Profile Job Results
Jobs
jp0rmo16p
Results Ready
Name
trocr_TrOCRDecoder
Target Device
- Dragonwing IQ-9075 EVK
- Qualcomm Linux 1.7
- Qualcomm® Dragonwing™ IQ-9075 | QCS9075
Creator
ai-hub-support@qti.qualcomm.com
Target Model
Input Specs
input_ids: int32[1, 1]index: int32[1]kv_0_attn_key: float32[1, 8, 19, 32]kv_0_attn_val: float32[1, 8, 19, 32]kv_0_cross_attn_key: float32[1, 8, 578, 32]kv_0_cross_attn_val: float32[1, 8, 578, 32]kv_1_attn_key: float32[1, 8, 19, 32]kv_1_attn_val: float32[1, 8, 19, 32]kv_1_cross_attn_key: float32[1, 8, 578, 32]kv_1_cross_attn_val: float32[1, 8, 578, 32]kv_2_attn_key: float32[1, 8, 19, 32]kv_2_attn_val: float32[1, 8, 19, 32]kv_2_cross_attn_key: float32[1, 8, 578, 32]kv_2_cross_attn_val: float32[1, 8, 578, 32]kv_3_attn_key: float32[1, 8, 19, 32]kv_3_attn_val: float32[1, 8, 19, 32]kv_3_cross_attn_key: float32[1, 8, 578, 32]kv_3_cross_attn_val: float32[1, 8, 578, 32]kv_4_attn_key: float32[1, 8, 19, 32]kv_4_attn_val: float32[1, 8, 19, 32]kv_4_cross_attn_key: float32[1, 8, 578, 32]kv_4_cross_attn_val: float32[1, 8, 578, 32]kv_5_attn_key: float32[1, 8, 19, 32]kv_5_attn_val: float32[1, 8, 19, 32]kv_5_cross_attn_key: float32[1, 8, 578, 32]kv_5_cross_attn_val: float32[1, 8, 578, 32]Completion Time
2/24/2026, 8:03:01 AM
Versions
- TensorFlow Lite: 2.17.0
- QAIRT: v2.43.0.260127150333_193827
- QNN TfLite Delegate: v2.43.0.260127150333_193827
- Qualcomm Linux: 1.7-ver.1.1
- AI Hub: aihub-2026.02.14.0
Estimated Inference Time
2.60 ms
Estimated Peak Memory Usage
0 ‑ 83 MB
Compute Units
NPU
399
| Stage | Time | Memory |
|---|---|---|
First App Load | 5.82 s | 320‑321 MB |
Subsequent App Load | 232 ms | 76 MB |
Inference | 2.60 ms | 0‑83 MB |
| TensorFlow Lite | Value |
|---|---|
| number_of_threads | 4 |
| QNN Delegate | Value |
|---|---|
| backend_type | kHtpBackend |
| log_level | kLogLevelWarn |
| htp_options.performance_mode | kHtpBurst |
| htp_options.precision | kHtpFp16 |
| htp_options.optimization_strategy | kHtpOptimizeForInference |
| htp_options.useConvHmx | true |
Sign up to run this model on a hosted Qualcomm® device!
Run on device