MCPcopy Create free account

hub / github.com/tile-ai/tilelang / functions

Functions7,254 in github.com/tile-ai/tilelang

Functionfence_proxy_async
()
tilelang/contrib/cutedsl/cpasync.py:334
Functionfence_proxy_async
Issue a shared memory fence for asynchronous proxy operations. Ensures that prior asynchronous operations (e.g. TMA stores) are visible to su
tilelang/language/builtin.py:337
Functionfinalize_reducer
Finalize a reducer buffer by emitting the `tl.tileop.finalize_reducer` intrinsic. This returns a TVM `tirx.Call` handle that finalizes the g
tilelang/language/reduce_op.py:300
Functionfind_arg_idx_from_buffer_chain
traverse to find the arg index from the buffer
tilelang/carver/matmul_analysis.py:164
Functionfind_last_producer_from_buffer
(sch, main_block, buffer: tirx.Buffer)
tilelang/carver/matmul_analysis.py:141
Functionfind_layers
(module, layers=None, name="")
examples/bitnet-1.58b/modeling_bitnet.py:59
Functionfind_lib_path
Find tile lang library Parameters ---------- name : str The name of the library optional: boolean Whether the librar
tilelang/libinfo.py:13
Functionfind_topo_sort_priority
(output_node_list)
tilelang/carver/roller/node.py:543
Functionfind_var_from_func
(func, var: str)
tilelang/carver/analysis.py:161
Functionfirst_arg
(*args)
testing/python/language/test_tilelang_language_eager_ast_signature.py:15
Functionflash_bwd
( Q: T.Tensor(shape, dtype), # type: ignore K: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_mha_bwd_bhsd.py:166
Functionflash_bwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(k_shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd_tma_reduce.py:182
Functionflash_bwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(k_shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd.py:167
Functionflash_bwd
( Q: T.Tensor(shape, dtype), # type: ignore K: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_mha_bwd_bshd.py:163
Functionflash_bwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(k_shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:250
Functionflash_bwd
( Q: T.Tensor(q_shape, dtype), K: T.Tensor(kv_shape, dtype), V: T.Tensor(kv_shape, dty
examples/attention_sink/example_gqa_sink_bwd_varlen.py:330
Functionflash_bwd
( Q: T.Tensor(shape, dtype), # type: ignore K: T.Tensor(shape, dtype), # type: ignore
examples/attention_sink/example_mha_sink_bwd_bhsd.py:225
Functionflash_bwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(kv_shape, dtype), # type: ignore
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:220
Functionflash_bwd_dsink
( Sinks: T.Tensor([heads], dtype), Delta: T.Tensor(shape, accum_dtype), lse: T.Tensor(
examples/attention_sink/example_gqa_sink_bwd_varlen.py:469
Functionflash_bwd_dsink
( Sinks: T.Tensor([heads], dtype), # type: ignore Delta: T.Tensor(shape, accum_dtype), # typ
examples/attention_sink/example_mha_sink_bwd_bhsd.py:319
Functionflash_bwd_dsink
( Sinks: T.Tensor([heads], dtype), # type: ignore Delta: T.Tensor(shape, accum_dtype), # typ
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:311
Functionflash_bwd_kernel
( Q: T.Tensor(q_shape, dtype), K: T.Tensor(kv_shape, dtype), V: T.Tensor(kv_shape, dty
examples/amd/example_amd_flash_attn_bwd.py:331
Functionflash_bwd_post
(dQ_in: T.Tensor(shape, accum_dtype), dQ_out: T.Tensor(shape, dtype))
examples/amd/example_amd_flash_attn_bwd.py:437
Functionflash_bwd_post
( dQ: T.Tensor(shape, accum_dtype), # type: ignore dQ_out: T.Tensor(shape, dtype), # type: i
examples/flash_attention/example_mha_bwd_bhsd.py:139
Functionflash_bwd_post
( dQ: T.Tensor(q_shape, accum_dtype), # type: ignore dK: T.Tensor(k_shape, accum_dtype), # t
examples/flash_attention/example_gqa_bwd_tma_reduce.py:142
Functionflash_bwd_post
( dQ: T.Tensor(shape, accum_dtype), # type: ignore dQ_out: T.Tensor(shape, dtype), # type: i
