{ "model_id": "whisper_small", "model_name": "Whisper-Small", "runtime": "precompiled_qnn_onnx", "precision": "float", "tool_versions": { "qairt": "2.45.0.260326154327" }, "model_files": { "encoder.onnx": { "inputs": { "input_features": { "shape": [ 1, 80, 3000 ], "dtype": "float16" } }, "outputs": { "k_cache_cross_0": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_0": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_1": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_1": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_2": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_2": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_3": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_3": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_4": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_4": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_5": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_5": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_6": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_6": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_7": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_7": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_8": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_8": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_9": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_9": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_10": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_10": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_11": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_11": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" } } }, "decoder.onnx": { "inputs": { "input_ids": { "shape": [ 1, 1 ], "dtype": "int32" }, "attention_mask": { "shape": [ 1, 1, 1, 200 ], "dtype": "float16" }, "k_cache_self_0_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_0_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_1_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_1_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_2_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_2_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_3_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_3_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_4_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_4_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_5_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_5_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_6_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_6_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_7_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_7_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_8_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_8_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_9_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_9_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_10_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_10_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_11_in": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_11_in": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_cross_0": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_0": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_1": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_1": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_2": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_2": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_3": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_3": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_4": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_4": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_5": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_5": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_6": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_6": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_7": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_7": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_8": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_8": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_9": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_9": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_10": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_10": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "k_cache_cross_11": { "shape": [ 12, 1, 64, 1500 ], "dtype": "float16" }, "v_cache_cross_11": { "shape": [ 12, 1, 1500, 64 ], "dtype": "float16" }, "position_ids": { "shape": [ 1 ], "dtype": "int32" } }, "outputs": { "logits": { "shape": [ 1, 51865, 1, 1 ], "dtype": "float16" }, "k_cache_self_0_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_0_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_1_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_1_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_2_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_2_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_3_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_3_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_4_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_4_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_5_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_5_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_6_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_6_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_7_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_7_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_8_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_8_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_9_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_9_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_10_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_10_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" }, "k_cache_self_11_out": { "shape": [ 12, 1, 64, 199 ], "dtype": "float16" }, "v_cache_self_11_out": { "shape": [ 12, 1, 199, 64 ], "dtype": "float16" } } } }, "chipset_attributes": { "aliases": [ "qualcomm-snapdragon-x2-elite", "sc8480xp" ], "marketing_name": "Snapdragon® X2 Elite", "world": "Compute", "supports_fp16": true, "htp_version": 81, "soc_model": 88, "reference_device": "Snapdragon X2 Elite CRD", "supports_weight_sharing": true, "name": "qualcomm-snapdragon-x2-elite" } }