v0.7.0r1

Update Performance Data

Signed-off-by: QAIHM Team <[email protected]>

qai_hub_models/models/aotgan/export.py

@@ -187,11 +187,11 @@ def export_model(
     # 5. Download the model asset to a local file
     if not skip_downloading:
         if target_runtime == TargetRuntime.QNN:
-            target_runtime_extension = ".so"
+            target_runtime_extension = "so"
         elif target_runtime == TargetRuntime.TFLITE:
-            target_runtime_extension = ".tflite"
+            target_runtime_extension = "tflite"
         elif target_runtime == TargetRuntime.ORT:
-            target_runtime_extension = ".onnx"
+            target_runtime_extension = "onnx"
 
         os.makedirs(output_path, exist_ok=True)
         target_model: hub.Model = compile_job.get_target_model()  # type: ignore

qai_hub_models/models/aotgan/perf.yaml

@@ -65,29 +65,14 @@ models:
         total_layers: 275
       job_id: jwgovdz45
       job_status: Passed
-    torchscript_onnx_ort:
-      inference_time: 'null'
-      throughput: 'null'
-      estimated_peak_memory_range:
-        min: 0
-        max: 0
-      primary_compute_unit: 'null'
-      precision: 'null'
-      layer_info:
-        layers_on_npu: 0
-        layers_on_gpu: 0
-        layers_on_cpu: 0
-        total_layers: 0
-      job_id: jygz7z2zp
-      job_status: Failed
     reference_device_info:
       name: Samsung Galaxy S23
       os: '13'
       form_factor: Phone
       os_name: Android
       manufacturer: Samsung
       chipset: Snapdragon® 8 Gen 2
-    timestamp: '2024-05-24T06:12:49.396717Z'
+    timestamp: '2024-05-29T18:59:39.212193Z'
   - torchscript_onnx_tflite:
       inference_time: 120767.0
       throughput: 8.280407727276492
@@ -118,29 +103,14 @@ models:
         total_layers: 275
       job_id: j1pvw2q7g
       job_status: Passed
-    torchscript_onnx_ort:
-      inference_time: 'null'
-      throughput: 'null'
-      estimated_peak_memory_range:
-        min: 0
-        max: 0
-      primary_compute_unit: 'null'
-      precision: 'null'
-      layer_info:
-        layers_on_npu: 0
-        layers_on_gpu: 0
-        layers_on_cpu: 0
-        total_layers: 0
-      job_id: jz5w9ywzp
-      job_status: Failed
     reference_device_info:
       name: Samsung Galaxy S24
       os: '14'
       form_factor: Phone
       os_name: Android
       manufacturer: Samsung
       chipset: Snapdragon® 8 Gen 3
-    timestamp: '2024-05-24T06:12:49.517726Z'
+    timestamp: '2024-05-29T18:59:39.339992Z'
   - torchscript_onnx_tflite:
       inference_time: 164352.0
       throughput: 6.084501557632398
@@ -178,7 +148,7 @@ models:
       os_name: Android
       manufacturer: Qualcomm
       chipset: Qcs8550
-    timestamp: '2024-05-24T06:12:49.635903Z'
+    timestamp: '2024-05-29T18:59:39.469404Z'
   - torchscript_onnx_qnn:
       inference_time: 145505.0
       throughput: 6.872616061303735
@@ -194,41 +164,11 @@ models:
         total_layers: 275
       job_id: j7gjl3d7p
       job_status: Passed
-    torchscript_onnx_ort:
-      inference_time: 'null'
-      throughput: 'null'
-      estimated_peak_memory_range:
-        min: 0
-        max: 0
-      primary_compute_unit: 'null'
-      precision: 'null'
-      layer_info:
-        layers_on_npu: 0
-        layers_on_gpu: 0
-        layers_on_cpu: 0
-        total_layers: 0
-      job_id: jmg94o0q5
-      job_status: Failed
-    torchscript_onnx_ort_dml_gpu:
-      inference_time: 739119.0
-      throughput: 1.3529621075902527
-      estimated_peak_memory_range:
-        min: 416927744
-        max: 416927744
-      primary_compute_unit: CPU
-      precision: fp32
-      layer_info:
-        layers_on_npu: 0
-        layers_on_gpu: 0
-        layers_on_cpu: 234
-        total_layers: 234
-      job_id: jnp18o2kg
-      job_status: Passed
     reference_device_info:
       name: Snapdragon X Elite CRD
       os: '11'
       form_factor: Compute
       os_name: Windows
       manufacturer: Qualcomm
       chipset: Snapdragon® X Elite
-    timestamp: '2024-05-24T06:12:49.749014Z'
+    timestamp: '2024-05-29T18:59:39.597982Z'

qai_hub_models/models/convnext_tiny/export.py

@@ -185,11 +185,11 @@ def export_model(
     # 5. Download the model asset to a local file
     if not skip_downloading:
         if target_runtime == TargetRuntime.QNN:
-            target_runtime_extension = ".so"
+            target_runtime_extension = "so"
         elif target_runtime == TargetRuntime.TFLITE:
-            target_runtime_extension = ".tflite"
+            target_runtime_extension = "tflite"
         elif target_runtime == TargetRuntime.ORT:
-            target_runtime_extension = ".onnx"
+            target_runtime_extension = "onnx"
 
         os.makedirs(output_path, exist_ok=True)
         target_model: hub.Model = compile_job.get_target_model()  # type: ignore

