From 35431ce1cf23e66964e25e8ba9e62a964323c337 Mon Sep 17 00:00:00 2001 From: cc <52520497+juncaipeng@users.noreply.github.com> Date: Tue, 27 Oct 2020 09:59:21 +0800 Subject: [PATCH] post_quant_static: remove the cache params, add optimize_model params (#471) * post_quant_static: remove the cache params, add optimize_model params, test=develop Co-authored-by: Bai Yifan --- docs/zh_cn/api_cn/quantization_api.rst | 8 +++----- paddleslim/quant/quanter.py | 12 +++++++----- 2 files changed, 10 insertions(+), 10 deletions(-) diff --git a/docs/zh_cn/api_cn/quantization_api.rst b/docs/zh_cn/api_cn/quantization_api.rst index 1b3358b1..1407a4f2 100644 --- a/docs/zh_cn/api_cn/quantization_api.rst +++ b/docs/zh_cn/api_cn/quantization_api.rst @@ -97,7 +97,7 @@ quant_post_dynamic quant_post_static --------------- -.. py:function:: paddleslim.quant.quant_post_static(executor,model_dir, quantize_model_path, batch_generator=None, sample_generator=None, model_filename=None, params_filename=None, save_model_filename='__model__', save_params_filename='__params__', batch_size=16, batch_nums=None, scope=None, algo='KL', quantizable_op_type=["conv2d","depthwise_conv2d","mul"], is_full_quantize=False, weight_bits=8, activation_bits=8, activation_quantize_type='range_abs_max', weight_quantize_type='channel_wise_abs_max', is_use_cache_file=False, cache_dir="./temp_post_training") +.. py:function:: paddleslim.quant.quant_post_static(executor,model_dir, quantize_model_path, batch_generator=None, sample_generator=None, model_filename=None, params_filename=None, save_model_filename='__model__', save_params_filename='__params__', batch_size=16, batch_nums=None, scope=None, algo='KL', quantizable_op_type=["conv2d","depthwise_conv2d","mul"], is_full_quantize=False, weight_bits=8, activation_bits=8, activation_quantize_type='range_abs_max', weight_quantize_type='channel_wise_abs_max', optimize_model=False) `源代码 `_ @@ -146,9 +146,7 @@ quant_post_static - **activation_bits(int)** - 激活值的量化比特位数, 默认值为8。 - **weight_quantize_type(str)** - weight的量化方式，可选 `abs_max` 或者 `channel_wise_abs_max` ,通常情况下选 `channel_wise_abs_max` 模型量化精度更高。 - **activation_quantize_type(str)** - 激活值的量化方式, 可选 `range_abs_max` 和 `moving_average_abs_max` 。设置激活量化方式不会影响计算scale的算法，只是影响在保存模型时使用哪种operator。 -- **is_use_cache_file(bool)** - 是否使用硬盘对中间结果进行存储。如果为False, 则将中间结果存储在内存中。默认值为False。 -- **cache_dir(str)** - 如果 ``'is_use_cache_file'`` 为True, 则将中间结果存储在此参数设置的路径下。默认值为 ``./temp_post_training`` 。 - +- **optimize_model(bool)** - 是否在量化之前对模型进行fuse优化。executor必须在cpu上执才可以设置该参数为True，然后会将`conv2d/depthwise_conv2d/conv2d_tranpose + batch_norm`进行fuse。 **返回** 无。 @@ -316,7 +314,7 @@ convert 'mul', 'conv2d', 'pool2d', 'depthwise_conv2d', 'elementwise_add', 'leaky_relu' ] - TRANSFORM_PASS_OP_TYPES = ['conv2d', 'depthwise_conv2d', 'mul'] + TRANSFORM_PASS_OP_TYPES = ['conv2d', 'depthwise_conv2d', 'mul', 'conv2d_transpose'] QUANT_DEQUANT_PASS_OP_TYPES = [ "pool2d", "elementwise_add", "concat", "softmax", "argmax", "transpose", diff --git a/paddleslim/quant/quanter.py b/paddleslim/quant/quanter.py index 32886172..0024ee31 100755 --- a/paddleslim/quant/quanter.py +++ b/paddleslim/quant/quanter.py @@ -321,6 +321,7 @@ def quant_post_static( activation_bits=8, activation_quantize_type='range_abs_max', weight_quantize_type='channel_wise_abs_max', + optimize_model=False, is_use_cache_file=False, cache_dir="./temp_post_training"): """ @@ -377,9 +378,11 @@ def quant_post_static( the model accuracy is usually higher when using 'channel_wise_abs_max'. is_full_quantize(bool): if True, apply quantization to all supported quantizable op type. If False, only apply quantization to the input quantizable_op_type. Default is False. - is_use_cache_file(bool): If False, all temp data will be saved in memory. If True, - all temp data will be saved to disk. Defalut: False. - cache_dir(str): When 'is_use_cache_file' is True, temp data will be save in 'cache_dir'. Default is './temp_post_training'. + optimize_model(bool, optional): If set optimize_model as True, it applies some + passes to optimize the model before quantization. So far, the place of + executor must be cpu it supports fusing batch_norm into convs. + is_use_cache_file(bool): This param is deprecated. + cache_dir(str): This param is deprecated. Returns: None @@ -401,8 +404,7 @@ def quant_post_static( activation_bits=activation_bits, activation_quantize_type=activation_quantize_type, weight_quantize_type=weight_quantize_type, - is_use_cache_file=is_use_cache_file, - cache_dir=cache_dir) + optimize_model=optimize_model) post_training_quantization.quantize() post_training_quantization.save_quantized_model( quantize_model_path, -- GitLab