Manual Tuning

Tuning Workflow

If you find that the accuracy of the model quantized based on the initial config.json file generated by the create_quant_config API call is not as expected, you can tune the configuration parameters until the accuracy meets your requirement. The workflow for manually tuning the parameters in the PTQ configuration file config.json goes through the following three phases:

1. Tune the amount of data used for calibration.
2. Skip quantizing certain layers.
3. Tune the quantization algorithm and parameters.

Specifically,

1. Run quantization based on the initial config.json file generated by the create_quant_config API call. If the accuracy of the quantized model is satisfactory, stop tuning the configuration parameters. Otherwise, go to 2.
2. Tweak the value of batch_num to tune the amount of data used for calibration.

   batch_num controls the batch number for quantization. Tune it based on the batch size and the dataset size. Generally:

   Generally, a larger value of batch_num indicates more data samples used for quantization and a smaller accuracy drop of the quantized model. However, excessive data does not necessarily improve accuracy, but certainly consumes more memory and reduces the quantization speed, hence resulting in insufficient memory, video RAM, and thread resources. An optimal tradeoff is achieved when the product of batch_num and batch_size (the number of images per batch) is 16 or 32.

3. Run quantization based on the new configuration generated in 2. If the accuracy of the quantized model is satisfactory, stop tuning the configuration parameters. Otherwise, go to 4.
4. Tweak the value of quant_enable to skip quantizing certain layers.

   quant_enable is the quantization switch of a specified layer. The value false indicates that the layer will be skipped during quantization; true, otherwise. Removing the layer configuration can also skip the layer.

   Quantizing a model can have a negative effect on accuracy. Layers sensitive to quantization will suffer from remarkable error increases once quantized and therefore should be left unquantized. Spot these layers as follows:

   1. In a model, the input layer, output layer, and layers with especially fewer parameters are likely to be quantization-sensitive.
   2. Use the Model Accuracy Analyzer to compare the output errors between the source model and the quantized model layer-wise (a cosine similarity of at least 0.99, for example) to locate the layers that reduce accuracy the most.
5. Run quantization based on the new configuration generated in 4. If the accuracy of the quantized model is satisfactory, stop tuning the configuration parameters. Otherwise, go to 6.
6. Tweak the values of activation_quant_params and weight_quant_params to tune the quantization algorithms and parameters.

   For details about the algorithm parameters, see Command-Line Options. For details about the algorithm, see PTQ Algorithms.

7. Run quantization based on the new configuration generated in 6. If the accuracy of the quantized model is satisfactory, stop tuning the configuration parameters. Otherwise, it indicates that your model is not suitable for quantization and the quantization configuration should be removed.

Quantization Configuration File

If the inference accuracy of the config.json quantization configuration file generated by the create_quant_config call does not meet the requirements, you need to tune the config.json file until the accuracy is as expected. The following is an example of the file content.

• Uniform quantization configuration file (with ifmr: IFMR algorithm for activation quantization used for activation quantization)

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68

  { "version":1, "batch_num":2, "activation_offset":true, "joint_quant":false, "do_fusion":true, "skip_fusion_layers":[], "tensor_quantize":[ { "layer_name": "maxpool_ld_default", "input_index":0, "activation_quant_params":{ "num_bits":8, "max_percentile":0.999999, "min_percentile":0.999999, "search_range":[ 0.7, 1.3 ], "search_step":0.01, "act_algo":"ifmr", "asymmetric":false } } ], "layer_name1":{ "quant_enable":true, "dmq_balancer_param":0.5, "activation_quant_params":{ "num_bits":8, "max_percentile":0.999999, "min_percentile":0.999999, "search_range":[ 0.7, 1.3 ], "search_step":0.01, "act_algo":"ifmr", "asymmetric":false }, "weight_quant_params":{ "num_bits":8, "wts_algo":"arq_quantize", "channel_wise":true } }, "layer_name2":{ "quant_enable":true, "dmq_balancer_param":0.5, "activation_quant_params":{ "num_bits":8, "max_percentile":0.999999, "min_percentile":0.999999, "search_range":[ 0.7, 1.3 ], "search_step":0.01, "act_algo":"ifmr", "asymmetric":false }, "weight_quant_params":{ "num_bits":8, "wts_algo":"arq_quantize", "channel_wise":false } } }

• Uniform quantization configuration file (with HFMG for Activation Quantization used for activation quantization)

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

  { "version":1, "batch_num":2, "activation_offset":true, "do_fusion":true, "skip_fusion_layers":[], "tensor_quantize":[ { "layer_name": "maxpool_ld_default", "input_index":0, "activation_quant_params":{ "num_bits":8, "max_percentile":0.999999, "min_percentile":0.999999, "search_range":[ 0.7, 1.3 ], "search_step":0.01 "act_algo":"hfmg" "asymmetric":false } } }, "layer_name1":{ "quant_enable":true, "dmq_balancer_param":0.5, "activation_quant_params":{ "num_bits":8, "act_algo":"hfmg", "num_of_bins":4096 "asymmetric":false }, "weight_quant_params":{ "num_bits":8, "wts_algo":"arq_quantize", "channel_wise":true } } }

Command-Line Options

The following tables describe the parameters in the configuration file.