Model: BAAI/URSA-0.6B-FSQ320 (transformer)
Device: npu:0
Dtype: float32
------------------------------------------------------------
Loaded 142/311 weights from transformer
Parameters: 794,127,360
--- CPU Inference ---
Output shape: [1, 16, 215669]
Logits[:5]: [2.6239, 2.1649, 6.9669, 1.5164, -6.7225]
Has NaN: False
--- NPU Inference (npu:0) ---
Output shape: [1, 16, 215669]
Logits[:5]: [2.6239, 2.1648, 6.9669, 1.5165, -6.7225]
Has NaN: False
--- Comparison ---
Cosine Similarity: 1.000143
Max Abs Error: 0.000124
--- Latency ---
Avg latency: 30.45 ms (10 runs)
Status: SUCCESS