Analysis Results

FP16 Overflow/Underflow Detection

Detects overflow/underflow errors in FP16 data of the comparison data. If overflow/underflow data exists, an expert suggestion is output.

Analysis result of Advisor:

Detection Type: FP16 overflow

Operator Index: 228

Expert Advice: Float16 data overflow occurs. Rectify the fault and perform comparison again.

Input Inconsistency Check

Checks whether the two batches of input data to be compared on the entire network are consistent. If inconsistency exists (the cosine similarity is less than 0.99), an expert suggestion is output.

Analysis result of Advisor:

Detection Type: Input inconsistent

Operator Index: 0

Expert Advice: The input data of NPUDump is inconsistent with that of GroundTruth. Use the same data or check the data preprocessing process.

Network-wide Consistency Check (Faulty Node Detection)

In the network-wide comparison result, checks whether the value of a layer is less than the threshold, whether the values of subsequent layers are less than the threshold, or whether the value of the last layer is less than the threshold (the cosine similarity is less than 0.99), and outputs quantized error correction suggestions.

Analysis result of Advisor:

Detection Type: global consistency

Operator Index: 1174

Expert Advice: The accuracy of some tensors is low, resulting in an unqualified final accuracy. This may be caused by quantization. Calibrate the data or contact Huawei for further diagnosis.

Network-wide Consistency Check (Single-Point Error Detection)

In the network-wide comparison result, checks that the value of a layer is less than the threshold (the cosine similarity is less than 0.99), but the final result meets the accuracy requirement, and outputs the expert suggestion.

Analysis result of Advisor:

Detection Type: global consistency

Operator Index: 195

Expert Advice: The accuracy of some tensors is low, while the final accuracy is qualified. This may be caused by Ascend internal optimization. Ignore or contact Huawei for further diagnosis.

Network-wide Consistency Check (Consistency Check)

Checks that all data in the comparison result meets the accuracy requirements, and outputs the expert suggestion.

Analysis result of Advisor:

Detection Type: global consistency

Operator Index: NA

Expert Advice: All data in the comparison result meets the accuracy requirements.

If data accuracy of the model is still not up to standard in practical application, please check the post-processing process of model outputs.