Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/fla-org/flash-linear-attention
/ functions
Functions
2,989 in github.com/fla-org/flash-linear-attention
⨍
Functions
2,989
◇
Types & classes
403
↳
Endpoints
219
↓ 1 callers
Function
yarn_get_mscale
(scale=1, mscale=1)
fla/layers/mla.py:44
Method
__bool__
(self)
fla/ops/common/intracard_cp.py:77
Method
__call__
( self, examples: list[list[int] | dict[str, Any]], )
legacy/training/flame/data.py:186
Method
__call__
(self, *args, **kwargs)
fla/models/modeling_layers.py:42
Method
__getitem__
(self, layer_idx: int)
fla/models/utils.py:208
Method
__getitem__
(self, layer_idx: int)
fla/models/utils.py:411
Method
__init__
(self, start_time: float = None, elapsed_time: float = None)
legacy/training/flame/logging.py:39
Method
__init__
( self, dataset: Dataset, tokenizer: PreTrainedTokenizer, context_len: int = 2
legacy/training/flame/data.py:27
Method
__init__
(self, search_dirs, test_dir, project_root=None)
scripts/find_dependent_tests.py:142
Method
__init__
(self, head_dim: int, *, device)
scripts/utils/autotune_generate.py:23
Method
__init__
(self, kernel_name: str, existing: KernelConfigFile | None = None)
scripts/utils/autotune_export.py:119
Method
__init__
(self, **kwargs)
evals/harness.py:18
Method
__init__
( self, model: PreTrainedModel, tokenizer: PreTrainedTokenizer, device: str =
evals/ppl.py:23
Method
__init__
(self, fn, arg_names, configs, key, reset_to_zero, restore_value, **kwargs)
fla/ops/utils/cache.py:345
Method
__init__
(self, operation_name: str)
fla/ops/backends/__init__.py:102
Method
__init__
( self, num_groups: int, hidden_size: int, elementwise_affine: bool = True,
fla/modules/layernorm.py:126
Method
__init__
( self, hidden_size: int, elementwise_affine: bool = True, bias: bool = False,
fla/modules/layernorm.py:972
Method
__init__
( self, num_groups: int, hidden_size: int, elementwise_affine: bool = True,
fla/modules/layernorm.py:1025
Method
__init__
( self, hidden_size, elementwise_affine: bool = True, bias: bool = False,
fla/modules/layernorm.py:1247
Method
__init__
( self, num_groups: int, hidden_size: int, elementwise_affine: bool = True,
fla/modules/layernorm.py:1303
Method
__init__
( self, hidden_size, elementwise_affine: bool = True, bias: bool = False,
fla/modules/layernorm.py:1369
Method
__init__
(self, *, sequence_dim: int = 1, use_local_output: bool = False)
fla/modules/layernorm.py:1425
Method
__init__
Initializes the BitLinear layer. Args: in_features: Size of each input sample. out_features: Size of each ou
fla/modules/fused_bitlinear.py:618
Method
__init__
(self, *, layouts: Placement | None = None)
fla/modules/parallel.py:21
Method
__init__
Arguments: ignore_index: int. If target == ignore_index, the loss is set to 0.0. label_smoothing: float l
fla/modules/fused_cross_entropy.py:377
Method
__init__
If group_size is not None, we do GroupNorm with each group having group_size elements. group_size=None is equivalent to group_size=hidden_size
fla/modules/layernorm_gated.py:472
Method
__init__
Args: reduction (`str`): Specifies the reduction to apply to the output: 'batchmean'. Default: 'batchmean'.
fla/modules/fused_kl_div.py:319
Method
__init__
( self, eps: float = 1e-6, output_dtype: torch.dtype | None = None, )
fla/modules/l2norm.py:278
Method
__init__
( self, hidden_size: int, elementwise_affine: bool = True, bias: bool = False,
fla/modules/fused_norm_gate.py:926
Method
__init__
( self, hidden_size: int, elementwise_affine: bool = True, bias: bool = False,
fla/modules/fused_norm_gate.py:1056
Method
__init__
( self, hidden_size: int, elementwise_affine: bool = True, eps: float = 1e-5,
fla/modules/fused_norm_gate.py:1076
Method
__init__
( self, hidden_size: int, elementwise_affine: bool = True, eps: float = 1e-5,
fla/modules/fused_norm_gate.py:1094
Method
__init__
( self, hidden_size: int, elementwise_affine: bool = True, eps: float = 1e-5,
fla/modules/fused_norm_gate.py:1154
Method
__init__
( self, hidden_size, elementwise_affine: bool = True, eps: float = 1e-5,
fla/modules/fused_norm_gate.py:1172
Method
__init__
( self, hidden_size: int, elementwise_affine: bool = True, eps: float = 1e-5,
fla/modules/fused_norm_gate.py:1231
Method
__init__
( self, *, input_layouts: Placement | None = None, output_layouts: Placement |
fla/modules/mlp.py:88
Method
__init__
Args: ignore_index: int. If target == ignore_index, the loss is set to 0.0. label_smoothing: float
fla/modules/fused_linear_cross_entropy.py:624
Method
__init__
interleaved: If True, rotate pairs of even and odd dimensions (GPT-J style) instead of 1st half and 2nd half (GPT-NeoX style).
