\u200E

开始使用
特性
文档
- API
- 使用指南
工具平台
- 工具
- 平台
资源

develop
3.0-beta
2.6
2.5
2.4
2.3
2.2
2.1
2.0
1.8

中文(简)
English(En)

Installation Guide
Guides
API Reference
- paddle
  - abs
  - abs_
  - acos
  - acos_
  - acosh
  - add
  - add_n
  - addmm
  - addmm_
  - all
  - allclose
  - amax
  - amin
  - angle
  - any
  - arange
  - argmax
  - argmin
  - argsort
  - as_complex
  - as_real
  - as_strided
  - asin
  - asinh
  - assign
  - atan
  - atan2
  - atan_
  - atanh
  - atleast_1d
  - atleast_2d
  - atleast_3d
  - batch
  - bernoulli
  - bernoulli_
  - bincount
  - binomial
  - bitwise_and
  - bitwise_and_
  - bitwise_left_shift
  - bitwise_left_shift_
  - bitwise_not
  - bitwise_not_
  - bitwise_or
  - bitwise_or_
  - bitwise_right_shift
  - bitwise_right_shift_
  - bitwise_xor
  - bitwise_xor_
  - block_diag
  - bmm
  - broadcast_shape
  - broadcast_tensors
  - broadcast_to
  - bucketize
  - cast
  - cast_
  - cauchy_
  - cdist
  - ceil
  - check_shape
  - chunk
  - clip
  - clone
  - column_stack
  - combinations
  - complex
  - concat
  - conj
  - copysign
  - copysign_
  - cos
  - cos_
  - cosh
  - count_nonzero
  - CPUPlace
  - create_parameter
  - crop
  - cross
  - CUDAPinnedPlace
  - CUDAPlace
  - cummax
  - cummin
  - cumprod
  - cumprod_
  - cumsum
  - cumsum_
  - cumulative_trapezoid
  - DataParallel
  - deg2rad
  - diag
  - diag_embed
  - diagflat
  - diagonal
  - diagonal_scatter
  - diff
  - digamma
  - digamma_
  - disable_signal_handler
  - disable_static
  - dist
  - divide
  - divide_
  - dot
  - dsplit
  - dstack
  - dtype
  - einsum
  - empty
  - empty_like
  - enable_grad
  - enable_static
  - equal
  - equal_
  - equal_all
  - erf
  - erf_
  - erfinv
  - exp
  - expand
  - expand_as
  - expm1
  - expm1_
  - eye
  - finfo
  - flatten
  - flatten_
  - flip
  - floor
  - floor_divide
  - floor_divide_
  - floor_mod
  - floor_mod_
  - flops
  - fmax
  - fmin
  - frac
  - frac_
  - frexp
  - full
  - full_like
  - gammainc
  - gammainc_
  - gammaincc
  - gammaincc_
  - gammaln
  - gammaln_
  - gather
  - gather_nd
  - gcd
  - gcd_
  - geometric_
  - get_cuda_rng_state
  - get_default_dtype
  - get_flags
  - get_rng_state
  - grad
  - greater_equal
  - greater_equal_
  - greater_than
  - greater_than_
  - heaviside
  - histogram
  - histogramdd
  - hsplit
  - hstack
  - hypot
  - hypot_
  - i0
  - i0_
  - i0e
  - i1
  - i1e
  - iinfo
  - imag
  - in_dynamic_mode
  - increment
  - index_add
  - index_add_
  - index_fill
  - index_fill_
  - index_put
  - index_put_
  - index_sample
  - index_select
  - inner
  - is_complex
  - is_empty
  - is_floating_point
  - is_grad_enabled
  - is_integer
  - is_tensor
  - isclose
  - isfinite
  - isin
  - isinf
  - isnan
  - isneginf
  - isposinf
  - isreal
  - kron
  - kthvalue
  - LazyGuard
  - lcm
  - lcm_
  - ldexp
  - ldexp_
  - lerp
  - less_equal
  - less_equal_
  - less_than
  - less_than_
  - lgamma
  - lgamma_
  - linspace
  - load
  - log
  - log10
  - log10_
  - log1p
  - log2
  - log2_
  - log_
  - logaddexp
  - logcumsumexp
  - logical_and
  - logical_and_
  - logical_not
  - logical_not_
  - logical_or
  - logical_or_
  - logical_xor
  - logit
  - logit_
  - logspace
  - logsumexp
  - masked_fill
  - masked_fill_
  - masked_scatter
  - masked_scatter_
  - masked_select
  - matmul
  - max
  - maximum
  - mean
  - median
  - meshgrid
  - min
  - minimum
  - mm
  - mode
  - Model
  - moveaxis
