Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/tile-ai/tilelang
/ functions
Functions
7,254 in github.com/tile-ai/tilelang
⨍
Functions
7,254
◇
Types & classes
525
↳
Endpoints
224
↓ 2 callers
Function
generate
Generates new tokens based on the given prompt tokens using the specified model. Args: model (Transformer): The transformer model us
examples/deepseek_v32/inference/generate.py:31
↓ 2 callers
Method
generate
( self, prompts: List[str], sampling_params: SamplingParams, images: Optional[
examples/bitnet-1.58b/vllm_workspace/conftest.py:443
↓ 2 callers
Function
generate_block_indices
Generate random block indices for the benchmark.
examples/deepseek_nsa/benchmark/benchmark_nsa_fwd.py:567
↓ 2 callers
Method
generate_greedy
( self, prompts: List[str], max_tokens: int, images: Optional[List[Image.Image
examples/bitnet-1.58b/vllm_workspace/conftest.py:248
↓ 2 callers
Function
generate_input
( dhidden: int, dexpert: int, nroutedexperts: int, nsharedexperts: int, nexpertspertoken: int, bs: int, se
examples/fusedmoe/example_fusedmoe_torch.py:137
↓ 2 callers
Function
generate_random_cu_seqlens
(per_cp_seqlen, cp_size=4, cp_rank=3, kv_stride=1, average_q_len=512)
examples/deepseek_v32/utils.py:199
↓ 2 callers
Function
generate_test_data
Generate test data for big fuse operator.
examples/deepseek_mhc/example_mhc_pre.py:340
↓ 2 callers
Function
generate_test_data
Generate test data for post operator.
examples/deepseek_mhc/example_mhc_post.py:78
↓ 2 callers
Function
generate_text
(model, tokenizer, prompt, max_length=100)
examples/bitnet-1.58b/eval_correctness.py:18
↓ 2 callers
Function
generate_text
(model, tokenizer, prompt, max_length=100)
examples/bitnet-1.58b/maint/create_bitblas_ckpt.py:33
↓ 2 callers
Method
generator
(self)
tilelang/autodd.py:580
↓ 2 callers
Function
get_access_vars
(region: list[Range])
tilelang/carver/matmul_analysis.py:444
↓ 2 callers
Function
get_annotated_mod
( func_or_mod: tirx.PrimFunc | tvm.IRModule, target: str | Target = "auto", target_host: str | Tar
tilelang/jit/adapter/utils.py:117
↓ 2 callers
Method
get_avaliable_tensorintrin_shapes
(self)
tilelang/carver/arch/rdna.py:96
↓ 2 callers
Method
get_block_extent
Returns the block extent for the given dimension. dim=0 corresponds to blockIdx.x, dim=1 to blockIdx.y, and dim=2 to blockIdx.z.
tilelang/language/kernel.py:181
↓ 2 callers
Method
get_buffer_dtype
(self, buffer: tirx.Buffer)
tilelang/carver/roller/node.py:329
↓ 2 callers
Function
get_buffer_elems
Get the number of elements in the buffer.
tilelang/utils/language.py:150
↓ 2 callers
Function
get_configs
()
testing/python/autotune/test_tilelang_autotune.py:27
↓ 2 callers
Function
get_cplus_compiler
Return the path to the default C/C++ compiler. Returns ------- out: Optional[str] The path to the default C/C++ compiler, or None
tilelang/contrib/cc.py:71
↓ 2 callers
Function
get_cuda_library_dirs
Return CUDA library directories that nvcc may need for host linking.
tilelang/contrib/nvcc.py:312
↓ 2 callers
Method
get_default_execution_backend
Get default execution backend from environment.
tilelang/env.py:380
↓ 2 callers
Method
get_default_target
Get default compilation target from environment.
tilelang/env.py:376
↓ 2 callers
Method
get_default_verbose
Get default verbose flag from environment.
tilelang/env.py:384
↓ 2 callers
Method
get_device_source
Returns the source code of the device module.
tilelang/jit/adapter/tvm_ffi.py:314
↓ 2 callers
Function
get_err_ratio
(x, y)
examples/kda/FLA_KDA/fla_utils.py:38
↓ 2 callers
Method
get_forward_vars
Retrieve the iteration variables associated with the layout. Returns ------- List[IterVar] A list of ite
tilelang/layout/layout.py:79
↓ 2 callers
Method
get_generated_module_source
Get the importable generated CuTeDSL Python module source.
tilelang/jit/adapter/cutedsl/adapter.py:247
↓ 2 callers
Function
get_git_commit_id
Get the current git commit hash by running git in the current file's directory.
