4-bit conformer with accurate quantization training for speech recognition
A method for training a model includes obtaining a plurality of training samples. Each respective training sample of the plurality of training samples includes a respective speech utterance and a respective textual utterance representing a transcription of the respective speech utterance. The method includes training, using quantization aware training with native integer operations, an automatic speech recognition (ASR) model on the plurality of training samples. The method also includes quantizing the trained ASR model to an integer target fixed-bit width. The quantized trained ASR model includes a plurality of weights. Each weight of the plurality of weights includes an integer with the target fixed-bit width. The method includes providing the quantized trained ASR model to a user device.
1. A computer-implemented method when executed on data processing hardware causes the data processing hardware to perform operations comprising:
obtaining a plurality of training samples, each respective training sample of the plurality of training samples comprising:
a respective speech utterance; and
a respective textual utterance representing a transcription of the respective speech utterance;
training, using quantization aware training with native integer operations, an automatic speech recognition (ASR) model on the plurality of training samples;
quantizing the trained ASR model to an integer target fixed-bit width, the quantized trained ASR model comprising a plurality of weights, each weight of the plurality of weights comprising an integer with the target fixed-bit width; and
providing the quantized trained ASR model to a user device.
2. The method of claim 1 , wherein the target fixed-bit width is four.
3. The method of claim 1 , wherein the ASR model further comprises a plurality of activations, each activation of the plurality of activations comprising an integer with the target fixed-bit width.
4. The method of claim 1 , wherein the ASR model further comprises a plurality of activations, each activation of the plurality of activations comprising an integer with a fixed bit width greater than the target fixed-bit width.
5. The method of claim 1 , wherein the ASR model further comprises a plurality of activations, each activation of the plurality of activations comprising a float value.
6. The method of claim 1 , wherein quantizing the trained ASR model comprises determining a scale factor based on an estimated max value of an axis to be quantized and the target fixed-bit width.
7. The method of claim 1 , wherein the ASR model comprises one or more multi-head attention layers.
8. The method of claim 7 , wherein the one or more multi-head attention layers comprise one or more conformer layers or one or more transformer layers.
9. The method of claim 1 , wherein:
the ASR model comprises a plurality of encoders and a plurality of decoders; and
quantizing the ASR model comprises quantizing the plurality of encoders and not quantizing the plurality of decoders.
10. The method of claim 1 , wherein:
the ASR model comprises an audio encoder; and
the audio encoder comprises a cascaded encoder comprising a first causal encoder and a second non-causal encoder.
11. A system comprising:
data processing hardware; and
memory hardware in communication with the data processing hardware, the memory hardware storing instructions that when executed on the data processing hardware cause the data processing hardware to perform operations comprising:
obtaining a plurality of training samples, each respective training sample of the plurality of training samples comprising:
a respective speech utterance; and
a respective textual utterance representing a transcription of the respective speech utterance;
training, using quantization aware training with native integer operations, an automatic speech recognition (ASR) model on the plurality of training samples;
quantizing the trained ASR model to an integer target fixed-bit width, the quantized trained ASR model comprising a plurality of weights, each weight of the plurality of weights comprising an integer with the target fixed-bit width; and
providing the quantized trained ASR model to a user device.
12. The system of claim 11 , wherein the target fixed-bit width is four.
13. The system of claim 11 , wherein the ASR model further comprises a plurality of activations, each activation of the plurality of activations comprising an integer with the target fixed-bit width.
14. The system of claim 11 , wherein the ASR model further comprises a plurality of activations, each activation of the plurality of activations comprising an integer with a fixed bit width greater than the target fixed-bit width.
15. The system of claim 11 , wherein the ASR model further comprises a plurality of activations, each activation of the plurality of activations comprising a float value.
16. The system of claim 11 , wherein quantizing the trained ASR model comprises determining a scale factor based on an estimated max value of an axis to be quantized and the target fixed-bit width.
17. The system of claim 11 , wherein the ASR model comprises one or more multi-head attention layers.
18. The system of claim 17 , wherein the one or more multi-head attention layers comprise one or more conformer layers or one or more transformer layers.
19. The system of claim 11 , wherein:
the ASR model comprises a plurality of encoders and a plurality of decoders; and
quantizing the ASR model comprises quantizing the plurality of encoders and not quantizing the plurality of decoders.
20. The system of claim 11 , wherein:
the ASR model comprises an audio encoder; and
the audio encoder comprises a cascaded encoder comprising a first causal encoder and a second non-causal encoder.