1 line
34 KiB
JavaScript
1 line
34 KiB
JavaScript
|
Search.setIndex({"docnames": ["contribute", "cutlass", "cutlass.emit", "cutlass.op", "cutlass.utils", "examples", "externals/00_basic_gemm", "externals/01_epilogue", "externals/02_pytorch_extension_grouped_gemm", "index", "install", "modules"], "filenames": ["contribute.md", "cutlass.rst", "cutlass.emit.rst", "cutlass.op.rst", "cutlass.utils.rst", "examples.rst", "externals/00_basic_gemm.nblink", "externals/01_epilogue.nblink", "externals/02_pytorch_extension_grouped_gemm.nblink", "index.rst", "install.md", "modules.rst"], "titles": ["Contributing", "CUTLASS", "Emitters", "Operations", "Utilities", "Examples", "Basic example of using the CUTLASS Python interface", "Example of using elementwise activation functions in the CUTLASS Python interface", "Exporting a CUTLASS grouped GEMM kernel to a PyTorch CUDA extension", "CUTLASS Python Interface", "Installation", "CUTLASS Python API"], "terms": {"thank": 0, "you": [0, 6, 9, 10], "your": 0, "interest": 0, "cutlass": [0, 2, 3, 4, 5, 10], "python": [0, 1, 2, 3, 5, 10], "interfac": [0, 1, 3, 5, 10], "base": [0, 1, 3, 4], "type": [0, 1, 3, 6, 7, 8, 11], "fall": [0, 6], "two": [0, 3, 6], "categori": 0, "want": [0, 6], "report": [0, 6], "bug": 0, "featur": [0, 6], "request": 0, "document": [0, 6], "issu": [0, 6], "file": [0, 2, 8], "an": [0, 1, 4, 5, 6, 9, 10], "describ": 0, "what": [0, 6, 9], "encount": 0, "see": [0, 3, 6, 7, 8], "chang": [0, 3, 5, 9, 10], "The": [0, 2, 3, 6, 8, 9, 10], "team": 0, "evalu": 0, "triag": 0, "them": [0, 7, 9, 10], "schedul": [0, 4, 6], "releas": [0, 9, 10], "If": [0, 2, 3, 6, 7, 9, 10], "believ": 0, "need": [0, 6, 7, 9], "prioriti": 0, "attent": 0, "comment": 0, "notifi": 0, "implement": 0, "fix": 0, "we": [0, 6, 7, 8, 9, 10], "welcom": [0, 9], "from": [0, 2, 3, 6, 7, 8], "commun": [0, 9], "recommend": [0, 9, 10], "via": [0, 1, 2, 3, 5, 6, 7, 9, 10], "pull": 0, "have": [0, 3, 6, 9, 10], "question": [0, 4, 6], "about": 0, "consid": [0, 1, 9], "ask": 0, "discuss": 0, "tab": [0, 9], "pleas": [0, 3, 6, 7, 8], "sure": 0, "search": [0, 9], "through": [0, 6, 7, 8], "both": [0, 4], "exist": [0, 3, 9], "whether": [0, 1, 2, 3, 4, 6, 7, 9], "ha": [0, 3, 8, 9, 10], "alreadi": [0, 3], "been": [0, 3, 6, 9, 10], "answer": 0, "emitt": [1, 9, 11], "oper": [1, 2, 4, 5, 7, 8, 9, 11], "util": [1, 2, 8, 11], "registri": 1, "elementwis": [1, 3, 9], "can": [1, 2, 3, 4, 6, 7, 8, 9, 10], "ad": [1, 8], "mani": [1, 3, 6, 8, 9], "kernel": [1, 2, 3, 4, 5, 7, 9], "cutla": 1, "code": [1, 3, 6, 9], "like": [1, 8], "follow": [1, 3, 8, 9, 10], "gemm": [1, 2, 5, 9, 11], "plan": [1, 2, 3, 6, 7, 8, 9, 10], "op": [1, 2, 3, 4, 6, 7, 8, 9], "element": [1, 2, 3, 4, 5, 6, 8, 9], "datatyp": [1, 3, 4], "f32": [1, 3], "layout": [1, 2, 3, 4, 6, 7, 8, 9], "layouttyp": [1, 2, 3, 6, 7, 8, 9], "rowmajor": [1, 2, 3, 6, 7, 8, 9], "activ": [1, 3, 5], "relu": [1, 3, 5, 9], "get_activation_epilogu": [1, 11], "element_output": 1, "elements_per_access": 1, "element_accumul": [1, 3, 6], "element_comput": 1, "sourc": [1, 2, 3, 4, 9], "return": [1, 2, 3, 4, 6, 7, 8], "correspond": [1, 8], "function": [1, 3, 4, 5, 6, 8], "data": [1, 3, 6, 9, 11], "align": [1, 3, 4, 11], "us": [1, 2, 3, 4, 8, 9, 10], "paramet": [1, 2, 3, 4, 6, 9], "output": [1, 2, 3, 6, 7], "int": [1, 2, 3, 4], "operand": [1, 3, 6], "c": [1, 2, 3, 6, 7, 8, 9], "accumul": [1, 3], "which": [1, 2, 3, 6, 8, 9, 10], "comput": [1, 2, 3, 4, 8], "should": [1, 2, 3, 9], "perform": [1, 3, 7, 9], "functor": [1, 3], "get_activ": [1, 7, 11], "list": [1, 3, 4, 7, 9], "avail": [1, 9], "class": [1, 3, 7], "contain": [1, 2, 8, 9, 10], "valid": [1, 3, 4], "given": [1, 3, 6, 7, 9], "capabl": [1, 2, 3, 4, 8], "library_default": 1, "archopt": [1, 11], "target_cc": 1, "kernel_cc": [1, 3], "operation_kind": [1, 4], "gemm_kind": 1, "allowed_math_oper": 1, "mathoper": [1, 4, 9], "multiply_add": [1, 9], "1": [1, 3, 6, 7, 8, 9, 10], "multiply_add_satur": 1, "2": [1, 3, 6, 7, 8, 9], "object": [1, 3], "structur": 1, "keep": 1, "track": 1, "devic": [1, 2, 3, 4, 6, 8], "run": [1, 2, 3, 5, 8, 9, 10], "gener": [1, 2, 3, 6, 7, 8], "operationkind": 1, "regist": 1, "includ": [1,
|