     - 0.50%
+     - 4.40%
 
 Notes: For all accuracy metrics a "-", (minus sign), indicates an accuracy drop.
 The Similarity metric is the distance from "perfect" and as such always positive.