examples/flash_attention/example_gqa_bwd.py:137
Functionflash_bwd_post
( dQ: T.Tensor(shape, accum_dtype), # type: ignore dQ_out: T.Tensor(shape, dtype), # type: i
examples/flash_attention/example_mha_bwd_bshd.py:136
Functionflash_bwd_post
( dQ: T.Tensor(q_shape, accum_dtype), # type: ignore dK: T.Tensor(k_shape, accum_dtype), # t
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:207
Functionflash_bwd_post
( dQ: T.Tensor(shape, accum_dtype), dQ_out: T.Tensor(shape, dtype), )
examples/attention_sink/example_gqa_sink_bwd_varlen.py:282
Functionflash_bwd_post
( dQ: T.Tensor(shape, accum_dtype), # type: ignore dQ_out: T.Tensor(shape, dtype), # type: i
examples/attention_sink/example_mha_sink_bwd_bhsd.py:184
Functionflash_bwd_post
( dQ: T.Tensor(shape, accum_dtype), # type: ignore dQ_out: T.Tensor(shape, dtype), # type: i
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:185
Functionflash_bwd_prep
(O: T.Tensor(shape, dtype), dO: T.Tensor(shape, dtype), Delta: T.Tensor([batch, heads, seq_len], accum_dtype))
examples/amd/example_amd_flash_attn_bwd.py:289
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), # type: ignore dO: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_mha_bwd_bhsd.py:99
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), # type: ignore dO: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd_tma_reduce.py:100
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), # type: ignore dO: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd.py:97
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), # type: ignore dO: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_mha_bwd_bshd.py:96
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), # type: ignore dO: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:150
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), dO: T.Tensor(shape, dtype), cu_seqlens_q: T.Tensor([batch
examples/attention_sink/example_gqa_sink_bwd_varlen.py:229
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), # type: ignore dO: T.Tensor(shape, dtype), # type: ignore
examples/attention_sink/example_mha_sink_bwd_bhsd.py:145
Functionflash_bwd_prep
( O: T.Tensor(shape, dtype), # type: ignore dO: T.Tensor(shape, dtype), # type: ignore
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:146
Functionflash_fwd
( Q: T.Tensor(shape, dtype), # type: ignore K: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_mha_bwd_bhsd.py:22
Functionflash_fwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(k_shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd_tma_reduce.py:25
Functionflash_fwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(k_shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd.py:24
Functionflash_fwd
( Q: T.Tensor(shape, dtype), # type: ignore K: T.Tensor(shape, dtype), # type: ignore
examples/flash_attention/example_mha_bwd_bshd.py:22
Functionflash_fwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(k_shape, dtype), # type: ignore
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:40
Functionflash_fwd
( Q: T.Tensor(shape, dtype), # type: ignore K: T.Tensor(shape, dtype), # type: ignore
examples/attention_sink/example_mha_sink_bwd_bhsd.py:52
Functionflash_fwd
( Q: T.Tensor(q_shape, dtype), # type: ignore K: T.Tensor(kv_shape, dtype), # type: ignore
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:53
Functionflash_split_ref
(Q, K, V, causal)
examples/flash_decoding/example_mha_inference.py:183
Functionflash_split_ref
(Q, Q_pe, KV, K_pe)
examples/deepseek_mla/torch_refs.py:6
Functionflashattn_bwd_pipeline
(batch, heads, seq_len, dim, is_causal, block_M, block_N, groups=1)
examples/flash_attention_sm100/gqa_bwd_bshd.py:237
Functionflashattn_bwd_pipeline
Pipeline (default): 128 threads, 2 stages.