version_provider.py:28
↓ 2 callers
Function
get_heuristic_config
()
examples/gemm/example_gemm_autotune.py:171
↓ 2 callers
Function
get_heuristic_config
()
examples/gemm/example_gemm_advanced_autotune.py:229
↓ 2 callers
Function
get_heuristic_config
()
examples/convolution/example_convolution_autotune.py:59
↓ 2 callers
Function
get_heuristic_config
()
examples/flash_decoding/example_gqa_decode.py:28
↓ 2 callers
Function
get_host_call
(is_device_c: bool = False)
tilelang/engine/lower.py:51
↓ 2 callers
Method
get_host_source
Returns the source code of the host module.
tilelang/jit/adapter/tvm_ffi.py:304
↓ 2 callers
Method
get_host_source
Get the cached host-side source code.
tilelang/jit/adapter/cutedsl/adapter.py:240
↓ 2 callers
Function
get_in_out_dtypes
Detect In/Out data types for the given block based on the analysis if read/write buffers.
tilelang/carver/matmul_analysis.py:409
↓ 2 callers
Function
get_mma_micro_size
(dtype: Literal["float16", "int8"])
tilelang/rocm/intrinsics/utils.py:18
↓ 2 callers
Function
get_msvc_environment_error
()
tilelang/contrib/msvc.py:244
↓ 2 callers
Method
get_node_reduce_step_candidates
Calculates reduction step candidates for each reduction axis in a PrimFuncNode. General idea : use factor first, since it does not require ex
tilelang/carver/roller/policy/default.py:241
↓ 2 callers
Function
get_nvcc_subprocess_env
Return the environment used by NVCC subprocesses. On Windows, the pip CUDA toolkit can provide nvcc without a system CUDA install, but nvcc s
tilelang/contrib/nvcc.py:22
↓ 2 callers
Function
get_nvrtc_version
()
tilelang/contrib/nvrtc.py:8
↓ 2 callers
Method
get_opt_shape
(self, name)
tilelang/carver/roller/node.py:286
↓ 2 callers
Method
get_or_create_node
(self, name)
tilelang/carver/roller/shape_inference/tir.py:104
↓ 2 callers
Function
get_ordered_axes
(region: list[Range])
tilelang/carver/matmul_analysis.py:337
↓ 2 callers
Method
get_reduce_axis
(self, block: SBlockRV)
tilelang/carver/roller/node.py:57
↓ 2 callers
Function
get_reduction_blocks
(sch: Schedule, blocks: list[SBlockRV])
tilelang/carver/analysis.py:263
↓ 2 callers
Function
get_rocm_arch
Utility function to get the AMD GPU architecture Parameters ---------- rocm_path : str The path to rocm installation directory
tilelang/contrib/rocm.py:231
↓ 2 callers
Function
get_sparse_attn_mask_from_topk
(x, topk, use_dense_for_last_block=False)
examples/blocksparse_attention/example_tilelang_block_sparse_attn.py:9
↓ 2 callers
Function
get_sparse_attn_mask_from_topk
(x, topk, use_dense_for_last_block=False)
examples/blocksparse_attention/block_sparse_attn_triton.py:13
↓ 2 callers
Function
get_sparse_attn_mask_from_topk
(x, topk, use_dense_for_last_block=False)
examples/seer_attention/block_sparse_attn_triton.py:14
↓ 2 callers
Function
get_target_arch
(compute_version: str | tuple[int, int])
tilelang/contrib/nvcc.py:453
↓ 2 callers
Method
get_tcgen5_blockscaled_instr_desc
Build the block-scaled instruction descriptor via FFI.
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:455
↓ 2 callers
Method
get_tcgen5_instr_desc
Build the 64-bit instruction descriptor for a ``tcgen05.mma`` PTX call.
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:584
↓ 2 callers
Function
get_tensor_supply
(supply_type: TensorSupplyType = TensorSupplyType.Integer)
tilelang/utils/tensor.py:42
↓ 2 callers
Method
get_thread_binding
(self)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:142
↓ 2 callers
Method
get_thread_size
Returns the extent (range size) of the thread dimension. If the Fragment was replicated over threads, this will reflect the n
tilelang/layout/fragment.py:109
↓ 2 callers
Function
get_type_hints
(func)
tilelang/language/eager/builder.py:863
↓ 2 callers
Function
get_wrong_kernel
(M: int = 4096)
testing/python/issue/test_tilelang_issue_1549.py:10
↓ 2 callers
Function
grouped_gemm
args: a (torch.Tensor): Input tensor of shape (M, K). b (torch.Tensor): Input tensor of shape (G, K, N).
examples/grouped_gemm/example_grouped_gemm_fwd.py:40
↓ 2 callers
Function
hadamard
(A, n, dtype)
examples/hadamard_transform/example_hadamard.py:35
↓ 2 callers
Function
has_common_reduce
(var: Var)
tilelang/carver/matmul_analysis.py:348
↓ 2 callers
Function
has_device_kernel_launch
Check if the attributes indicate a device kernel launch.
