Functions — github.com/ByteDance-Seed/AHN

CodeHub MCPcopy Create free account

hub / github.com/ByteDance-Seed/AHN / functions

Functions289 in github.com/ByteDance-Seed/AHN

⨍ Functions 289◇ Types & classes 48 ↳ Endpoints 2

↓ 1 callersFunctionprocess_args

eval/lveval/pred.py:222

↓ 1 callersFunctionprocess_pt_shard

(src: Path, dst: Path, old: str, new: str, dry: bool)

examples/scripts/utils/rename_sd.py:50

↓ 1 callersFunctionprocess_pt_shard

(src: Path, dst: Path, old: str, new: str, dry: bool)

examples/scripts/deprecated/rename_sd.py:50

↓ 1 callersFunctionprocess_safetensors_shard

(src: Path, dst: Path, old: str, new: str, dry: bool)

examples/scripts/utils/rename_sd.py:25

↓ 1 callersFunctionprocess_safetensors_shard

(src: Path, dst: Path, old: str, new: str, dry: bool)

examples/scripts/deprecated/rename_sd.py:25

↓ 1 callersFunctionread_index_json

If an index file exists (for sharded checkpoints), return {param_key: shard_file}.

examples/scripts/utils/split_weights.py:33

↓ 1 callersFunctionremove_articles

eval/longbench/metrics.py:25

↓ 1 callersFunctionremove_articles

eval/lveval/metrics.py:33

↓ 1 callersMethodreset_parameters

src/ahn/utils.py:62

↓ 1 callersFunctionrouge_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:120

↓ 1 callersFunctionscorer

(dataset, predictions, answers, all_classes)

eval/longbench/eval.py:89

↓ 1 callersFunctionscorer

(dataset, predictions, answers, gold_anss)

eval/lveval/eval.py:22

↓ 1 callersFunctionscorer_e

(dataset, predictions, answers, lengths, all_classes)

eval/longbench/eval.py:65

↓ 1 callersFunctionseed_everything

eval/longbench/pred.py:217

↓ 1 callersFunctionseed_everything

eval/lveval/utils.py:28

↓ 1 callersFunctionsingle_processing

(datasets, args)

eval/lveval/pred.py:190

↓ 1 callersFunctionstr_to_dtype

examples/scripts/inference.py:60

↓ 1 callersMethodtrim_cache

Evict key-value tokens from the cache after compression into memory.

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:1086

↓ 1 callersMethodtrim_cache

Evict key-value tokens from the cache after compression into memory.

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:1075

↓ 1 callersFunctiontruncate_prompt

(tokenizer, prompt, max_length)

eval/lveval/utils.py:133

( self, in_features: int, out_features: int, num_groups: int, use_bias

src/ahn/utils.py:30

(self, num_heads: int, head_dim: int, use_dimwise_pos: bool = False)

src/ahn/utils.py:184

Qwen3RMSNorm is equivalent to T5LayerNorm

src/ahn/transformer/qwen3/modeling_qwen3.py:64

(self, config: Qwen3Config, layer_idx: int)

src/ahn/transformer/qwen3/modeling_qwen3.py:174

(self, config: Qwen3Config, layer_idx: int)

src/ahn/transformer/qwen3/modeling_qwen3.py:258

(self, config: Qwen3Config, device=None)

src/ahn/transformer/qwen3/modeling_qwen3.py:317

(self, config: Qwen3Config)

src/ahn/transformer/qwen3/modeling_qwen3.py:476

src/ahn/transformer/qwen3/modeling_qwen3.py:834

src/ahn/transformer/qwen3/modeling_qwen3.py:964

src/ahn/transformer/qwen3/modeling_qwen3.py:1057

src/ahn/transformer/qwen3/modeling_qwen3.py:1141

( self, vocab_size=151936, hidden_size=4096, intermediate_size=22016,

src/ahn/transformer/qwen3/configuration_qwen3.py:152

(self, config: Qwen2Config, layer_idx: int)

src/ahn/transformer/qwen2/modeling_qwen2.py:138

Qwen2RMSNorm is equivalent to T5LayerNorm

src/ahn/transformer/qwen2/modeling_qwen2.py:212

(self, config: Qwen2Config, layer_idx: int)

src/ahn/transformer/qwen2/modeling_qwen2.py:232

(self, config: Qwen2Config, device=None)

src/ahn/transformer/qwen2/modeling_qwen2.py:289

(self, config: Qwen2Config)

src/ahn/transformer/qwen2/modeling_qwen2.py:448

src/ahn/transformer/qwen2/modeling_qwen2.py:740

src/ahn/transformer/qwen2/modeling_qwen2.py:870

src/ahn/transformer/qwen2/modeling_qwen2.py:963

src/ahn/transformer/qwen2/modeling_qwen2.py:1047

( self, vocab_file=None, merges_file=None, tokenizer_file=None, unk_to

src/ahn/transformer/qwen2/tokenization_qwen2_fast.py:83

( self, vocab_size=151936, hidden_size=4096, intermediate_size=22016,

src/ahn/transformer/qwen2/configuration_qwen2.py:132

( self, vocab_file, merges_file, errors="replace", unk_token="<|endoft

src/ahn/transformer/qwen2/tokenization_qwen2.py:137

( self, loss_type: Optional[str] = "ce", _layer_implementation: Optional[str] = "Qwen2

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:74

(self, *args, **kwargs)

