MCPcopy Create free account

hub / github.com/fla-org/flash-linear-attention / functions

Functions2,989 in github.com/fla-org/flash-linear-attention

Functiontest_cp4_single_sequence
CP4: single long sequence spanning all ranks.
tests/context_parallel/test_cp_gdn.py:356
Functiontest_cp4_state_v_first
CP4: state_v_first=True with single long sequence.
tests/context_parallel/test_cp_kda.py:574
Functiontest_cp4_state_v_first
CP4: state_v_first=True with single long sequence.
tests/context_parallel/test_cp_gdn.py:445
Functiontest_cp4_worst_case_many_len1
CP4: globally length-1 sequences + short tails across multiple ranks. Stress test for cross-sequence gradient isolation in CP backward.
tests/context_parallel/test_cp_conv.py:533
Functiontest_cp8_single_sequence
CP8: single long sequence spanning all ranks.
tests/context_parallel/test_cp_kda.py:508
Functiontest_cp8_single_sequence
CP8: single long sequence spanning all ranks.
tests/context_parallel/test_cp_gdn.py:370
Functiontest_cu_seqlens_cpu_matches_device
(name, impl, ref)
tests/ops/utils/test_index.py:128
Functiontest_cumsum_local_scalar_fwd
( B: int, T: int, H: int, chunk_size: int, dtype: torch.dtype, )
tests/ops/test_comba.py:44
Functiontest_decode
(B: int, Sq: int, Skv: int, H: int, HQ: int, D: int, W, dtype: torch.dtype)
tests/ops/test_parallax.py:285
Functiontest_decode_cache_layout_shapes
Pre-rescaled cache has documented shapes; r_cache size == ceil(T/C).
tests/ops/test_wall_attn.py:448
Functiontest_decode_matches_parallel
The decode kernel must reproduce the (reference-verified) training kernel for the corresponding positions: decode(last m queries | full KV) == par
tests/ops/test_parallax.py:314
Functiontest_decode_matches_training_forward
Decode at position t reproduces the training forward output at that row.
tests/ops/test_wall_attn.py:330
Functiontest_decode_matches_training_forward_long
Long-context stability: per-block reference must keep exp2 finite.
tests/ops/test_wall_attn.py:347
Functiontest_decode_one_step
Single-query decode kernel: must match the training kernel at the last position and the prefill-shaped decode kernel.
tests/ops/test_parallax.py:345
Functiontest_decode_streaming_matches_full_forward
End-to-end serving pattern: prefill the cache, then decode token-by-token, appending each new (k_tilde, v) to a pre-allocated buffer, and compare
tests/ops/test_wall_attn.py:389
Functiontest_decode_with_scalar_gate
Wall + FoX-style scalar gate: decode matches training forward.
tests/ops/test_wall_attn.py:367
Functiontest_decoding_one_step
( B: int, H: int, D: int, gate_range: tuple[float, float], max_CG_step: int, dtype: to
tests/ops/test_mesa.py:230
Functiontest_deltaformer_attn
Test DeltaFormer pre-attention by comparing fused implementation with naive reference.
tests/ops/test_deltaformer.py:32
Functiontest_deltaformer_attn_varlen
( H: int, D: int, cu_seqlens: list[int], dtype: torch.dtype, )
tests/ops/test_deltaformer.py:88
Functiontest_deprecate_kwarg_can_raise_when_both_names_are_set
()
tests/test_utils.py:92
Functiontest_deprecate_kwarg_renames_old_keyword_and_warns
()
tests/test_utils.py:78
Functiontest_dg_nonzero_after_backward
()
tests/ops/test_wall_attn.py:176
Functiontest_edge_cases
()
tests/ops/utils/test_index.py:163
Functiontest_fast_conv_varlen
( N: int, T: int, D: int, W: int, activation: str, has_bias: bool, has_residual: b
tests/modules/test_conv.py:806
Functiontest_fla_cache_autotune_handles_none_restore_value
A None argument listed in restore_value must not crash autotuning. Triton's default pre_hook clones every ``restore_value`` argument during bench
tests/ops/test_cache.py:39
Functiontest_fla_layer_reset
()
tests/models/test_cache.py:259
Functiontest_flashkda_chunk
(B, T, H, D, monkeypatch)
tests/ops/test_kda.py:1235
Functiontest_flashkda_chunk_varlen
(H, D, cu_seqlens, monkeypatch)
tests/ops/test_kda.py:1274
Functiontest_full_split_import_contract_when_runtime_dependencies_available
(tmp_path: Path)
