megengine.quantization.LSQ¶
- class LSQ(dtype, enable=True, eps=1e-05, **kwargs)[源代码]¶
LSQ: https://arxiv.org/pdf/1902.08153.pdf Estimating and scaling the task loss gradient at each weight and activation layer’s quantizer step size
- 参数
dtype (
Union[str,QuantDtypeMeta]) – a string orQuantDtypeMetaindicating the target quantization dtype of input.enable (
bool) – whether donormal_forwardorfake_quant_forward.eps (
float) – a small value to avoid division by zero. Default: 1e-5
Methods
apply(fn)Applies function
fnto all the modules within this module, including itself.buffers([recursive])Returns an iterable for the buffers of the module.
children(**kwargs)Returns an iterable for all the submodules that are direct attributes of this module.
disable()disable_quantize([value])Sets
module'squantize_disabledattribute and returnmodule.enable()eval()Sets training mode of all the modules within this module (including itself) to
False.fake_quant_forward(inp[, qparams])forward(inp[, qparams])load_state_dict(state_dict[, strict])Loads a given dictionary created by
state_dictinto this module.modules(**kwargs)Returns an iterable for all the modules within this module, including itself.
named_buffers([prefix, recursive])Returns an iterable for key buffer pairs of the module, where
keyis the dotted path from this module to the buffer.named_children(**kwargs)Returns an iterable of key-submodule pairs for all the submodules that are direct attributes of this module, where 'key' is the attribute name of submodules.
named_modules([prefix])Returns an iterable of key-module pairs for all the modules within this module, including itself, where 'key' is the dotted path from this module to the submodules.
named_parameters([prefix, recursive])Returns an iterable for key
Parameterpairs of the module, wherekeyis the dotted path from this module to theParameter.named_tensors([prefix, recursive])Returns an iterable for key tensor pairs of the module, where
keyis the dotted path from this module to the tensor.normal_forward(inp[, qparams])parameters([recursive])Returns an iterable for the
Parameterof the module.register_forward_hook(hook)Registers a hook to handle forward results.
Registers a hook to handle forward inputs.
replace_param(params, start_pos[, seen])Replaces module's parameters with
params, used byParamPacktoset_qparams(qparams)state_dict([rst, prefix, keep_var])tensors([recursive])Returns an iterable for the
Tensorof the module.train([mode, recursive])Sets training mode of all the modules within this module (including itself) to
mode.Sets all parameters' grads to zero