MCPcopy Create free account

hub / github.com/OpenSparseLLMs/Linear-MoE / functions

Functions1,498 in github.com/OpenSparseLLMs/Linear-MoE

⨍ Functions 1,498◇ Types & classes 227
Method__init__
(self, num_local_experts, config: TransformerConfig, submodules: MLPSubmodules)
linear_moe/model/deepseek_v2/moe/experts.py:185
Method__init__
(self, config: TransformerConfig, layer_number: int = None)
linear_moe/model/deepseek_v2/moe/moe_layer.py:38
Method__init__
Initialize the zero token dropping router. Args: config (TransformerConfig): The configuration for the transformer model.
linear_moe/model/deepseek_v2/moe/router.py:183
Method__init__
(self)
linear_moe/model/llama3/transformer_layer.py:60
Method__init__
( self, config: TransformerConfig, submodules: TransformerLayerSubmodules, lay
linear_moe/model/llama3/transformer_layer.py:71
Method__init__
Llama3RMSNorm is equivalent to T5LayerNorm
linear_moe/model/llama3/rms_norm.py:5
Method__init__
(self, config, num_tokentypes=0, parallel_output=True,
linear_moe/model/llama3/gpt_model.py:58
Method__init__
(self, hidden_size, vocab_size, max_sequence_length,
linear_moe/model/llama3/language_model.py:133
Method__init__
(self, config, encoder_attn_mask_type, num_tokentypes=0,
linear_moe/model/llama3/language_model.py:327
Method__init__
(self, config, is_expert=False)
linear_moe/model/llama3/transformer_legacy.py:101
Method__init__
(self, config)
linear_moe/model/llama3/transformer_legacy.py:203
Method__init__
(self, layer_number, config, attn_mask_type=AttnMaskType.padding)
linear_moe/model/llama3/transformer_legacy.py:313
Method__init__
(self, causal=False, softmax_scale=None, attention_dropout=0.0, device=None, dtype=None)
linear_moe/model/llama3/transformer_legacy.py:456
Method__init__
(self, config, layer_number, attention_type=AttnType.self_attn, attn_mask_ty
linear_moe/model/llama3/transformer_legacy.py:515
Method__init__
(self, config, layer_number, layer_type=LayerType.encoder, self_attn_mask_ty
linear_moe/model/llama3/transformer_legacy.py:870
Method__init__
(self, layer_number)
linear_moe/model/llama3/transformer_legacy.py:1310
Method__init__
(self, config, model_type, layer_type=LayerType.encoder, self_attn_mask_type
linear_moe/model/llama3/transformer_legacy.py:1394
Method__init__
( self, config: TransformerConfig, transformer_layer_spec: ModuleSpec, vocab_s
linear_moe/model/llama3/model.py:53
Method__init__
( self, config: TransformerConfig, submodules: HybridTransformerBlockSubmodules,
linear_moe/model/llama3/hybrid/hybrid_transformer_block.py:90
Method__init__
( self, config: TransformerConfig, hybrid_transformer_layer_spec: ModuleSpec,
linear_moe/model/llama3/hybrid/hybrid_model.py:47
Method__init__
( self, config: TransformerConfig, submodules: SelfAttentionSubmodules, layer_
linear_moe/model/llama3/transformer/attention.py:356
Method__init__
( self, config: TransformerConfig, submodules: CrossAttentionSubmodules, layer
linear_moe/model/llama3/transformer/attention.py:522
Method__init__
( self, config: TransformerConfig, submodules: MLPSubmodules, is_expert: bool
linear_moe/model/llama3/transformer/mlp.py:58
Method__init__
( self, hidden_size: int, elementwise_affine: bool = True, bias: bool = False,
linear_moe/model/common_modules/layernorm.py:538
Method__init__
( self, num_groups: int, hidden_size: int, elementwise_affine: bool = True,
linear_moe/model/common_modules/layernorm.py:580
Method__init__
( self, hidden_size, elementwise_affine: bool = True, bias: bool = False,
linear_moe/model/common_modules/layernorm.py:804
Method__init__
( self, num_groups: int, hidden_size: int, elementwise_affine: bool = True,
linear_moe/model/common_modules/layernorm.py:849
Method__init__
( self, hidden_size, elementwise_affine: bool = True, bias: bool = False,
linear_moe/model/common_modules/layernorm.py:900
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).