examples/flash_attention_sm100/mha_bwd_bshd.py:227
Functionflashattn_bwd_warp
(batch, heads, seq_len, dim, is_causal, block_M, block_N, groups=1)
examples/flash_attention_sm100/gqa_bwd_bshd.py:252
Functionflashattn_bwd_warp
ts: 256 threads, 2 stages. Use --variant ts to enable.
examples/flash_attention_sm100/mha_bwd_bshd.py:232
Functionflashattn_gqa_decode_no_split
( Q: T.Tensor(shape_q, dtype), K: T.Tensor(shape_k, dtype), V: T.Tensor(shape_v, dtype
examples/flash_decoding/example_gqa_decode_varlen_logits.py:54
Functionflashattn_gqa_decode_no_split
( Q: T.Tensor(shape_q, dtype), K: T.Tensor(shape_k, dtype), V: T.Tensor(shape_v, dtype
examples/flash_decoding/example_gqa_decode.py:191
Functionflashattn_gqa_decode_split
( Q: T.Tensor(shape_q, dtype), K: T.Tensor(shape_k, dtype), V: T.Tensor(shape_v, dtype
examples/flash_decoding/example_gqa_decode.py:65
Functionflashattn_mha_inference
( Q: T.Tensor(shape_q, dtype), K: T.Tensor(shape_kv, dtype), V: T.Tensor(shape_kv, dty
examples/flash_decoding/example_mha_inference.py:21
Functionfloor
Take floor of float input x. Parameters ---------- x : PrimExpr Input argument. span : Optional[Span] The location o
tilelang/language/tir/op.py:2629
Methodflush_binds
()
tilelang/language/eager/ast.py:375
Functionfma2
Packed fused multiply-add (x * y + z).
tilelang/language/math_intrinsics.py:382
Functionfmod
Return the remainder of x divided by y with the same sign as x. Parameters ---------- x : PrimExpr Input argument. y : PrimEx
tilelang/language/tir/op.py:3210
Functionfn
()
examples/deepseek_v32/sparse_mla_bwd.py:319
Functionfn
()
examples/dsa_sparse_finetune/sparse_mla_bwd.py:333
Functionfn
()
examples/dsa_hisa/pool_mqa_fp8.py:227
Functionfn
()
examples/dsa_hisa/block_sparse_mqa_fp8.py:236
Functionfn
()
examples/dsa_hisa/fp8_block_mean_pooling.py:126
Functionfn
()
examples/dsa_hisa/clean_and_maintain_logits.py:100
Functionfoo
( x: T.Buffer( [ 32, ], T.int32, ), )
testing/python/issue/test_tilelang_issue_1198.py:7
Functionformat_list
(ls)
examples/deepseek_mhc/example_mhc_bwd.py:268
Methodforward
(ctx, q, k, v, block_sparse_dense, sm_scale)
benchmark/blocksparse_attention/benchmark_triton_block_sparse_fmha.py:240
Methodforward
(ctx, q, k, v, block_indices, block_counts, block_size, window_size, scale, offsets)
examples/deepseek_nsa/example_triton_nsa_fwd_varlen.py:222
Methodforward
(ctx, q, k, v, block_indices, block_counts, block_size, window_size, scale, offsets)
examples/deepseek_nsa/example_tilelang_nsa_fwd_varlen.py:203
Methodforward
( ctx, q, k, v, block_indices, block_counts, block_siz
examples/deepseek_nsa/example_tilelang_nsa_bwd.py:616
Methodforward
(ctx, q, k, v, block_indices, block_size, scale, offsets)
examples/deepseek_nsa/example_triton_nsa_bwd.py:120
Methodforward
(ctx, q, k, v, block_indices, block_size, scale, offsets)
examples/deepseek_nsa/example_triton_nsa_fwd.py:120
Methodforward
(ctx, q, k, v, block_indices, block_size, scale, offsets)
examples/deepseek_nsa/benchmark/benchmark_nsa_fwd.py:114
Methodforward
Forward pass for parallel embedding layer. Args: x (torch.Tensor): Input tensor containing token indices. Retur
