o 闦im@sddlZddlmZddlmZmZmZmZmZm Z ddl Z ddl m Z m Z ddlmZmZmZmZmZmZmZmZmZmZgdZeddZGd d d e jZ dd e jd edefddZGdddeZGdddeZ dS)N) namedtuple)AnyCallableDictListOptionalTuple))sparse_semi_structured_from_dense_cutlass'sparse_semi_structured_to_dense_cutlass) fallback_dispatchersemi_sparse_addmmsemi_sparse_detachsemi_sparse_indicessemi_sparse_linearsemi_sparse_mmsemi_sparse_scaled_mm semi_sparse_tsemi_sparse_valuessemi_sparse_view)SparseSemiStructuredTensor!SparseSemiStructuredTensorCUTLASS$SparseSemiStructuredTensorCUSPARSELTto_sparse_semi_structured_SEMI_STRUCTURED_SPARSE_CONFIGz=sparse_min_rows sparse_min_cols dense_min_rows dense_min_colsc@s"eZdZUdZdZeed<eej e fed<dZ e ed<dZ e ed<dZe ed<eed <eeefed <eejed <eejed <eejed <eejed<eejed<e ed<eed<gdZe   d3dejd eejd eejd eejdeejdeejde dede fddZdefddZdeeeeeje ee fffddZedeeje ee fdejfddZejjZede fdd Z!ed4d5d"d#Z"ed$ejdd!fd%d&Z#ed'ejdejfd(d)Z$d*d+Z%ed$ejddfd,d-Z&d!d.d/ejd0eejdejfd1d2Z'd!S)6ra This class implementes semi-structured sparsity as a Tensor subclass. Semi-structured sparsity describes a sparsity pattern where n in every 2n elements are sparse, depending on the datatype. It is also referred to as 2:4 sparsity or fine-grained structured sparsity. There are two backends available for semi_structred sparsity, either cuSPARSELt or CUTLASS. This class is meant to serve as a base class for both implementations. SparseSemiStructuredCUTLASS and SparseSemiStructuredCUSPARSELT both inherit from this class and define three backend-specific items. Note that as such, this class cannot be insantiated directly. -`_DTYPE_SHAPE_CONSTRAINTS` - A dictionary holding backend specific dense/sparse min shape constraints - `def from_dense()` - backend specific compression routines - `def _mm()` - backend specifc mm op (either torch._cslt_sparse_mm or torch._sparse_semi_structured_(mm|addmm)) r_DEFAULT_ALG_ID_DTYPE_SHAPE_CONSTRAINTSF_FORCE_CUTLASS_FUSE_TRANSPOSE_PROTOTYPE_WARNING_SHOWNBACKENDSPARSE_DISPATCHpackedmetapacked_tmeta_tcompressed_swizzled_bitmaskfuse_transpose_cusparseltalg_id_cusparselt)r!r"r#r$r%shape requires_gradc Cs|jstdtd|_|tj||dur|} n |dur$|} ntd| j | j | j | d} tj j ||fi| } || _|| _|| _|| _|| _|| _|| _| S)a0 Create a new instance of the tensor subclass from the compressed sparse representation. We have the option to create the subclass with the compressed representations of both X and X', for training. For inference, we only need a single representation (either X or X'), while the corresponding other set will be None. Depending on the backend selected, certain fields will be set to None. (CUSPARSELT vs CUTLASS) Args: shape: The shape of the original dense tensor packed: The compressed representation of the original dense tensor meta: The metadata of the original dense tensor, if it is stored separately packed_t: The compressed representation of the transposed original dense tensor meta_t: The metadata of the transposed original dense tensor, if it is stored separately compressed_swizzled_bitmask: The masks used by the CUTLASS backend to determine which threads should participate in the computation. Used for pointwise ops. fuse_transpose_cusparselt: When running with cuSPARSELt, we have the option to fuse a transposition with a matmul, which is useful in the case of 2:4 sparse training. alg_id_cusparselt: The algorithm id to use when using cuSPARSELT, will have effect on performance Returns: torch.Tensor: A torch.Tensor wrapper subclass. Raises: ValueError: If all of the tensor arguments are None. zThe PyTorch API of SparseSemiStructuredTensor is in prototype stage and will change in the near