tilelang/engine/lower.py:26
↓ 2 callers
Function
hisa_indexer
Run the full hisa prefill pipeline. Returns: ``[M, topk_tokens]`` int32 — each row is this query's top ``topk_tokens`` K positions, expressed
examples/dsa_hisa/hisa.py:11
↓ 2 callers
Function
host_codegen
Generate host-side code from the lowered IR module. Parameters ---------- host_mod : tvm.IRModule The host-side IR module to comp
tilelang/engine/lower.py:194
↓ 2 callers
Function
impl
(func: Callable[_P, _T])
tilelang/language/eager/builder.py:852
↓ 2 callers
Function
indexer_bwd_interface
( q: torch.Tensor, weights: torch.Tensor, k: torch.Tensor, attn_score: torch.Tensor, index
examples/dsa_sparse_finetune/indexer_bwd.py:170
↓ 2 callers
Function
indexer_topk_reducesum_interface
( q: torch.Tensor, weights: torch.Tensor, k: torch.Tensor, topk: int, offsets: torch.Tenso
examples/dsa_sparse_finetune/indexer_topk_reducesum.py:191
↓ 2 callers
Method
init_wgmma_a_desc
Emit TIR to initialize a pre-allocated WGMMA A descriptor (SS variant). Parameters ---------- desc_a : Buffer A d
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:431
↓ 2 callers
Method
initialize_jit_mode
(self, *args: _P.args, **kwargs: _P.kwargs)
tilelang/jit/__init__.py:332
↓ 2 callers
Function
interleave_weight
Interleave the weight to the target data type. Args: qweight (_type_): _description_ nbits (int, optional): _description_. Defaul
tilelang/quantize/utils.py:71
↓ 2 callers
Function
is_ampere_arch
(arch: TileDevice)
tilelang/carver/arch/cuda.py:25
↓ 2 callers
Method
is_autotune_cache_disabled
(self)
tilelang/env.py:367
↓ 2 callers
Method
is_boolean
Checks if the parameter represents a boolean type. Returns: bool: True if parameter is a boolean type, False otherwise
tilelang/engine/param.py:115
↓ 2 callers
Function
is_common_reduce
(var: Var)
tilelang/carver/matmul_analysis.py:345
↓ 2 callers
Function
is_f8f6f4_family
Whether *value* is a tcgen05 ``kind::f8f6f4`` / ``mxf8f6f4`` operand dtype.
tilelang/language/dtypes.py:304
↓ 2 callers
Method
is_float8
Checks if the parameter represents a float8 type. Returns: bool: True if parameter is a float8 type, False otherwise
tilelang/engine/param.py:91
↓ 2 callers
Method
is_gemm_rr
(self)
tilelang/rocm/op/gemm/gemm_wmma.py:167
↓ 2 callers
Method
is_gemm_rr
(self)
tilelang/rocm/op/gemm/gemm_mfma.py:247
↓ 2 callers
Method
is_gemm_rr
(self)
tilelang/cuda/op/gemm/gemm_mma.py:230
↓ 2 callers
Method
is_gemm_rr
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_mma.py:261
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/rocm/op/gemm/gemm_wmma.py:164
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/rocm/op/gemm/gemm_mfma.py:244
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/tileop/gemm_sp/gemm_sp_wgmma.py:157
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/cuda/op/gemm/gemm_mma_sm70.py:179
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/cuda/op/gemm/gemm_wgmma.py:174
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/cuda/op/gemm/gemm_mma.py:227
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_wgmma.py:157
↓ 2 callers
Method
is_gemm_rs
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_mma.py:258
↓ 2 callers
Method
is_gemm_sr
(self)
tilelang/rocm/op/gemm/gemm_wmma.py:161
↓ 2 callers
Method
is_gemm_sr
(self)
tilelang/rocm/op/gemm/gemm_mfma.py:241
↓ 2 callers
Method
is_gemm_sr
(self)
tilelang/cuda/op/gemm/gemm_mma.py:224
↓ 2 callers
Method
is_gemm_sr
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_mma.py:255
↓ 2 callers
Method
is_gemm_ss
(self)
tilelang/rocm/op/gemm/gemm_wmma.py:158
↓ 2 callers
Method
is_gemm_ss
(self)
tilelang/rocm/op/gemm/gemm_mfma.py:238
↓ 2 callers
Method
is_gemm_ss
(self)
tilelang/tileop/gemm_sp/gemm_sp_wgmma.py:151
↓ 2 callers
Method
is_gemm_ss
(self)
tilelang/cuda/op/gemm/gemm_mma_sm70.py:173
↓ 2 callers
Method
is_gemm_ss
(self)
tilelang/cuda/op/gemm/gemm_wgmma.py:168
↓ 2 callers
Method
is_gemm_ss
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_wgmma.py:151
↓ 2 callers
Method
is_gemm_ss
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_mma.py:252
↓ 2 callers
Method
is_gemm_ts
Return True if A is in tensor memory and B is in shared memory (TS variant).
tilelang/tileop/gemm/gemm_base.py:65
← previous
next →
1,101–1,200 of 7,254, ranked by callers