examples/deepseek_v32/inference/model.py:113
Methodforward
Forward pass for the custom linear layer. Args: x (torch.Tensor): Input tensor. Returns: torch.Tens
examples/deepseek_v32/inference/model.py:202
Methodforward
Forward pass for column parallel linear layer. Args: x (torch.Tensor): Input tensor. Returns: torch
examples/deepseek_v32/inference/model.py:231
Methodforward
Forward pass for row parallel linear layer. Args: x (torch.Tensor): Input tensor. Returns: torch.Te
examples/deepseek_v32/inference/model.py:267
Methodforward
Forward pass for RMSNorm. Args: x (torch.Tensor): Input tensor. Returns: torch.Tensor: Normalized t
examples/deepseek_v32/inference/model.py:301
Methodforward
(self, x: torch.Tensor)
examples/deepseek_v32/inference/model.py:336
Methodforward
(self, x: torch.Tensor, qr: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optio
examples/deepseek_v32/inference/model.py:478
Methodforward
Forward pass for the MLP layer. Args: x (torch.Tensor): Input tensor. Returns: torch.Tensor: Output
examples/deepseek_v32/inference/model.py:670
Methodforward
Forward pass for the gating mechanism. Args: x (torch.Tensor): Input tensor. Returns: Tuple[torch.T
examples/deepseek_v32/inference/model.py:716
Methodforward
Forward pass for the Expert layer. Args: x (torch.Tensor): Input tensor. Returns: torch.Tensor: Out
examples/deepseek_v32/inference/model.py:774
Methodforward
Forward pass for the MoE module. Args: x (torch.Tensor): Input tensor. Returns: torch.Tensor: Outpu
examples/deepseek_v32/inference/model.py:825
Methodforward
Forward pass for the Transformer block. Args: x (torch.Tensor): Input tensor. start_pos (int): Starting posi
examples/deepseek_v32/inference/model.py:877
Methodforward
Forward pass for the Transformer model. Args: tokens (torch.Tensor): Input tensor of token IDs with shape (batch_size, s
examples/deepseek_v32/inference/model.py:938
Methodforward
(ctx, a, b, batch_sizes, block_M, block_N, block_K, num_stages, threads)
examples/grouped_gemm/example_grouped_gemm_bwd.py:59
Methodforward
(ctx, x, loss)
examples/dsa_sparse_finetune/dsa.py:15
Methodforward
( ctx, q: torch.Tensor, kv: torch.Tensor, index_q: torch.Tensor, index
examples/dsa_sparse_finetune/dsa.py:116
Methodforward
(ctx, input, indices)
examples/flash_attention/bert_padding.py:10
Methodforward
(ctx, values, indices, first_axis_dim)
examples/flash_attention/bert_padding.py:41
Methodforward
(ctx, input, indices)
examples/flash_attention/bert_padding.py:65
Methodforward
(ctx, q, k, v, causal)
examples/flash_attention/example_mha_bwd_bhsd.py:249
Methodforward
(ctx, q, k, v, causal, groups=1, use_atomic=True)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:351
Methodforward
(ctx, q, k, v, causal, groups=1, use_atomic=True)
examples/flash_attention/example_gqa_bwd.py:341
Methodforward
(ctx, q, k, v, causal)
examples/flash_attention/example_mha_bwd_bshd.py:246
Methodforward
( ctx, q, k, v, seqlens_q, seqlens_k, cu_seqlens_q, cu_seqlens_k, max_seqlen_q, max_seqlen_k, causal,
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:489
← previousnext →4,401–4,500 of 7,254, ranked by callers