fla/modules/rotary.py:337
Method
__init__
( self, head_dim: int, )
fla/modules/feature_map.py:49
Method
__init__
( self, head_dim: int, nu: int = 4, )
fla/modules/feature_map.py:108
Method
__init__
( self, head_dim: int, )
fla/modules/feature_map.py:124
Method
__init__
( self, head_dim: int, feature_dim: int, )
fla/modules/feature_map.py:138
Method
__init__
( self, head_dim: int, sketch_size: int | None = None, degree: int | None = 2,
fla/modules/feature_map.py:155
Method
__init__
( self, head_dim: int, )
fla/modules/feature_map.py:195
Method
__init__
( self, head_dim: int, use_gamma: bool | None = True, use_beta: bool | None =
fla/modules/feature_map.py:212
Method
__init__
( self, )
fla/modules/feature_map.py:254
Method
__init__
( self, )
fla/modules/feature_map.py:265
Method
__init__
( self, )
fla/modules/feature_map.py:276
Method
__init__
( self, )
fla/modules/feature_map.py:287
Method
__init__
( self, )
fla/modules/feature_map.py:298
Method
__init__
( self, hidden_size: int, kernel_size: int, bias: bool = False, activa
fla/modules/conv/short_conv.py:52
Method
__init__
Complex exponential positional embeddings for implicit long convolution filters.
fla/modules/conv/long_conv.py:83
Method
__init__
Long convolution with implicit filter parameterized by an MLP.
fla/modules/conv/long_conv.py:127
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, expand_k: float = 0.5,
fla/layers/gla.py:75
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/path_attn.py:31
Method
__init__
( self, mode: str = 'chunk', hidden_size: str = 1024, expand_k: float = 1.0,
fla/layers/linear_attn.py:28
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, expand_k: float = 1.,
fla/layers/gsa.py:32
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/deltaformer.py:63
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/attn.py:39
Method
__init__
( self, hidden_size: int = 2048, expand_v: float = 1.0, head_dim: int = 128,
fla/layers/gdn2.py:88
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, num_heads: int | None =
fla/layers/hgrn2.py:33
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 64, num_kv_heads: int | None
fla/layers/nsa.py:29
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, head_dim: int | None = 6
fla/layers/rwkv7.py:34
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 16, q_lora_rank: int | None
fla/layers/mla.py:55
Method
__init__
( self, hidden_size: int = 2048, head_dim: int = 256, num_heads: int = 4,
fla/layers/mom.py:285
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/parallax.py:64
Method
__init__
( self, num_heads: int, head_dim: int = 64, hidden_size: int = 2048, s
fla/layers/log_linear_mamba2.py:280
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 16, head_dim: int = 128,
fla/layers/mesa_net.py:59
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, expand_k: float = 1.0,
fla/layers/multiscale_retention.py:70
Method
__init__
( self, hidden_size: int = 2048, state_size: int = 16, conv_kernel: int = 4,
fla/layers/mamba.py:54
Method
__init__
( self, hidden_size: int = 2048, expand_v: float = 2, head_dim: int = 256,
fla/layers/gated_deltaproduct.py:35
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, expand_k: float = 1.,
fla/layers/simple_gla.py:68
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/wall_attn.py:69
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/bitattn.py:40
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/forgetting_attn.py:32
Method
__init__
( self, hidden_size: int = 1024, expand_k: float = 0.5, expand_v: float = 1.0,
fla/layers/abc.py:28
Method
__init__
( self, hidden_size: int = 2048, state_size: int = 128, expand: int = 2,
fla/layers/mamba3.py:65
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, expand_k: float = 1.,
fla/layers/raven.py:44
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, qkv_bias: bool = False,
fla/layers/rodimus.py:257
Method
__init__
( self, mode: str = 'chunk', d_model: int = None, hidden_size: int = 1024,
fla/layers/delta_net.py:77
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, num_heads: int | None =
fla/layers/lightnet.py:32
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/moba.py:92
Method
__init__
( self, dim: int, base: float = 10000.0, scale_base: float | None = None,
fla/layers/yoco.py:34
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/yoco.py:228
Method
__init__
( self, hidden_size: int = 2048, num_heads: int = 32, num_kv_heads: int | None
fla/layers/yoco.py:318
Method
__init__
( self, num_heads: int | None = None, head_dim: int = 64, hidden_size: int = 2
fla/layers/mamba2.py:113
Method
__init__
( self, hidden_size: int, feature_dim: int = 16, num_key_value_heads: int = 12
fla/layers/based.py:24
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, expand_k: float = 0.5,
fla/layers/rwkv6.py:32
Method
__init__
( self, input_dim: int, output_dim: int, low_rank_dim: int | None = None,
fla/layers/rwkv6.py:300
Method
__init__
( self, input_dim: int, output_dim: int, low_rank_dim: int | None = None,
fla/layers/rwkv6.py:335
Method
__init__
( self, hidden_size: int, l_max: int = 2048, feature_dim: int = 16, nu
fla/layers/rebased.py:25
Method
__init__
( self, hidden_size: int = 2048, expand_v: float = 1, head_dim: int = 128,
fla/layers/kda.py:83
Method
__init__
( self, hidden_size: int = 2048, expand_v: float = 2, head_dim: int = 256,
fla/layers/comba.py:82
Method
__init__
( self, hidden_size: int = 2048, expand_v: float = 2, head_dim: int = 256,
fla/layers/gated_deltanet.py:87
Method
__init__
( self, mode: str = 'chunk', hidden_size: int = 1024, expand_ratio: int | None
fla/layers/hgrn.py:31
Method
__init__
(self)
fla/models/utils.py:34
Method
__init__
( self, seen_tokens: int = 0, )
fla/models/utils.py:198
Method
__init__
(self, seen_tokens: int = 0, **kwargs)
fla/models/utils.py:366
Method
__init__
(self, *args, **kwargs)
fla/models/utils.py:475
Method
__init__
(self, *inputs, **kwargs)
fla/models/mla/modeling_mla.py:169
← previous
next →
1,001–1,100 of 2,989, ranked by callers