From 6fd1b4d2929d2a70a20bb3af6d91480242651233 Mon Sep 17 00:00:00 2001 From: Le-soleile <3516093767@qq.com> Date: Tue, 27 Jan 2026 15:37:31 +0800 Subject: [PATCH] update doc --- doc/tensor_body.md | 428 +++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 418 insertions(+), 10 deletions(-) diff --git a/doc/tensor_body.md b/doc/tensor_body.md index 6041264..16ab283 100644 --- a/doc/tensor_body.md +++ b/doc/tensor_body.md @@ -1,16 +1,424 @@ +##### tensor_body.h 头文件 API 兼容性 -##### tensor_body.h头文件API兼容情况 +**状态说明：** +- ✅ 已支持 +- 🚧 正在开发 +- ⏳ 尚未开发 +- ❌ 不准备支持 +**按照首字母进行排序** -##### tensor_body.h 头文件 API 兼容性 +| torch API | paddle API 兼容性 | 测试用例状态 | 优先级 | 备注 | +|-----------------------------|----------------|--------|-----|--------------------------------------| +| `abs` | ✅ | ✅ | P2 | | +| `absolute` | ✅ | ✅ | P2 | | +| `acos` | ⏳ | ⏳ | P2 | | +| `acosh` | ⏳ | ⏳ | P2 | | +| `add` | ⏳ | ⏳ | P2 | | +| `addbmm` | ⏳ | ⏳ | P2 | | +| `addcdiv` | ⏳ | ⏳ | P2 | | +| `addcmul` | ⏳ | ⏳ | P2 | | +| `addmm` | ⏳ | ⏳ | P2 | | +| `addmv` | ⏳ | ⏳ | P2 | | +| `addr` | ⏳ | ⏳ | P2 | | +| `adjoint` | ⏳ | ⏳ | P2 | | +| `alias` | ⏳ | ⏳ | P2 | | +| `align_as` | ⏳ | ⏳ | P2 | | +| `align_to` | ⏳ | ⏳ | P2 | | +| `all` | ✅ | ✅ | P2 | | +| `allclose` | ✅ | ✅ | P2 | | +| `amax` | ⏳ | ⏳ | P2 | | +| `amin` | ⏳ | ⏳ | P2 | | +| `angle` | ⏳ | ⏳ | P2 | | +| `any` | 🚧 | 🚧 | P2 | | +| `arccos` | ⏳ | ⏳ | P2 | | +| `arccosh` | ⏳ | ⏳ | P2 | | +| `arcsin` | ⏳ | ⏳ | P2 | | +| `arcsinh` | ⏳ | ⏳ | P2 | | +| `arctan` | ⏳ | ⏳ | P2 | | +| `arctan2` | ⏳ | ⏳ | P2 | | +| `arctanh` | ⏳ | ⏳ | P2 | | +| `argmax` | ⏳ | ⏳ | P2 | | +| `argmin` | ⏳ | ⏳ | P2 | | +| `argsort` | ⏳ | ⏳ | P2 | | +| `argwhere` | ⏳ | ⏳ | P2 | | +| `as_strided` | ✅ | ✅ | P2 | | +| `as_strided_` | ✅ | ✅ | P2 | | +| `as_strided__symint` | ⏳ | ⏳ | P2 | | +| `as_strided_scatter` | ✅ | ✅ | P2 | | +| `as_strided_scatter_symint` | ⏳ | ⏳ | P2 | | +| `as_strided_symint` | ⏳ | ⏳ | P2 | | +| `asin` | ⏳ | ⏳ | P2 | | +| `asinh` | ⏳ | ⏳ | P2 | | +| `atan` | ⏳ | ⏳ | P2 | | +| `atan2` | ⏳ | ⏳ | P2 | | +| `atanh` | ⏳ | ⏳ | P2 | | +| `backward` | ⏳ | ⏳ | P2 | | +| `baddbmm` | ⏳ | ⏳ | P2 | | +| `bernoulli` | ⏳ | ⏳ | P2 | | +| `bincount` | ⏳ | ⏳ | P2 | | +| `bincount_symint` | ⏳ | ⏳ | P2 | | +| `bitwise_and` | ⏳ | ⏳ | P2 | | +| `bitwise_left_shift` | ⏳ | ⏳ | P2 | | +| `bitwise_not` | ⏳ | ⏳ | P2 | | +| `bitwise_or` | ⏳ | ⏳ | P2 | | +| `bitwise_right_shift` | ✅ | ✅ | P2 | | +| `bitwise_xor` | ⏳ | ⏳ | P2 | | +| `bmm` | ⏳ | ⏳ | P2 | | +| `broadcast_to` | ⏳ | ⏳ | P2 | | +| `broadcast_to_symint` | ⏳ | ⏳ | P2 | | +| `ccol_indices` | ⏳ | ⏳ | P2 | | +| `ceil` | ⏳ | ⏳ | P2 | | +| `chalf` | ⏳ | ⏳ | P2 | | +| `cholesky` | ⏳ | ⏳ | P2 | | +| `cholesky_inverse` | ⏳ | ⏳ | P2 | | +| `cholesky_solve` | ⏳ | ⏳ | P2 | | +| `clamp` | ✅ | ✅ | P2 | 已实现 Scalar 和 Tensor 版本 | +| `clamp_max` | ✅ | ✅ | P2 | 已实现 Scalar 和 Tensor 版本 | +| `clamp_min` | ✅ | ✅ | P2 | 已实现 Scalar 和 Tensor 版本 | +| `clip` | ⏳ | ⏳ | P2 | | +| `clone` | ✅ | ✅ | P2 | | +| `coalesce` | ⏳ | ⏳ | P2 | | +| `col_indices` | ⏳ | ⏳ | P2 | | +| `conj` | ⏳ | ⏳ | P2 | | +| `conj_physical` | ⏳ | ⏳ | P2 | | +| `contiguous` | ✅ | ✅ | P2 | | +| `copysign` | ⏳ | ⏳ | P2 | | +| `corrcoef` | ⏳ | ⏳ | P2 | | +| `cos` | ⏳ | ⏳ | P2 | | +| `cosh` | ⏳ | ⏳ | P2 | | +| `count_nonzero` | ⏳ | ⏳ | P2 | | +| `cov` | ⏳ | ⏳ | P2 | | +| `cpu` | ✅ | ✅ | P1 | | +| `cross` | ⏳ | ⏳ | P2 | | +| `crow_indices` | ⏳ | ⏳ | P2 | | +| `cuda` | ✅ | ✅ | P2 | | +| `cumprod` | ⏳ | ⏳ | P2 | | +| `cumsum` | ⏳ | ⏳ | P2 | | +| `data` | ✅ | ✅ | P1 | torch准备遗弃该接口，建议使用 data_ptr() | +| `deg2rad` | ⏳ | ⏳ | P2 | | +| `dense_dim` | ⏳ | ⏳ | P2 | | +| `dequantize` | ⏳ | ⏳ | P2 | | +| `det` | ⏳ | ⏳ | P2 | | +| `detach` | ⏳ | ⏳ | P2 | | +| `diag` | ⏳ | ⏳ | P2 | | +| `diag_embed` | ⏳ | ⏳ | P2 | | +| `diagflat` | ⏳ | ⏳ | P2 | | +| `diagonal` | ⏳ | ⏳ | P2 | | +| `diagonal_scatter` | ⏳ | ⏳ | P2 | | +| `diff` | ⏳ | ⏳ | P2 | | +| `digamma` | ⏳ | ⏳ | P2 | | +| `dist` | ⏳ | ⏳ | P2 | | +| `div` | ⏳ | ⏳ | P2 | | +| `divide` | ⏳ | ⏳ | P2 | | +| `dot` | ⏳ | ⏳ | P2 | | +| `eq` | ⏳ | ⏳ | P2 | | +| `equal` | ⏳ | ⏳ | P2 | | +| `erf` | ⏳ | ⏳ | P2 | | +| `erfc` | ⏳ | ⏳ | P2 | | +| `erfinv` | ⏳ | ⏳ | P2 | | +| `exp` | ⏳ | ⏳ | P2 | | +| `exp2` | ⏳ | ⏳ | P2 | | +| `expand` | 🚧 | 🚧 | P2 | | +| `expand_as` | 🚧 | 🚧 | P2 | | +| `expand_symint` | ⏳ | ⏳ | P2 | | +| `expm1` | ⏳ | ⏳ | P2 | | +| `fix` | ⏳ | ⏳ | P2 | | +| `flatten` | ⏳ | ⏳ | P2 | | +| `flip` | ⏳ | ⏳ | P2 | | +| `fliplr` | ⏳ | ⏳ | P2 | | +| `flipud` | ⏳ | ⏳ | P2 | | +| `float_power` | ⏳ | ⏳ | P2 | | +| `floor` | ⏳ | ⏳ | P2 | | +| `floor_divide` | ⏳ | ⏳ | P2 | | +| `fmax` | ⏳ | ⏳ | P2 | | +| `fmin` | ⏳ | ⏳ | P2 | | +| `fmod` | ⏳ | ⏳ | P2 | | +| `frac` | ⏳ | ⏳ | P2 | | +| `gather` | ⏳ | ⏳ | P2 | | +| `gcd` | ⏳ | ⏳ | P2 | | +| `ge` | ⏳ | ⏳ | P2 | | +| `ger` | ⏳ | ⏳ | P2 | | +| `greater` | ⏳ | ⏳ | P2 | | +| `greater_equal` | ⏳ | ⏳ | P2 | | +| `gt` | ⏳ | ⏳ | P2 | | +| `hardshrink` | ⏳ | ⏳ | P2 | | +| `hardshrink_backward` | ⏳ | ⏳ | P2 | | +| `hash_tensor` | ⏳ | ⏳ | P2 | | +| `heaviside` | ⏳ | ⏳ | P2 | | +| `hip` | ⏳ | ⏳ | P2 | | +| `histc` | ⏳ | ⏳ | P2 | | +| `hypot` | ⏳ | ⏳ | P2 | | +| `i0` | ⏳ | ⏳ | P2 | | +| `igamma` | ⏳ | ⏳ | P2 | | +| `igammac` | ⏳ | ⏳ | P2 | | +| `index` | ✅ | ✅ | P2 | | +| `index_add` | ⏳ | ⏳ | P2 | | +| `index_copy` | ⏳ | ⏳ | P2 | | +| `index_fill` | ⏳ | ⏳ | P2 | | +| `index_put` | ⏳ | ⏳ | P2 | | +| `index_reduce` | ⏳ | ⏳ | P2 | | +| `index_select` | ✅ | ⏳ | P2 | | +| `indices` | ⏳ | ⏳ | P2 | | +| `inner` | ⏳ | ⏳ | P2 | | +| `int_repr` | ⏳ | ⏳ | P2 | | +| `inverse` | ⏳ | ⏳ | P2 | | +| `is_coalesced` | ⏳ | ⏳ | P2 | | +| `is_distributed` | ⏳ | ⏳ | P2 | | +| `is_nonzero` | ⏳ | ⏳ | P2 | | +| `is_pinned` | ✅ | ✅ | P2 | | +| `is_same_size` | ⏳ | ⏳ | P2 | | +| `is_set_to` | ⏳ | ⏳ | P2 | | +| `is_variable` | ❌ | ❌ | P2 | torch准备遗弃该接口，现在所有 tensor 都是 variable | +| `isclose` | ⏳ | ⏳ | P2 | | +| `isfinite` | ⏳ | ⏳ | P2 | | +| `isinf` | ⏳ | ⏳ | P2 | | +| `isnan` | ⏳ | ⏳ | P2 | | +| `isneginf` | ⏳ | ⏳ | P2 | | +| `isposinf` | ⏳ | ⏳ | P2 | | +| `isreal` | ⏳ | ⏳ | P2 | | +| `istft` | ⏳ | ⏳ | P2 | | +| `item` | ✅ | ✅ | P2 | | +| `kron` | ⏳ | ⏳ | P2 | | +| `lcm` | ⏳ | ⏳ | P2 | | +| `ldexp` | ⏳ | ⏳ | P2 | | +| `le` | ⏳ | ⏳ | P2 | | +| `lerp` | ⏳ | ⏳ | P2 | | +| `less` | ⏳ | ⏳ | P2 | | +| `less_equal` | ⏳ | ⏳ | P2 | | +| `lgamma` | ⏳ | ⏳ | P2 | | +| `log` | ⏳ | ⏳ | P2 | | +| `log10` | ⏳ | ⏳ | P2 | | +| `log1p` | ⏳ | ⏳ | P2 | | +| `log2` | ⏳ | ⏳ | P2 | | +| `log_softmax` | ⏳ | ⏳ | P2 | | +| `logaddexp` | ⏳ | ⏳ | P2 | | +| `logaddexp2` | ⏳ | ⏳ | P2 | | +| `logcumsumexp` | ⏳ | ⏳ | P2 | | +| `logdet` | ⏳ | ⏳ | P2 | | +| `logical_and` | ⏳ | ⏳ | P2 | | +| `logical_not` | ⏳ | ⏳ | P2 | | +| `logical_or` | ⏳ | ⏳ | P2 | | +| `logical_xor` | ⏳ | ⏳ | P2 | | +| `logit` | ⏳ | ⏳ | P2 | | +| `logsumexp` | ⏳ | ⏳ | P2 | | +| `lt` | ⏳ | ⏳ | P2 | | +| `lu_solve` | ⏳ | ⏳ | P2 | | +| `mH` | ⏳ | ⏳ | P2 | | +| `mT` | ⏳ | ⏳ | P2 | | +| `masked_fill` | ⏳ | ⏳ | P2 | | +| `masked_scatter` | ⏳ | ⏳ | P2 | | +| `masked_select` | ⏳ | ⏳ | P2 | | +| `matmul` | ⏳ | ⏳ | P2 | | +| `matrix_H` | ⏳ | ⏳ | P2 | | +| `matrix_exp` | ⏳ | ⏳ | P2 | | +| `matrix_power` | ⏳ | ⏳ | P2 | | +| `max` | ⏳ | ⏳ | P2 | | +| `maximum` | ⏳ | ⏳ | P2 | | +| `mean` | ⏳ | ⏳ | P2 | | +| `median` | ⏳ | ⏳ | P2 | | +| `meta` | ✅ | ✅ | P1 | | +| `metal` | ⏳ | ⏳ | P2 | | +| `min` | ⏳ | ⏳ | P2 | | +| `minimum` | ⏳ | ⏳ | P2 | | +| `mm` | ⏳ | ⏳ | P2 | | +| `moveaxis` | ⏳ | ⏳ | P2 | | +| `movedim` | ⏳ | ⏳ | P2 | | +| `msort` | ⏳ | ⏳ | P2 | | +| `mul` | ⏳ | ⏳ | P2 | | +| `multinomial` | ⏳ | ⏳ | P2 | | +| `multinomial_symint` | ⏳ | ⏳ | P2 | | +| `multiply` | ⏳ | ⏳ | P2 | | +| `mutable_grad` | ❌ | ❌ | P2 | | +| `mv` | ⏳ | ⏳ | P2 | | +| `mvlgamma` | ⏳ | ⏳ | P2 | | +| `nan_to_num` | ⏳ | ⏳ | P2 | | +| `nanmean` | ⏳ | ⏳ | P2 | | +| `nanmedian` | ⏳ | ⏳ | P2 | | +| `nanquantile` | ⏳ | ⏳ | P2 | | +| `nansum` | ⏳ | ⏳ | P2 | | +| `narrow` | ⏳ | ⏳ | P2 | | +| `narrow_copy` | ⏳ | ⏳ | P2 | | +| `narrow_copy_symint` | ⏳ | ⏳ | P2 | | +| `narrow_symint` | ⏳ | ⏳ | P2 | | +| `ne` | ⏳ | ⏳ | P2 | | +| `neg` | ⏳ | ⏳ | P2 | | +| `negative` | ⏳ | ⏳ | P2 | | +| `new_empty` | 🚧 | 🚧 | P2 | | +| `new_empty_strided` | ⏳ | ⏳ | P2 | | +| `new_empty_strided_symint` | ⏳ | ⏳ | P2 | | +| `new_empty_symint` | ⏳ | ⏳ | P2 | | +| `new_full` | 🚧 | 🚧 | P2 | | +| `new_full_symint` | ⏳ | ⏳ | P2 | | +| `new_ones` | 🚧 | 🚧 | P2 | | +| `new_ones_symint` | ⏳ | ⏳ | P2 | | +| `new_zeros` | 🚧 | 🚧 | P2 | | +| `new_zeros_symint` | ⏳ | ⏳ | P2 | | +| `nextafter` | ⏳ | ⏳ | P2 | | +| `nonzero` | ⏳ | ⏳ | P2 | | +| `nonzero_static` | ⏳ | ⏳ | P2 | | +| `nonzero_static_symint` | ⏳ | ⏳ | P2 | | +| `norm` | ⏳ | ⏳ | P2 | | +| `not_equal` | ⏳ | ⏳ | P2 | | +| `numpy_T` | ⏳ | ⏳ | P2 | | +| `orgqr` | ⏳ | ⏳ | P2 | | +| `ormqr` | ⏳ | ⏳ | P2 | | +| `outer` | ⏳ | ⏳ | P2 | | +| `permute` | ✅ | ✅ | P2 | | +| `pin_memory` | ✅ | ✅ | P2 | | +| `pinverse` | ⏳ | ⏳ | P2 | | +| `polygamma` | ⏳ | ⏳ | P2 | | +| `positive` | ⏳ | ⏳ | P2 | | +| `pow` | ⏳ | ⏳ | P2 | | +| `prelu` | ⏳ | ⏳ | P2 | | +| `prod` | ⏳ | ⏳ | P2 | | +| `put` | ⏳ | ⏳ | P2 | | +| `q_per_channel_axis` | ⏳ | ⏳ | P2 | | +| `q_per_channel_scales` | ⏳ | ⏳ | P2 | | +| `q_per_channel_zero_points` | ⏳ | ⏳ | P2 | | +| `q_scale` | ⏳ | ⏳ | P2 | | +| `q_zero_point` | ⏳ | ⏳ | P2 | | +| `quantile` | ⏳ | ⏳ | P2 | | +| `rad2deg` | ⏳ | ⏳ | P2 | | +| `ravel` | ⏳ | ⏳ | P2 | | +| `reciprocal` | ⏳ | ⏳ | P2 | | +| `record_stream` | ✅ | ✅ | P2 | 仅 CUDA 版本 | +| `refine_names` | ⏳ | ⏳ | P2 | | +| `relu` | ⏳ | ⏳ | P2 | | +| `remainder` | ⏳ | ⏳ | P2 | | +| `rename` | 🚧 | 🚧 | P2 | | +| `renorm` | ⏳ | ⏳ | P2 | | +| `repeat` | ⏳ | ⏳ | P2 | | +| `repeat_interleave` | ⏳ | ⏳ | P2 | | +| `repeat_interleave_symint` | ⏳ | ⏳ | P2 | | +| `repeat_symint` | ⏳ | ⏳ | P2 | | +| `reshape` | ✅ | ✅ | P2 | | +| `reshape_as` | ⏳ | ⏳ | P2 | | +| `reshape_symint` | ⏳ | ⏳ | P2 | | +| `resize_` | 🚧 | 🚧 | P2 | | +| `resize__symint` | ⏳ | ⏳ | P2 | | +| `resize_as_` | ⏳ | ⏳ | P2 | | +| `resize_as_sparse_` | ⏳ | ⏳ | P2 | | +| `resolve_conj` | ⏳ | ⏳ | P2 | | +| `resolve_neg` | ⏳ | ⏳ | P2 | | +| `roll` | ⏳ | ⏳ | P2 | | +| `roll_symint` | ⏳ | ⏳ | P2 | | +| `rot90` | ⏳ | ⏳ | P2 | | +| `round` | ⏳ | ⏳ | P2 | | +| `row_indices` | ⏳ | ⏳ | P2 | | +| `rsqrt` | ⏳ | ⏳ | P2 | | +| `scatter` | ⏳ | ⏳ | P2 | | +| `scatter_add` | ⏳ | ⏳ | P2 | | +| `scatter_reduce` | ⏳ | ⏳ | P2 | | +| `select` | ⏳ | ⏳ | P2 | | +| `select_scatter` | ⏳ | ⏳ | P2 | | +| `select_scatter_symint` | ⏳ | ⏳ | P2 | | +| `select_symint` | ⏳ | ⏳ | P2 | | +| `sgn` | ⏳ | ⏳ | P2 | | +| `sigmoid` | ⏳ | ⏳ | P2 | | +| `sign` | ⏳ | ⏳ | P2 | | +| `signbit` | ⏳ | ⏳ | P2 | | +| `sin` | ⏳ | ⏳ | P2 | | +| `sinc` | ⏳ | ⏳ | P2 | | +| `sinh` | ⏳ | ⏳ | P2 | | +| `size` | ⏳ | ⏳ | P2 | | +| `slice` | ✅ | ✅ | P2 | | +| `slice_inverse` | ⏳ | ⏳ | P2 | | +| `slice_inverse_symint` | ⏳ | ⏳ | P2 | | +| `slice_scatter` | ⏳ | ⏳ | P2 | | +| `slice_scatter_symint` | ⏳ | ⏳ | P2 | | +| `slice_symint` | ⏳ | ⏳ | P2 | | +| `smm` | ⏳ | ⏳ | P2 | | +| `softmax` | ⏳ | ⏳ | P2 | | +| `sparse_dim` | ⏳ | ⏳ | P2 | | +| `sparse_mask` | ⏳ | ⏳ | P2 | | +| `sparse_resize_` | ⏳ | ⏳ | P2 | | +| `sparse_resize_and_clear_` | ⏳ | ⏳ | P2 | | +| `sqrt` | ⏳ | ⏳ | P2 | | +| `square` | ⏳ | ⏳ | P2 | | +| `squeeze` | ✅ | ✅ | P2 | | +| `sspaddmm` | ⏳ | ⏳ | P2 | | +| `std` | ✅ | ✅ | P2 | | +| `stft` | ⏳ | ⏳ | P2 | | +| `stride` | ⏳ | ⏳ | P2 | | +| `sub` | ⏳ | ⏳ | P2 | | +| `subtract` | ⏳ | ⏳ | P2 | | +| `sum` | ⏳ | ⏳ | P2 | | +| `sum_to_size` | ⏳ | ⏳ | P2 | | +| `sum_to_size_symint` | ⏳ | ⏳ | P2 | | +| `swapaxes` | ⏳ | ⏳ | P2 | | +| `swapdims` | ⏳ | ⏳ | P2 | | +| `t` | ⏳ | ⏳ | P2 | | +| `take` | ⏳ | ⏳ | P2 | | +| `take_along_dim` | ⏳ | ⏳ | P2 | | +| `tan` | ⏳ | ⏳ | P2 | | +| `tanh` | ⏳ | ⏳ | P2 | | +| `tensor_data` | ✅ | ✅ | P2 | | +| `tile` | ⏳ | ⏳ | P2 | | +| `tile_symint` | ⏳ | ⏳ | P2 | | +| `to` | ✅ | ✅ | P1 | | +| `toBackend` | ✅ | ✅ | P2 | torch准备遗弃该接口（TODO标记 | +| `toType` | ✅ | ✅ | P1 | | +| `to_dense` | ⏳ | ⏳ | P2 | | +| `to_mkldnn` | ⏳ | ⏳ | P2 | | +| `to_padded_tensor` | ⏳ | ⏳ | P2 | | +| `to_padded_tensor_symint` | ⏳ | ⏳ | P2 | | +| `to_sparse` | ⏳ | ⏳ | P2 | | +| `to_sparse_bsc` | ⏳ | ⏳ | P2 | | +| `to_sparse_bsr` | ⏳ | ⏳ | P2 | | +| `to_sparse_csc` | ⏳ | ⏳ | P2 | | +| `to_sparse_csr` | ⏳ | ⏳ | P2 | | +| `trace` | ⏳ | ⏳ | P2 | | +| `transpose` | ✅ | ⏳ | P2 | | +| `tril` | ⏳ | ⏳ | P2 | | +| `triu` | ⏳ | ⏳ | P2 | | +| `true_divide` | ⏳ | ⏳ | P2 | | +| `trunc` | ⏳ | ⏳ | P2 | | +| `type` | ❌ | ❌ | P1 | torch准备遗弃该接口 | +| `type_as` | ⏳ | ⏳ | P2 | | +| `unflatten` | ⏳ | ⏳ | P2 | | +| `unflatten_symint` | ⏳ | ⏳ | P2 | | +| `unfold` | ⏳ | ⏳ | P2 | | +| `unsqueeze` | ✅ | ✅ | P2 | | +| `values` | ⏳ | ⏳ | P2 | | +| `var` | ✅ | ✅ | P2 | | +| `variable_data` | ✅ | ✅ | P2 | | +| `vdot` | ⏳ | ⏳ | P2 | | +| `ve` | ⏳ | ⏳ | P2 | | +| `view` | ✅ | ⏳ | P2 | | +| `view_as` | ⏳ | ⏳ | P2 | | +| `view_symint` | ⏳ | ⏳ | P2 | | +| `vulkan` | ⏳ | ⏳ | P2 | | +| `where` | ⏳ | ⏳ | P2 | | +| `xlogy` | ⏳ | ⏳ | P2 | | -✅ 表示已经支持 -❌ 表示不准备支持 +--- -**按照首字母进行排序** +### 兼容性统计 + +| 状态 | 数量 | +|------|------| +| ✅ 已支持 | 37 | +| 🚧 正在开发 | 12 | +| ⏳ 尚未开发 | 329 | +| ❌ 不准备支持 | 3 | +| **总计** | **381** | + +--- + +### 备注 + +1. **优先级说明**： + - P0: 核心功能，必须支持 + - P1: 常用功能，高优先级 + - P2: 进阶功能，中优先级 + - P3: 边缘功能，低优先级 + +2. **实现说明**： + - 所有标记为 ✅ 的接口均已实现并测试通过 + - 部分接口（如 `data`、`toBackend`、`is_variable`）在 torch 中准备遗弃，但为了兼容性仍提供支持 + - `clamp` 系列接口已实现 Scalar 和 Tensor 两种版本 -| torch API | paddle API 兼容性 | 测试用例状态 | 优先级 | 备注 | -|------------------|------------------|------------|-------|------| -| `cpu` | - [ ] | - [ ] | P1 | | -| `toBackend` | ✅ | ✅ | | | -| `type` | ❌ | ❌ | | torch准备遗弃该接口 | +3. **命名空间**： + - 所有接口位于 `at::Tensor` 命名空间下 + - 兼容 PyTorch 的 `ATen` API 接口