tests/test_split_package_release.py:273
Functiontest_fused
( B: int, T: int, D: int, V: int, reduction: str, accumulate_grad_in_fp32: bool, d
tests/modules/test_kl_div.py:27
Functiontest_fused_chunk
( B: int, T: int, H: int, D: int, scale: float, dtype: torch.dtype, )
tests/ops/test_ttt.py:121
Functiontest_fused_chunk
( B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/ops/test_linear_attn.py:177
Functiontest_fused_chunk
( B: int, T: int, H: int, D: int, dtype: torch.dtype, scale: float, gate_logit_nor
tests/ops/test_simple_gla.py:466
Functiontest_fused_chunk
( B: int, T: int, H: int, K: int, expand_ratio: int, dtype: torch.dtype, )
tests/ops/test_retention.py:157
Functiontest_fused_chunk_varlen
( H: int, D: int, cu_seqlens: list[int], dtype: torch.dtype, )
tests/ops/test_simple_gla.py:542
Functiontest_fused_chunk_varlen
( H: int, K: int, expand_ratio: int, cu_seqlens: list[int], dtype: torch.dtype, )
tests/ops/test_retention.py:210
Functiontest_fused_cross_entropy
( B: int, T: int, D: int, V: int, reduction: str, dtype: torch.dtype )
tests/modules/test_cross_entropy.py:27
Functiontest_fused_cross_entropy_softcap
( B: int, T: int, D: int, V: int, softcap: float, reduction: str, dtype: torch.dty
tests/modules/test_cross_entropy.py:66
Functiontest_fused_grpos
(B: int, T: int, V: int, dtype: torch.dtype, inplace: bool, repeat: int)
tests/modules/test_grpo.py:21
Functiontest_fused_k_update
( B: int, T: int, H: int, D: int, dtype: torch.dtype, ka_shape: int, )
tests/ops/test_rwkv7.py:239
Functiontest_fused_linear_cross_entropy
( B: int, T: int, D: int, V: int, scale: float, reduction: str, accumulate_grad_in
tests/modules/test_cross_entropy.py:109
Functiontest_fused_linear_cross_entropy_l2_warp
( B: int, T: int, H: int, V: int, l2_penalty_factor: float, dtype: torch.dtype, )
tests/modules/test_l2warp.py:28
Functiontest_fused_linear_cross_entropy_softcap
( B: int, T: int, D: int, V: int, softcap: float, reduction: str, dtype: torch.dty
tests/modules/test_cross_entropy.py:164
Functiontest_fused_mul_recurrent_fwd
( B: int, T: int, H: int, D: int, scale: float, dtype: torch.dtype, )
tests/ops/test_rwkv7.py:90
Functiontest_fused_recurrent
(B, T, H, K, V, scale, use_qk_l2norm_in_kernel, dtype)
tests/ops/test_gdn2.py:90
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, scale: float, gate_logit_normalizer: float, dtyp
tests/ops/test_comba.py:74
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, scale: float | None, dtype: torch.dtype, )
tests/ops/test_linear_attn.py:31
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, scale: float, dtype: torch.dtype, )
tests/ops/test_dplr_delta.py:212
Functiontest_fused_recurrent
( B: int, T: int, H: int, HV: int, D: int, scale: float, gate_logit_normalizer: fl
tests/ops/test_kda.py:104
Functiontest_fused_recurrent
( B: int, T: int, D: int, dtype: torch.dtype, )
tests/ops/test_hgrn.py:32
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, M: int, gate_logit_normalizer: float, dtype: tor
tests/ops/test_gsa.py:40
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, gate_logit_normalizer: float, dtype: torch.dtype, )
tests/ops/test_gla.py:39
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, scale: float, gate_logit_normalizer: float, dtyp
tests/ops/test_simple_gla.py:38
Functiontest_fused_recurrent
( B: int, T: int, H: int, HV: int, D: int, scale: float, gate_logit_normalizer: fl
tests/ops/test_gdn.py:38
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, scale: float, gate_logit_normalizer: float, dtyp
tests/ops/test_oja.py:210
Functiontest_fused_recurrent
( B: int, T: int, H: int, D: int, scale: float, dtype: torch.dtype, )
tests/ops/test_iplr_delta.py:135
Functiontest_fused_recurrent_beta_sigmoid_in_kernel
fused_recurrent_gated_delta_rule with use_beta_sigmoid_in_kernel=True matches manual sigmoid.
tests/ops/test_gdn.py:485
Functiontest_fused_recurrent_gate_in_kernel
use_gate_in_kernel=True must match the manually-activated gate path.