linear_moe/model/common_modules/rotary.py:155
Method__init__
( self, head_dim: int )
linear_moe/model/common_modules/feature_map.py:45
Method__init__
( self, head_dim: int, nu: int = 4 )
linear_moe/model/common_modules/feature_map.py:104
Method__init__
( self, head_dim: int )
linear_moe/model/common_modules/feature_map.py:120
Method__init__
( self, head_dim: int, feature_dim: int )
linear_moe/model/common_modules/feature_map.py:134
Method__init__
( self, head_dim: int, sketch_size: Optional[int] = None, degree: Optional[int
linear_moe/model/common_modules/feature_map.py:151
Method__init__
( self, head_dim: int )
linear_moe/model/common_modules/feature_map.py:191
Method__init__
( self, head_dim: int, use_gamma: Optional[bool] = True, use_beta: Optional[bo
linear_moe/model/common_modules/feature_map.py:208
Method__init__
( self, )
linear_moe/model/common_modules/feature_map.py:250
Method__init__
( self, )
linear_moe/model/common_modules/feature_map.py:261
Method__init__
( self, )
linear_moe/model/common_modules/feature_map.py:272
Method__init__
( self, )
linear_moe/model/common_modules/feature_map.py:283
Method__init__
( self, )
linear_moe/model/common_modules/feature_map.py:294
Method__init__
Initializes the dataset. Args: path(str): The path of the dataset file. tokenizer(object): The tokenizer obje
linear_moe/data/bloom.py:23
Method__init__
(self, path, max_padding_length, split='train')
linear_moe/data/llama.py:63
Method__init__
(self, name, data_prefix, documents, index
linear_moe/data/llama.py:228
Method__init__
Initializes the dataset. Args: path(str): The path of the dataset file. tokenizer(object): The tokenizer obje
linear_moe/data/glm.py:29
Method__init__
(self, datapaths, tokenizer, max_source_seq_length, max_target_seq_length)
linear_moe/data/glm.py:129
Method__init__
(self, datapaths, max_source_length, max_target_length)
linear_moe/data/glm.py:248
Method__init__
(self, data_path: str)
linear_moe/data/qwen_vl.py:102
Method__init__
Args: datapath (str): The path of the dataset. max_padding_length (int): The maximum length to pad the input sequence
linear_moe/data/starcoder.py:25
Method__init__
( self, do_resize: bool = True, size: Dict[str, int] = None, resample: PILImag
linear_moe/data/llava/cvcuda_image_processing_clip.py:101
Method__init__
(self, data_path: str)
linear_moe/data/llava/mm_pretrain_dataset.py:176
Method__init__
(self, keywords, tokenizer, input_ids)
linear_moe/data/llava/mm_utils.py:88
Method__init__
( self, vocab_file, errors="replace", image_start_tag='<img>', image_e
linear_moe/tokenizer/tokenization_qwen_vl.py:97
Method__init__
(self, img_rgb, metadata=None, scale=1.0)
linear_moe/tokenizer/tokenization_qwen_vl.py:489
Method__init__
( self, vocab_file, unk_token="<unk>", bos_token="<s>", eos_token="</s
linear_moe/tokenizer/tokenization_baichuan.py:37
Method__init__
(self, tokenizer_json_file)
linear_moe/tokenizer/jiebabpe_tokenizer.py:21
Method__init__
(self, tokenizer_path, extra_vocab_size)
linear_moe/tokenizer/__init__.py:237
Method__init__
(self, tokenizer_path, extra_vocab_size)
linear_moe/tokenizer/__init__.py:327
Method__init__
(self, tokenizer_path, extra_vocab_size)
linear_moe/tokenizer/__init__.py:400
Method__init__
(self, name)
linear_moe/tokenizer/icetk_glm130b_tokenizer.py:29
Method__init__
( self, path='~/.icetk_models', max_blank_length=80, byte_fallback=True, )
linear_moe/tokenizer/icetk_glm130b_tokenizer.py:85
Method__init__
( self, vocab_file, unk_token="<unk>", bos_token="<|startoftext|>", eo
linear_moe/tokenizer/tokenization_yi.py:34
Method__len__
(self)
linear_moe/data/bloom.py:40
Method__len__
(self)
linear_moe/data/llama.py:117
Method__len__
(self)
linear_moe/data/llama.py:264
Method__len__
Returns the number of sequences in the dataset.
