Profile Job Results
Jobs
jp4vezlqp
Results Ready
Name
whisper_small_quantized_w8a16_WhisperSmallDecoderQuantizable
Target Device
- Dragonwing RB3 Gen 2 Vision Kit
- Qualcomm Linux 1.6
- Qualcomm® Dragonwing™ RB3 Gen 2 | QCS6490
Creator
ai-hub-support@qti.qualcomm.com
Target Model
Input Specs
input_ids: int32[1, 1]attention_mask: uint16[1, 1, 1, 200]k_cache_self_0_in: uint8[12, 1, 64, 199]v_cache_self_0_in: uint8[12, 1, 199, 64]k_cache_self_1_in: uint8[12, 1, 64, 199]v_cache_self_1_in: uint8[12, 1, 199, 64]k_cache_self_2_in: uint8[12, 1, 64, 199]v_cache_self_2_in: uint8[12, 1, 199, 64]k_cache_self_3_in: uint8[12, 1, 64, 199]v_cache_self_3_in: uint8[12, 1, 199, 64]k_cache_self_4_in: uint8[12, 1, 64, 199]v_cache_self_4_in: uint8[12, 1, 199, 64]k_cache_self_5_in: uint8[12, 1, 64, 199]v_cache_self_5_in: uint8[12, 1, 199, 64]k_cache_self_6_in: uint8[12, 1, 64, 199]v_cache_self_6_in: uint8[12, 1, 199, 64]k_cache_self_7_in: uint8[12, 1, 64, 199]v_cache_self_7_in: uint8[12, 1, 199, 64]k_cache_self_8_in: uint8[12, 1, 64, 199]v_cache_self_8_in: uint8[12, 1, 199, 64]k_cache_self_9_in: uint8[12, 1, 64, 199]v_cache_self_9_in: uint8[12, 1, 199, 64]k_cache_self_10_in: uint8[12, 1, 64, 199]v_cache_self_10_in: uint8[12, 1, 199, 64]k_cache_self_11_in: uint8[12, 1, 64, 199]v_cache_self_11_in: uint8[12, 1, 199, 64]k_cache_cross_0: uint8[12, 1, 64, 1500]v_cache_cross_0: uint8[12, 1, 1500, 64]k_cache_cross_1: uint8[12, 1, 64, 1500]v_cache_cross_1: uint8[12, 1, 1500, 64]k_cache_cross_2: uint8[12, 1, 64, 1500]v_cache_cross_2: uint8[12, 1, 1500, 64]k_cache_cross_3: uint8[12, 1, 64, 1500]v_cache_cross_3: uint8[12, 1, 1500, 64]k_cache_cross_4: uint8[12, 1, 64, 1500]v_cache_cross_4: uint8[12, 1, 1500, 64]k_cache_cross_5: uint8[12, 1, 64, 1500]v_cache_cross_5: uint8[12, 1, 1500, 64]k_cache_cross_6: uint8[12, 1, 64, 1500]v_cache_cross_6: uint8[12, 1, 1500, 64]k_cache_cross_7: uint8[12, 1, 64, 1500]v_cache_cross_7: uint8[12, 1, 1500, 64]k_cache_cross_8: uint8[12, 1, 64, 1500]v_cache_cross_8: uint8[12, 1, 1500, 64]k_cache_cross_9: uint8[12, 1, 64, 1500]v_cache_cross_9: uint8[12, 1, 1500, 64]k_cache_cross_10: uint8[12, 1, 64, 1500]v_cache_cross_10: uint8[12, 1, 1500, 64]k_cache_cross_11: uint8[12, 1, 64, 1500]v_cache_cross_11: uint8[12, 1, 1500, 64]position_ids: int32[1]Completion Time
12/6/2025, 1:49:47 PM
Options
--qairt_version 2.37Versions
- ONNX Runtime: 1.23.0
- QAIRT: v2.37.1.250807093845_124904
- Qualcomm Linux: 1.6-ver.1.2
- AI Hub: aihub-2025.11.18.0
Estimated Inference Time
32.7 ms
Estimated Peak Memory Usage
27 ‑ 59 MB
Compute Units
NPU
2878
| Stage | Time | Memory |
|---|---|---|
First App Load | 389 ms | 226‑227 MB |
Subsequent App Load | 376 ms | 226‑227 MB |
Inference | 32.7 ms | 27‑59 MB |
| ONNX Runtime | Value |
|---|---|
| execution_mode | SEQUENTIAL |
| intra_op_num_threads | 0 |
| inter_op_num_threads | 0 |
| enable_memory_pattern | false |
| enable_cpu_memory_arena | false |
| graph_optimization_level | ENABLE_ALL |
| QNN Execution Provider | Value |
|---|---|
| htp_performance_mode | "burst" |
| htp_graph_finalization_optimization_mode | "2" |
| enable_htp_fp16_precision | "0" |
| capture_network_visualizations | false |
| context_priority | "normal" |
| offload_graph_io_quantization | "1" |
Sign up to run this model on a hosted Qualcomm® device!
Run on device