  - multigammaln
  - multigammaln_
  - multinomial
  - multiplex
  - multiply
  - multiply_
  - mv
  - nan_to_num
  - nan_to_num_
  - nanmean
  - nanmedian
  - nanquantile
  - nansum
  - neg
  - neg_
  - nextafter
  - no_grad
  - nonzero
  - normal
  - normal_
  - not_equal
  - numel
  - ones
  - ones_like
  - outer
  - ParamAttr
  - pdist
  - poisson
  - polar
  - polygamma
  - polygamma_
  - pow
  - pow_
  - prod
  - put_along_axis
  - quantile
  - rad2deg
  - rand
  - randint
  - randint_like
  - randn
  - randperm
  - rank
  - real
  - reciprocal
  - reduce_as
  - renorm
  - renorm_
  - repeat_interleave
  - reshape
  - reshape_
  - roll
  - rot90
  - round
  - row_stack
  - rsqrt
  - save
  - scale
  - scatter
  - scatter_
  - scatter_nd
  - scatter_nd_add
  - searchsorted
  - seed
  - select_scatter
  - set_cuda_rng_state
  - set_default_dtype
  - set_flags
  - set_grad_enabled
  - set_printoptions
  - set_rng_state
  - sgn
  - shape
  - shard_index
  - sign
  - signbit
  - sin
  - sin_
  - sinc
  - sinc_
  - sinh
  - sinh_
  - slice
  - slice_scatter
  - sort
  - split
  - sqrt
  - square
  - square_
  - squeeze
  - squeeze_
  - stack
  - standard_gamma
  - standard_normal
  - stanh
  - std
  - strided_slice
  - subtract
  - sum
  - summary
  - t
  - t_
  - take
  - take_along_axis
  - tan
  - tan_
  - tanh
  - tanh_
  - Tensor
  - tensor_split
  - tensordot
  - tile
  - to_tensor
  - tolist
  - topk
  - trace
  - transpose
  - transpose_
  - trapezoid
  - tril
  - tril_
  - tril_indices
  - triu
  - triu_
  - triu_indices
  - trunc
  - trunc_
  - unbind
  - unflatten
  - unfold
  - uniform
  - unique
  - unique_consecutive
  - unsqueeze
  - unsqueeze_
  - unstack
  - vander
  - var
  - view
  - view_as
  - vsplit
  - vstack
  - where
  - where_
  - zeros
  - zeros_like
- paddle.amp
- paddle.audio
  - backends
  - datasets
    - ESC50
    - TESS
  - features
  - functional
  - info
  - load
  - save
- paddle.autograd
- paddle.callbacks
- paddle.device
- paddle.distributed
- paddle.distribution
- paddle.fft
  - fft
  - fft2
  - fftfreq
  - fftn
  - fftshift
  - hfft
  - hfft2
  - hfftn
  - ifft
  - ifft2
  - ifftn
  - ifftshift
  - ihfft
  - ihfft2
  - ihfftn
  - irfft
  - irfft2
  - irfftn
  - rfft
  - rfft2
  - rfftfreq
  - rfftn
- paddle.geometric
- paddle.hub
  - help
  - list
  - load
- paddle.incubate
- paddle.inference
- paddle.io
- paddle.jit
- paddle.linalg
  - cholesky
  - cholesky_solve
  - cond
  - corrcoef
  - cov
  - det
  - eig
  - eigh
  - eigvals
  - eigvalsh
  - householder_product
  - inv
  - lstsq
  - lu
  - lu_unpack
  - matrix_exp
  - matrix_norm
  - matrix_power
  - matrix_rank
  - multi_dot
  - norm
  - ormqr
  - pca_lowrank
  - pinv
  - qr
  - slogdet
  - solve
  - svd
  - svd_lowrank
  - triangular_solve
  - vector_norm
- paddle.metric
  - Accuracy
  - accuracy
  - Auc
  - Metric
  - Precision
  - Recall
- paddle.nn
- paddle.onnx
  - export
- paddle.optimizer
  - Adadelta
  - Adagrad
  - Adam
  - Adamax
  - AdamW
  - ASGD
  - Lamb
  - LBFGS
  - lr
  - Momentum
  - NAdam
  - Optimizer
  - RAdam
  - RMSProp
  - Rprop
  - SGD
- paddle.profiler
- paddle.quantization
  - BaseObserver
  - BaseQuanter
  - PTQ
  - QAT
  - QuantConfig
  - quanter
- paddle.regularizer
  - L1Decay
  - L2Decay
- paddle.signal
  - istft
  - stft
- paddle.sparse
  - abs
  - add
  - addmm
  - asin
  - asinh
  - atan
  - atanh
  - cast
  - coalesce
  - deg2rad
