Qualcomm® AI HubAI Hub

Profile Job Results

Jobs
jg9mm77l5
Results Ready
Name
qwen2_5_vl_7b_instruct_vision_encoder_qwen2_5_vl_7b_instruct
Target Device
  • Dragonwing IQ-9075 EVK
  • Qualcomm Linux 1.9
  • Qualcomm® Dragonwing™ IQ-9075 | QCS9075
Creator
ai-hub-support@qti.qualcomm.com
Input Specs
pixel_values: uint16[864, 1176]
position_ids_cos: uint16[864, 40]
position_ids_sin: uint16[864, 40]
window_attention_mask: uint16[1, 864, 864]
full_attention_mask: uint16[1, 864, 864]
Completion Time
8/29/2026, 8:28:24 AM
Versions
  • QAIRT: v2.45.0.260326154327
  • QNN Backend API: 5.45.0
  • QNN Core API: 2.34.0
  • Qualcomm Linux: 1.9-ver.1.0
  • AI Hub: aihub-2026.08.14.0
Estimated Inference Time
459 ms
Estimated Peak Memory Usage
2 ‑ 10 MB
Compute Units
NPU
2071
StageTimeMemory
First App Load
509 ms1‑2 MB
Subsequent App Load
507 ms3‑4 MB
Inference
459 ms2‑10 MB
QNNValue
context_options.htp_options.performance_modeBURST
default_graph_options.htp_options.optimizations[0].typeFINALIZE_OPTIMIZATION_FLAG
default_graph_options.htp_options.optimizations[0].value3.0
default_graph_options.htp_options.precisionFLOAT16
default_graph_options.htp_options.vtcm_size0

Sign up to run this model on a hosted Qualcomm® device!

Run on device