linear_moe/data/glm.py:79
Method__len__
(self)
linear_moe/data/glm.py:143
Method__len__
(self)
linear_moe/data/glm.py:259
Method__len__
(self)
linear_moe/data/qwen_vl.py:110
Method__len__
(self)
linear_moe/data/starcoder.py:80
Method__len__
(self)
linear_moe/data/llava/mm_pretrain_dataset.py:194
Method__len__
(self)
linear_moe/tokenizer/tokenization_qwen_vl.py:170
Method__post_init__
(self)
linear_moe/model/mixtral/transformer_config.py:111
Method__post_init__
(self)
linear_moe/model/qwen2/transformer_config.py:111
Method__post_init__
(self)
linear_moe/model/llama3/transformer_config.py:111
Method__repr__
(self)
linear_moe/sequence_modeling/rwkv7/lora_mlp.py:45
Method__repr__
(self)
linear_moe/sequence_modeling/rwkv6/dd_lerp_linear.py:31
Method__repr__
(self)
linear_moe/sequence_modeling/rwkv6/dd_lerp_linear.py:68
Method__repr__
(self)
linear_moe/sequence_modeling/rwkv6/dd_lerp_linear.py:106
Method__repr__
(self)
linear_moe/model/common_modules/layernorm.py:558
Method__repr__
(self)
linear_moe/model/common_modules/layernorm.py:605
Method__repr__
(self)
linear_moe/model/common_modules/layernorm.py:648
Method__repr__
(self)
linear_moe/model/common_modules/layernorm.py:824
Method__repr__
(self)
linear_moe/model/common_modules/layernorm.py:874
Method__repr__
(self)
linear_moe/model/common_modules/layernorm.py:920
Method__repr__
(self)
linear_moe/model/common_modules/feature_map.py:90
Method__setstate__
(self, d)
linear_moe/tokenizer/tokenization_baichuan.py:77
Method__setstate__
(self, d)
linear_moe/tokenizer/tokenization_yi.py:90
Method_add_tokens
(self, new_tokens: Union[List[str], List[AddedToken]], special_tokens: bool = False)
linear_moe/tokenizer/tokenization_qwen_vl.py:192
Function_bias_dropout_add
(x, bias, residual, prob)
linear_moe/model/llama3/transformer_legacy.py:842
Function_bwd_dm_calculate
( Q, DO, DKV, n: tl.constexpr, d: tl.constexpr, e: tl.constexpr, BLOCK: tl.constex
linear_moe/sequence_modeling/lasp2/lasp2_with_mask_triton_op.py:298
Function_bwd_dm_calculate
( Q, DO, DKV, n: tl.constexpr, d: tl.constexpr, e: tl.constexpr, BLOCK: tl.constex
linear_moe/sequence_modeling/lasp2/lasp2_without_mask_triton_op.py:207
Function_bwd_dm_cumsum
( DKV, d: tl.constexpr, e: tl.constexpr, NUM_BLOCK: tl.constexpr, D_FBLOCK: tl.constexpr,
linear_moe/sequence_modeling/lasp2/lasp2_with_mask_triton_op.py:367
Function_bwd_dm_cumsum
( DKV, d: tl.constexpr, e: tl.constexpr, NUM_BLOCK: tl.constexpr, D_FBLOCK: tl.constexpr,
linear_moe/sequence_modeling/lasp2/lasp2_without_mask_triton_op.py:276
Function_bwd_dm_update
( DKV, GDKV, d: tl.constexpr, e: tl.constexpr, NUM_BLOCK: tl.constexpr, D_FBLOCK: tl.c
linear_moe/sequence_modeling/lasp2/lasp2_with_mask_triton_op.py:403
Function_bwd_dm_update
( DKV, GDKV, d: tl.constexpr, e: tl.constexpr, NUM_BLOCK: tl.constexpr, D_FBLOCK: tl.c
linear_moe/sequence_modeling/lasp2/lasp2_without_mask_triton_op.py:312
Function_bwd_inter_kernel
( K, V, DO, DQ, DK, DV, KV, DKV, n: tl.constexpr, d: tl.constexpr,
linear_moe/sequence_modeling/lasp2/lasp2_with_mask_triton_op.py:593
Function_bwd_intra_kernel
( Q, K, V, DO, DQ, DK, DV, n: tl.constexpr, d: tl.constexpr, e: tl.con
linear_moe/sequence_modeling/lasp2/lasp2_with_mask_triton_op.py:446
Function_bwd_kernel
( K, V, DO, DQ, DK, DV, KV, DKV, n: tl.constexpr, d: tl.constexpr,
linear_moe/sequence_modeling/lasp2/lasp2_without_mask_triton_op.py:355
Method_checkpointed_attention_forward
Forward method with selective activation checkpointing.
linear_moe/model/deepseek_v2/transformer/attention.py:125
← previousnext →801–900 of 1,498, ranked by callers