  - divide
  - expm1
  - is_same_shape
  - isnan
  - log1p
  - masked_matmul
  - matmul
  - multiply
  - mv
  - neg
  - nn
    - BatchNorm
    - Conv2D
    - Conv3D
    - functional
      - attention
      - conv2d
      - conv3d
      - leaky_relu
      - max_pool3d
      - relu
      - relu6
      - softmax
      - subm_conv2d
      - subm_conv2d_igemm
      - subm_conv3d
      - subm_conv3d_igemm
    - LeakyReLU
    - MaxPool3D
    - ReLU
    - ReLU6
    - Softmax
    - SubmConv2D
    - SubmConv3D
    - SyncBatchNorm
  - pca_lowrank
  - pow
  - rad2deg
  - reshape
  - sin
  - sinh
  - slice
  - sparse_coo_tensor
  - sparse_csr_tensor
  - sqrt
  - square
  - subtract
  - sum
  - tan
  - tanh
  - transpose
- paddle.static
- paddle.sysconfig
  - get_include
  - get_lib
- paddle.Tensor
  - Overview
  - acosh_
  - add_
  - apply
  - apply_
  - asin_
  - asinh_
  - astype
  - atanh_
  - backward
  - ceil_
  - clear_grad
  - clip_
  - coalesce
  - cosh_
  - cpu
  - create_tensor
  - cuda
  - dim
  - erfinv_
  - exp_
  - exponential_
  - fill_
  - fill_diagonal_
  - fill_diagonal_tensor
  - fill_diagonal_tensor_
  - floor_
  - gradient
  - item
  - lerp_
  - log1p_
  - logical_xor_
  - ndimension
  - not_equal_
  - pin_memory
  - put_along_axis_
  - reciprocal_
  - register_hook
  - round_
  - rsqrt_
  - scale_
  - set_value
  - sigmoid
  - sigmoid_
  - sqrt_
  - subtract_
  - to
  - to_dense
  - to_sparse_coo
  - top_p_sampling
  - uniform_
  - value
  - values
  - zero_
- paddle.text
  - Conll05st
  - Imdb
  - Imikolov
  - Movielens
  - UCIHousing
  - viterbi_decode
  - ViterbiDecoder
  - WMT14
  - WMT16
- paddle.utils
  - cpp_extension
  - deprecated
  - dlpack
    - from_dlpack
    - to_dlpack
  - download
    - get_weights_path_from_url
  - require_version
  - run_check
  - try_import
  - unique_name
    - generate
    - guard
    - switch
- paddle.version
  - cuda
  - cudnn
  - nccl
  - show
  - xpu
  - xpu_xccl
  - xpu_xhpc
- paddle.vision
  - datasets
    - Cifar10
    - Cifar100
    - DatasetFolder
    - FashionMNIST
    - Flowers
    - ImageFolder
    - MNIST
    - VOC2012
  - get_image_backend
  - image_load
  - models
  - ops
    - box_coder
    - decode_jpeg
    - deform_conv2d
    - DeformConv2D
    - distribute_fpn_proposals
    - generate_proposals
    - matrix_nms
    - nms
    - prior_box
    - psroi_pool
    - PSRoIPool
    - read_file
    - roi_align
    - roi_pool
    - RoIAlign
    - RoIPool
    - yolo_box
    - yolo_loss
  - set_image_backend
  - transforms
    - adjust_brightness
    - adjust_contrast
    - adjust_hue
    - affine
    - BaseTransform
    - BrightnessTransform
    - center_crop
    - CenterCrop
    - ColorJitter
    - Compose
    - ContrastTransform
    - crop
    - erase
    - Grayscale
    - hflip
    - HueTransform
    - Normalize
    - normalize
    - pad
    - Pad
    - perspective
    - RandomAffine
    - RandomCrop
    - RandomErasing
    - RandomHorizontalFlip
    - RandomPerspective
    - RandomResizedCrop
    - RandomRotation
    - RandomVerticalFlip
    - Resize
    - resize
    - rotate
    - SaturationTransform
    - to_grayscale
    - to_tensor
    - ToTensor
    - Transpose
    - vflip
Contribution Guidelines
- Custom Device Support
3.0 Beta Release Note

PrecisionType

»
PrecisionType
Edit on GitHub

PrecisionType¶

paddle.inference. PrecisionType: alias of paddle.base.libpaddle.AnalysisConfig.Precision