future. Please open a Github issue for features requests and see our documentation on the torch.sparse module for further information about the project.TNz3At least one of packed or packed_t must be provided)devicedtypelayoutr))rwarningswarn UserWarning_load_dispatch_tabletorch_dynamoallow_in_graph ValueErrorr*r+r,Tensor_make_wrapper_subclassr!r"r#r$r%r&r') clsr(r!r"r#r$r%r&r'r)previous_tensorkwargstensorr;Z/home/ubuntu/transcripts/venv/lib/python3.10/site-packages/torch/sparse/semi_structured.py__new__Ks6'  z"SparseSemiStructuredTensor.__new__returncCs$t|dsJ|jjd|jdS)Nr(z(shape=))hasattr __class____name__r(selfr;r;r<__repr__sz#SparseSemiStructuredTensor.__repr__cs4ttfddj}jjjjf}||fS)Ncst|duSN)getattr)xrCr;r<sz?SparseSemiStructuredTensor.__tensor_flatten__..)listfilter __slots__r(r&r'r))rD inner_tensors tensor_metar;rCr<__tensor_flatten__sz-SparseSemiStructuredTensor.__tensor_flatten__rNc CsN|\}}}}|||dd|dd|dd|dd|dd|||d S)Nr!r"r#r$r% r(r!r"r#r$r%r&r'r))get) r7rMrN outer_size outer_strider(r&r'r)r;r;r<__tensor_unflatten__s     z/SparseSemiStructuredTensor.__tensor_unflatten__cCs:|j|jvrt|jd|jd|j|j||||S)NzI only supports a specific set of operations, can't perform requested op (r?)_overloadpacketr NotImplementedErrorrB)r7functypesargsr9r;r;r<__torch_dispatch__s z-SparseSemiStructuredTensor.__torch_dispatch__NcCst|dddurXtjjjttjjjttjjjt tjjj t tjjj t tjjj ttjjjttjjjttjjjttjjjttjjjttjjjt tjjjti |_|durZ|j|dSdSdS)zT Loads the op overload sparse dispatch table for the current class. r N)rGr1opsatenvaluesrindicesr is_same_sizer detach_detachr trviewrmmrmatmuladdmmr linearr_to_copy _scaled_mmrr update)r7custom_dispatch_tabler;r;r<r0s&             z/SparseSemiStructuredTensor._load_dispatch_tableoriginal_tensorcCs|js td|jd|dkrtd|d|s$td|j|jvr6td|jd|d |j\}}|j|jj}|j|jj }||ksY||sY||ksY||rhtd |jd |d |d dS)z_ Assert that the given tensor is valid for semi-structured sparse compression. zError original_tensor.device= z= is not supported! Only CUDA tensors are currently supported.zError original_tensor.dim = z; is not supported! Only 2d tensors are currently supported.zXError original_tensor is not contiguous!Only contiguous tensors are currently supported.zError original_tensor.dtype z is not a supported dtype for !zError original_tensor.shape zS is not supported! Both dimensions must be larger or equal than and a multiple of (z, r?N) is_cuda RuntimeErrorr*dim is_contiguousr+rr(sparse_min_rowssparse_min_cols)r7rlmnmin_rowsmin_colsr;r;r< _validate_device_dim_dtype_shapes8      z;SparseSemiStructuredTensor._validate_device_dim_dtype_shape dense_inputcCs|dksJ|j\}}|j|jj}|j|jj}||ks#||r(| |nd}||ks2||r7| |nd}|s=|rItjj |d|d|fS|S)z Calculates padding for dense tensor and pads tensor if necessary. If padding is not required, this function returns the original tensor. rmr) rqr(rr+dense_min_rowsdense_min_colsr1nn functionalpad)r7rzrurvrwrxto_pad_mto_pad_nr;r;r<_pad_dense_inputs z+SparseSemiStructuredTensor._pad_dense_inputcCs&|jd}t|tj||j|jdS)N)r+r*)r(r1rdeyer+r*)rDcolr;r;r<to_dense+s z#SparseSemiStructuredTensor.to_densecCtrFrVr7rlr;r;r< from_dense/sz%SparseSemiStructuredTensor.from_densebiasBrcKrrFr)rDrrr9r;r;r<_mm3szSparseSemiStructuredTensor._mm)FrFrF)r>N)(rB __module__ __qualname____doc__rint__annotations__rr1r+rrboolrrstrrrr5rL staticmethodSizer=rErrrO classmethodrT_C_disabled_torch_function_impl__torch_function__rrZr0ryrrrrr;r;r;r<r&s        R  *rFrl transposedr>cCs4|r tjdtddtjrtjjntjj}| |S)a This function converts a dense tensor into a sparse semi-structured tensor. It will return a SparseSemiStructuredTensor, a subclass of torch.Tensor. This function will