tests/ops/test_gdn2.py:125
Functiontest_fused_recurrent_gate_in_kernel
fused_recurrent_kda with use_gate_in_kernel=True matches manual gate path.
tests/ops/test_kda.py:298
Functiontest_fused_recurrent_gate_in_kernel
fused_recurrent_gated_delta_rule with use_gate_in_kernel=True matches manual gate.
tests/ops/test_gdn.py:422
Functiontest_fused_recurrent_gate_in_kernel_varlen
Varlen fused_recurrent_gated_delta_rule with use_gate_in_kernel=True.
tests/ops/test_gdn.py:545
Functiontest_fused_recurrent_state_v_first
state_v_first stores the state transposed to [V, K]; output must match.
tests/ops/test_gdn2.py:149
Functiontest_fused_recurrent_state_v_first
( B: int, T: int, H: int, D: int, scale: float, gate_logit_normalizer: float, dtyp
tests/ops/test_kda.py:228
Functiontest_fused_recurrent_state_v_first
( B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/ops/test_gla.py:110
Functiontest_fused_recurrent_state_v_first
( B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/ops/test_simple_gla.py:108
Functiontest_fused_recurrent_state_v_first
( B: int, T: int, H: int, HV: int, D: int, scale: float, gate_logit_normalizer: fl
tests/ops/test_gdn.py:350
Functiontest_fused_recurrent_state_v_first
( B: int, T: int, H: int, K: int, expand_ratio: int, dtype: torch.dtype, )
tests/ops/test_retention.py:327
Functiontest_fused_recurrent_use_beta_sigmoid_in_kernel
( B: int, T: int, H: int, D: int, allow_neg_eigval: bool, dtype: torch.dtype, )
tests/ops/test_kda.py:164
Functiontest_fused_recurrent_varlen
Packed varlen recurrent run must equal running each sequence on its own.
tests/ops/test_gdn2.py:192
Functiontest_fused_recurrent_varlen
( D: int, cu_seqlens: list[int], dtype: torch.dtype, )
tests/ops/test_hgrn.py:80
Functiontest_fused_recurrent_varlen
( H: int, D: int, M: int, cu_seqlens: list[int], dtype: torch.dtype, )
tests/ops/test_gsa.py:119
Functiontest_fused_recurrent_varlen
( H: int, D: int, cu_seqlens: list[int], dtype: torch.dtype, )
tests/ops/test_gla.py:177
Functiontest_fused_recurrent_varlen
( H: int, D: int, scale: float, gate_logit_normalizer: float, cu_seqlens: list[int], d
tests/ops/test_simple_gla.py:173
Functiontest_fused_recurrent_vllm_decode
Test vLLM-style decoding with continuous batching and paged state storage.
tests/ops/test_kda.py:374
Functiontest_fused_recurrent_with_initial_state
()
tests/ops/test_gdn2.py:162
Functiontest_fused_rejects_trainable_teacher
()
tests/modules/test_kl_div.py:65
Functiontest_fused_rwkv7_addcmul
( B: int, T: int, H: int, D: int, dtype: torch.dtype, use_g: bool, )
tests/ops/test_rwkv7.py:150
Functiontest_g_gradient_matches_finite_differences
dL/dg for the Triton Wall path vs central finite differences. The loss is accumulated in fp64; the step is sized for fp32 softmax logits.
tests/ops/test_wall_attn.py:188
Functiontest_gate
( B: int, T: int, H: int, D: int, HAS_BIAS: bool, LOWER_BOUND: float | None, )
tests/ops/test_kda.py:1040
Functiontest_gate
( B: int, T: int, HV: int, HAS_BIAS: bool, )
tests/ops/test_gdn.py:1081
Functiontest_gate_output_correction
( B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/ops/test_rwkv7.py:275
Functiontest_gelu
(B: int, T: int, D: int, noncontiguous: bool, compile: bool)
tests/modules/test_activation.py:136
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_mom.py:68
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_gsa.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_samba.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_deltanet.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_linear_attn.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_retnet.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_parallax.py:68
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_gla.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_rodimus.py:54
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_mla.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_nsa.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_bitnet.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_rwkv7.py:55
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_mamba3.py:104
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, use_rope: bool, dtype: torch.dtype, )
tests/models/test_modeling_raven.py:66
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_wall_transformer.py:66
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_deltaformer.py:67
Functiontest_generation
( L: int, B: int, T: int, H: int, D: int, dtype: torch.dtype, )
tests/models/test_modeling_kda.py:66
← previousnext →2,701–2,800 of 2,989, ranked by callers