qai_hub_models/models/convnext_tiny/perf.yaml

@@ -87,7 +87,7 @@ models:
       os_name: Android
       manufacturer: Samsung
       chipset: Snapdragon® 8 Gen 2
-    timestamp: '2024-05-24T06:12:49.806533Z'
+    timestamp: '2024-05-29T18:59:39.628526Z'
   - torchscript_onnx_tflite:
       inference_time: 3967.0
       throughput: 252.07965717166624
@@ -140,7 +140,7 @@ models:
       os_name: Android
       manufacturer: Samsung
       chipset: Snapdragon® 8 Gen 3
-    timestamp: '2024-05-24T06:12:49.806634Z'
+    timestamp: '2024-05-29T18:59:39.628637Z'
   - torchscript_onnx_tflite:
       inference_time: 5705.0
       throughput: 175.28483786152498
@@ -178,7 +178,7 @@ models:
       os_name: Android
       manufacturer: Qualcomm
       chipset: Qcs8550
-    timestamp: '2024-05-24T06:12:49.806715Z'
+    timestamp: '2024-05-29T18:59:39.628718Z'
   - torchscript_onnx_qnn:
       inference_time: 3927.0
       throughput: 254.64731347084287
@@ -209,26 +209,11 @@ models:
         total_layers: 189
       job_id: j2p0ro32p
       job_status: Passed
-    torchscript_onnx_ort_dml_gpu:
-      inference_time: 35785.0
-      throughput: 27.94466955428252
-      estimated_peak_memory_range:
-        min: 1454080
-        max: 1454080
-      primary_compute_unit: CPU
-      precision: fp32
-      layer_info:
-        layers_on_npu: 0
-        layers_on_gpu: 0
-        layers_on_cpu: 202
-        total_layers: 202
-      job_id: j1p87j0z5
-      job_status: Passed
     reference_device_info:
       name: Snapdragon X Elite CRD
       os: '11'
       form_factor: Compute
       os_name: Windows
       manufacturer: Qualcomm
       chipset: Snapdragon® X Elite
-    timestamp: '2024-05-24T06:12:49.806795Z'
+    timestamp: '2024-05-29T18:59:39.628784Z'

qai_hub_models/models/convnext_tiny_w8a16_quantized/export.py

@@ -193,11 +193,11 @@ def export_model(
     # 5. Download the model asset to a local file
     if not skip_downloading:
         if target_runtime == TargetRuntime.QNN:
-            target_runtime_extension = ".so"
+            target_runtime_extension = "so"
         elif target_runtime == TargetRuntime.TFLITE:
-            target_runtime_extension = ".tflite"
+            target_runtime_extension = "tflite"
         elif target_runtime == TargetRuntime.ORT:
-            target_runtime_extension = ".onnx"
+            target_runtime_extension = "onnx"
 
         os.makedirs(output_path, exist_ok=True)
         target_model: hub.Model = compile_job.get_target_model()  # type: ignore

qai_hub_models/models/convnext_tiny_w8a16_quantized/perf.yaml

@@ -63,7 +63,7 @@ models:
       os_name: Android
       manufacturer: Samsung
       chipset: Snapdragon® 8 Gen 2
-    timestamp: '2024-05-24T06:12:50.094158Z'
+    timestamp: '2024-05-29T18:59:39.928338Z'
   - torchscript_onnx_qnn:
       inference_time: 2291.0
       throughput: 436.4906154517678
@@ -86,7 +86,7 @@ models:
       os_name: Android
       manufacturer: Samsung
       chipset: Snapdragon® 8 Gen 3
-    timestamp: '2024-05-24T06:12:50.094352Z'
+    timestamp: '2024-05-29T18:59:39.928544Z'
   - torchscript_onnx_qnn:
       inference_time: 3277.0
       throughput: 305.15715593530666
@@ -109,7 +109,7 @@ models:
       os_name: Android
       manufacturer: Qualcomm
       chipset: Qcs8550
-    timestamp: '2024-05-24T06:12:50.094519Z'
+    timestamp: '2024-05-29T18:59:39.928727Z'
   - torchscript_onnx_qnn:
       inference_time: 11165.0
       throughput: 89.56560680698611
@@ -132,15 +132,15 @@ models:
       os_name: Android
       manufacturer: Qualcomm
       chipset: Qcs6490
-    timestamp: '2024-05-24T06:12:50.216033Z'
+    timestamp: '2024-05-29T18:59:40.073528Z'
   - reference_device_info:
       name: RB5 (Proxy)
       os: '12'
       form_factor: Iot
       os_name: Android
       manufacturer: Qualcomm
       chipset: Qcs8250
-    timestamp: '2024-05-24T06:12:50.334252Z'
+    timestamp: '2024-05-29T18:59:40.202310Z'
   - torchscript_onnx_qnn:
       inference_time: 3621.0
       throughput: 276.16680475006905
@@ -163,4 +163,4 @@ models:
       os_name: Windows
       manufacturer: Qualcomm
       chipset: Snapdragon® X Elite
-    timestamp: '2024-05-24T06:12:50.334505Z'
+    timestamp: '2024-05-29T18:59:40.203221Z'

qai_hub_models/models/convnext_tiny_w8a8_quantized/export.py

@@ -193,11 +193,11 @@ def export_model(
     # 5. Download the model asset to a local file
     if not skip_downloading:
         if target_runtime == TargetRuntime.QNN:
-            target_runtime_extension = ".so"
+            target_runtime_extension = "so"
         elif target_runtime == TargetRuntime.TFLITE:
-            target_runtime_extension = ".tflite"
+            target_runtime_extension = "tflite"
         elif target_runtime == TargetRuntime.ORT:
-            target_runtime_extension = ".onnx"
+            target_runtime_extension = "onnx"
 
         os.makedirs(output_path, exist_ok=True)
         target_model: hub.Model = compile_job.get_target_model()  # type: ignore