check to ensure the dense tensor has the right dtype, size, dims, and device. We currently only support semi-structured sparse tensors for 2d CUDA tensors. Additionally, your tensor must be a positive multiple of the mininum sparse block size, given in `_DTYPE_TO_SHAPE_CONSTRAINTS` for each dtype (float32, float16, bfloat16, int8). Args: original_tensor (Tensor): the dense tensor to convert transposed (bool, optional): deprecated arg to be removed in another release. Do not use. Returns: SparseSemiStructuredTensor: A sparse semi-structured tensor created from the given original_tensor Raises: None Example: >>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA) >>> A = torch.Tensor([0, 0, 1, 1]).tile((128, 32)).half().cuda() tensor([[0., 0., 1., ..., 0., 1., 1.], [0., 0., 1., ..., 0., 1., 1.], [0., 0., 1., ..., 0., 1., 1.], ..., [0., 0., 1., ..., 0., 1., 1.], [0., 0., 1., ..., 0., 1., 1.], [0., 0., 1., ..., 0., 1., 1.]], device='cuda:0', dtype=torch.float16) >>> A_sparse = to_sparse_semi_structured(A) SparseSemiStructuredTensor(shape=torch.Size([128, 128])) >>> A_sparse.values() tensor([[1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], ..., [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.]], device='cuda:0', dtype=torch.float16), >>> A_sparse.indices() tensor([[-4370, -4370, -4370, ..., -4370, -4370, -4370], [-4370, -4370, -4370, ..., -4370, -4370, -4370], [-4370, -4370, -4370, ..., -4370, -4370, -4370], ..., [-4370, -4370, -4370, ..., -4370, -4370, -4370], [-4370, -4370, -4370, ..., -4370, -4370, -4370], [-4370, -4370, -4370, ..., -4370, -4370, -4370]], device='cuda:0', dtype=torch.int16)) zSetting transpose from `to_sparse_semi_structured` is deprecated and will be removed in a future release. `SparseSemiStructuredTensor` only support contiguous input tensors.rm) stacklevel) r-r. FutureWarningrrr1sparserrr)rlrSPARSE_SUBCLASSr;r;r<r=s1  rc seZdZdZdZejeddddejeddddej eddddej eddddiZ e d ej d dfd d Zfd dZe dd ej d dfddZdddej deej d ej fddZZS)ra This class implements semi-structured sparsity for the CUTLASS backend. In this implementation, the specified elements and metadata are stored seprately, in packed and meta respectively. When _FORCE_CUTLASS is set, or when cuSPARSELt is not available, this subclass calls into _sparse_semi_structured_(mm|addmm) and sparse_semi_structured_from_dense for conversion to the compressed format. cutlass @rlr>c Cs0||t|\}}||j||ddd|jdS)Nr!r"r#r$r%r))ryr r(r))r7rlsparse_tensor_cutlassmeta_tensor_cutlassr;r;r<rs z,SparseSemiStructuredTensorCUTLASS.from_densecs<|jdur |jdus J|jjdkrt|j|jStS)Nrm)r"r!ndimr superrrCrAr;r<rs z*SparseSemiStructuredTensorCUTLASS.to_denserc Cs2tj||dd\}}}}}||j|||||ddS)a This function takes in a unpruned dense tensor and runs a (branchless) static sort across a 4x4 tile. It greedily picks the largest values in the tile, upholding the 2:4 sparsity constraint across both rows and columns. The algorithm used to prune the matrix is implemented in `_sparse_semi_structured_tile`. Then it creates the packed and meta tensors for the compressed sparse representation of the pruned dense tensor. It also calculates the packed_t and meta_t tensors for the compressed sparse representation of the transposed pruned dense tensor. Since we cannot transpose the compressed representations, we store both for the fw/bw pass respectively. Finally, this function also computes a compressed swizzled bitmask that encodes the sparsity pattern This can be used in the backward pass to mask the gradients. [9 1 7 4] [9 0 7 0] [1 2 3 0] [0 2 0 0] [8 3 5 4] -> prune 4x4 tile -> [8 0 0 4] -> pack to