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:113

(self, config, do_train: bool = False)

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:720

( self, config: Qwen2Config, layer_idx: int, ahn_cls_name: str = "GatedDeltaNet" )

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:1049

( self, loss_type: Optional[str] = "ce", _layer_implementation: Optional[str] = "Qwen3

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:76

(self, *args, **kwargs)

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:115

(self, config, do_train: bool = False)

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:713

( self, config: Qwen3Config, layer_idx: int, ahn_cls_name: str = "GatedDeltaNe

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:1034

( self, mode: str = 'chunk', d_model: int = None, hidden_size: int = 1024,

src/ahn/rnn/delta_net.py:82

( self, num_heads: int = 8, head_dim: int = 128, hidden_size: int = 1024,

src/ahn/rnn/mamba2.py:120

( self, hidden_size: int = 2048, expand_v: float = 1, head_dim: int = 256,

src/ahn/rnn/gated_deltanet.py:93

Method_convert_id_to_token

Converts an index (integer) in a token (str) using the vocab.

src/ahn/transformer/qwen2/tokenization_qwen2.py:278

Method_convert_token_to_id

Converts a token (str) in an id using the vocab.

src/ahn/transformer/qwen2/tokenization_qwen2.py:273

Method_init_weights

src/ahn/transformer/qwen3/modeling_qwen3.py:385

Method_init_weights

src/ahn/transformer/qwen2/modeling_qwen2.py:357

Method_init_weights

(model: nn.Module)

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:755

Method_init_weights

(model: nn.Module)

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:748

Function_init_worker

eval/lveval/pred.py:41

src/train.py:26

Method_tokenize

Tokenize a string.

src/ahn/transformer/qwen2/tokenization_qwen2.py:262

Method_update_model_kwargs_for_generation

( self, outputs: ModelOutput, model_kwargs: Dict[str, Any], is_encoder_decoder

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:988

Method_update_model_kwargs_for_generation

( self, outputs: ModelOutput, model_kwargs: Dict[str, Any], is_encoder_decoder

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:973

Functionapply_mask_to_padding_states

Tunes out the hidden states for padding tokens, see https://github.com/state-spaces/mamba/issues/66

src/ahn/rnn/mamba2.py:49

Functionclassification_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:104

Functioncode_sim_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:94

Functioncompare_models

(new_model, model, atol=1e-6, rtol=1e-5)

examples/scripts/utils/model_compare.py:3

Methodconvert_tokens_to_string

Converts a sequence of tokens (string) in a single string.

src/ahn/transformer/qwen2/tokenization_qwen2.py:283

Functioncount_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:58

Methodctx_sliding_window_mask

(b, h, q_idx, kv_idx)

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:428

Methodctx_sliding_window_mask

(b, h, q_idx, kv_idx)

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:431

Functioncustom_sort

eval/lveval/eval.py:17

Methoddefault_forward

( self, hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor] = None,

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:1167

Methoddefault_forward

( self, hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor] = None,

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:1157

Functiondump_preds_results_once

(pred, save_path)

eval/lveval/utils.py:45

Functioneager_attention_forward

( module: nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attentio

src/ahn/transformer/qwen3/modeling_qwen3.py:145

Functioneager_attention_forward

( module: nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attentio

src/ahn/transformer/qwen2/modeling_qwen2.py:109

src/ahn/rnn/delta_net.py:33

src/ahn/rnn/gated_deltanet.py:37

Functionevaluate

eval/lveval/pred.py:88

Methodextra_repr

src/ahn/transformer/qwen3/modeling_qwen3.py:79

Methodextra_repr

src/ahn/transformer/qwen2/modeling_qwen2.py:227

src/ahn/utils.py:65

( self, hidden_states: torch.Tensor, q_states: torch.Tensor, k_states: torch.T

src/ahn/utils.py:158

( self, pos_ratio: torch.FloatTensor, h_mem: torch.Tensor, h_local: torch.Tensor )

src/ahn/utils.py:198

(self, hidden_states)

src/ahn/transformer/qwen3/modeling_qwen3.py:72

src/ahn/transformer/qwen3/modeling_qwen3.py:94

( self, hidden_states: torch.Tensor, position_embeddings: Tuple[torch.Tensor, torch.Te

src/ahn/transformer/qwen3/modeling_qwen3.py:206

( self, hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor] = None,

src/ahn/transformer/qwen3/modeling_qwen3.py:273

(self, x, position_ids)

src/ahn/transformer/qwen3/modeling_qwen3.py:336

( self, input_ids: Optional[torch.LongTensor] = None, attention_mask: Optional[torch.T

src/ahn/transformer/qwen3/modeling_qwen3.py:500

r""" labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the masked lan

src/ahn/transformer/qwen3/modeling_qwen3.py:865

r""" labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for computing the sequence classification/regression

src/ahn/transformer/qwen3/modeling_qwen3.py:981

r""" labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for computing the sequence classification/regression

src/ahn/transformer/qwen3/modeling_qwen3.py:1086

r""" start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for position (index) of the start of the la

src/ahn/transformer/qwen3/modeling_qwen3.py:1157

src/ahn/transformer/qwen2/modeling_qwen2.py:58

( self, hidden_states: torch.Tensor, position_embeddings: Tuple[torch.Tensor, torch.Te

src/ahn/transformer/qwen2/modeling_qwen2.py:152

← previous next →101–200 of 289, ranked by callers