CUTLASS semi-structured -> packed [1 2 6 2] [0 0 6 2] -> metadata -> pack to transposed CUTLASS -> packed_t semi-structured representation -> metadata_t -> compute swizzled bitmask -> compressed_swizzled_bitmask The equivalent PyTorch code to create the same five outputs from the dense tensor can be found below: ``` from torch.sparse import SparseSemiStructuredTensorCUTLASS from torch.sparse._semi_structured_conversions import _sparse_semi_structured_tile, _compute_compressed_swizzled_bitmask pruned = _sparse_semi_structured_tile(dense) packed_cutlass, meta_cutlass = sparse_semi_structured_from_dense_cutlass(pruned) packed_t_cutlass, meta_t_cutlass = sparse_semi_structured_from_dense_cutlass(pruned.t().contiguous()) bitmask = _compute_compressed_swizzled_bitmask(pruned) SparseSemiStructuredTensorCUTLASS(dense.shape, packed_cutlass, meta_cutlass, packed_t_cutlass, meta_t_cutlass, bitmask) ``` T algorithm use_cutlassFrr1_sparse_semi_structured_tiler(r7rlrr!r"r#r$r%r;r;r<prune_dense_static_sorts$1z9SparseSemiStructuredTensorCUTLASS.prune_dense_static_sortNrrrcKst|tr td|jj}|jdks|jdkrtd|d|jdus)|jdur1td|d|dur?t |j|j|}n t ||j|j|}|d|j dS)NZ`SparseSemiStructuredTensor @ SparseSemiStructuredTensor` is not supported by the hardwarerm`)` matmul: Broadcasting is not implemented$` matmul: operation is not supportedr) isinstancerr4rArBrrVr!r"r1_sparse_semi_structured_mm_sparse_semi_structured_addmmr()rDrrr9cls_nameresr;r;r<rs&    z%SparseSemiStructuredTensorCUTLASS._mmr)rBrrrrr1int8rfloat16bfloat16float32rrr5rrrrr __classcell__r;r;rr<rs<    ?rc @seZdZdZdZejeddddejeddddej eddddej eddddiZ e dej ddfdd Ze ddej dd fd d Zdddej deej dej fddZdS)ra The cuSPARSELt backend expects the specified elements and the metadata to be stored in a single tensor: packed = [ specified elements of original tensor | metadata ] For an original tensor of size (m, k) we expect the first m * k // 2 elements to be the kept elements The rest of the tensor is metadata. Since there is only one tensor, we only use the packed and packed_t attributes respectively. cuSPARSELt also supports transposition fusion, which is necessary for performant 2:4 sparse training, as well as specifying alg_id, a config that affects the performance of the matmul depending on matmul sizes. cusparseltrrrrlr>c Cs2||||jt|ddddtjtj|jd S)NrP)ryr(r1_cslt_compressrrrr)rr;r;r<r s z/SparseSemiStructuredTensorCUSPARSELT.from_denserrc Cs2tj||dd\}}}}}||j|||||ddS)a This function does the same thing as described in SparseSemiStructuredCUTLASS, but uses the cuSPASRELt metadata layout and sparse matmul. The only functional difference is that cuSPARSELt stores `metadata` and `packed` together into a single tensor. [9 1 7 4] [9 0 7 0] [1 2 3 0] [0 2 0 0] [8 3 5 4] -> prune 4x4 tile -> [8 0 0 4] -> pack to cuSPARSELT semi-structured -> packed [1 2 6 2] [0 0 6 2] -> pack to transposed cuSPARSELt -> packed_t semi-structured representation -> compute swizzled bitmask -> compressed_swizzled_bitmask The equivalent PyTorch code to create the same three outputs from the dense tensor can be found below: ``` from torch.sparse import SparseSemiStructuredTensorCUSPARSELT from torch.sparse._semi_structured_conversions import _sparse_semi_structured_tile, _compute_compressed_swizzled_bitmask pruned = _sparse_semi_structured_tile(dense) packed_cusparselt = torch._cslt_compress(pruned) packed_t_cusparselt = torch._cslt_compress(pruned.t().contiguous()) bitmask = _compute_compressed_swizzled_bitmask(pruned) SparseSemiStructuredTensorCUSPARSELT(dense.shape, packed_cutlass, None, packed_t_cutlass, None, bitmask) ``` Frrrrr;r;r<r1s$(zs2  0 D