Functions — github.com/ByteDance-Seed/AHN

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

Functions289 in github.com/ByteDance-Seed/AHN

⨍ Functions 289◇ Types & classes 48 ↳ Endpoints 2

Methodforward

(self, hidden_states)

src/ahn/transformer/qwen2/modeling_qwen2.py:220

Methodforward

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

src/ahn/transformer/qwen2/modeling_qwen2.py:245

Methodforward

(self, x, position_ids)

src/ahn/transformer/qwen2/modeling_qwen2.py:308

Methodforward

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

src/ahn/transformer/qwen2/modeling_qwen2.py:472

Methodforward

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

src/ahn/transformer/qwen2/modeling_qwen2.py:771

Methodforward

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

src/ahn/transformer/qwen2/modeling_qwen2.py:887

Methodforward

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

src/ahn/transformer/qwen2/modeling_qwen2.py:992

Methodforward

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

src/ahn/transformer/qwen2/modeling_qwen2.py:1063

Methodforward

(self, *args, **kwargs)

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:208

Methodforward

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

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:521

Methodforward

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

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:779

Methodforward

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

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:1134

Methodforward

(self, *args, **kwargs)

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:211

Methodforward

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

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:524

Methodforward

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

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:772

Methodforward

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

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:1124

Methodforward

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

src/ahn/rnn/delta_net.py:176

Methodforward

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

src/ahn/rnn/mamba2.py:191

Methodforward

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

src/ahn/rnn/gated_deltanet.py:174

Methodfrom_pretrained

(cls, *args, **kwargs)

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:744

Methodget_beta_alpha

(self, hidden_states: torch.Tensor)

src/ahn/rnn/delta_net.py:172

Methodget_beta_alpha

(self, hidden_states: torch.Tensor)

src/ahn/rnn/gated_deltanet.py:167

Methodget_decoder

(self)

src/ahn/transformer/qwen3/modeling_qwen3.py:858

Methodget_decoder

(self)

src/ahn/transformer/qwen2/modeling_qwen2.py:764

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen3/modeling_qwen3.py:492

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen3/modeling_qwen3.py:843

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen3/modeling_qwen3.py:973

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen3/modeling_qwen3.py:1073

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen3/modeling_qwen3.py:1149

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen2/modeling_qwen2.py:464

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen2/modeling_qwen2.py:749

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen2/modeling_qwen2.py:879

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen2/modeling_qwen2.py:979

Methodget_input_embeddings

(self)

src/ahn/transformer/qwen2/modeling_qwen2.py:1055

Methodget_output_embeddings

(self)

src/ahn/transformer/qwen3/modeling_qwen3.py:849

Methodget_output_embeddings

(self)

src/ahn/transformer/qwen2/modeling_qwen2.py:755

Functionget_score_one_code_run

Returns the score of one example in Code.Run.

eval/longbench/metrics.py:165

Methodget_vocab

(self)

src/ahn/transformer/qwen2/tokenization_qwen2.py:215

Functionload_jsonl

(data_path)

eval/lveval/utils.py:57

Functionload_model_and_tokenizer_serial

(num_gpus, args)

eval/lveval/pred.py:153

Methodmem_forward_inference

Args: hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` attention_mask (

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:1362

Methodmem_forward_inference

Args: hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` attention_mask (

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:1352

Methodmem_forward_train

Args: hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` attention_mask (

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:1215

Methodmem_forward_train

Args: hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` attention_mask (

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:1205

Functionmultiple_processing_once

(num_gpus, dataset, shared_dict, device_dict, args)

eval/lveval/pred.py:131

Functionnew_backward

Custom backward function with retain_graph=True by default. Calls the original backward method.

src/ahn/utils.py:17

Methodprepare_for_tokenization

(self, text, **kwargs)

src/ahn/transformer/qwen2/tokenization_qwen2.py:337

Functionqa_f1_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:147

Functionqa_f1_score

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:103

Functionqa_f1_score_factrecall

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:110

Functionqa_f1_score_with_gold_ans

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:121

Functionqa_f1_zh_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:156

Functionqa_f1_zh_score

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:138

Functionqa_f1_zh_score_factrecall

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:147

Functionqa_f1_zh_score_with_gold_ans

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:159

Functionregister_custom_accelerator

()

src/ahn/utils.py:10

Functionreshape_into_chunks

Padding input_tensor with `pad_size` on the seq_len dim (dim=1) and simultaneously splitting it into chunk sequences. Assumes that we on

src/ahn/rnn/mamba2.py:71

Functionretrieval_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:68

Functionretrieval_zh_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:81

Functionrouge_zh_score

(prediction, ground_truth, **kwargs)

eval/longbench/metrics.py:129

Functionrouge_zh_score

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:73

Functionrouge_zh_score_blacklist

(prediction, ground_truth, gold_ans=None, **kwargs)

eval/lveval/metrics.py:79

Methodsample_num_attn_sinks

( sliding_window: int, candidates: list = [0, 32, 64, 128, 512, 2048,

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:479

Methodsample_num_attn_sinks

( sliding_window: int, candidates: list = [0, 32, 64, 128, 512, 2048,

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:482

Methodsample_window_size

( seq_len: str, candidates: list = [32, 64, 128, 256, 512, 1024, 2048,

src/ahn/transformer/qwen2_ahn/qwen2_ahn.py:459

Methodsample_window_size

( seq_len: str, candidates: list = [32, 64, 128, 256, 512, 1024, 2048,

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:462

Methodsave_pretrained

(self, save_directory, state_dict, **kwargs)

src/ahn/transformer/qwen3_ahn/qwen3_ahn.py:729

Methodsave_vocabulary

(self, save_directory: str, filename_prefix: Optional[str] = None)

src/ahn/transformer/qwen2/tokenization_qwen2_fast.py:132

Methodsave_vocabulary

(self, save_directory: str, filename_prefix: Optional[str] = None)

src/ahn/transformer/qwen2/tokenization_qwen2.py:308

Functionsegment_sum

More stable segment sum calculation. Uses cumulative sums and masking instead of direct subtractions.

src/ahn/rnn/mamba2.py:92

Methodset_decoder

(self, decoder)

src/ahn/transformer/qwen3/modeling_qwen3.py:855

Methodset_decoder

(self, decoder)

src/ahn/transformer/qwen2/modeling_qwen2.py:761

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen3/modeling_qwen3.py:495

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen3/modeling_qwen3.py:846

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen3/modeling_qwen3.py:976

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen3/modeling_qwen3.py:1076

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen3/modeling_qwen3.py:1152

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen2/modeling_qwen2.py:467

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen2/modeling_qwen2.py:752

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen2/modeling_qwen2.py:882

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen2/modeling_qwen2.py:982

Methodset_input_embeddings

(self, value)

src/ahn/transformer/qwen2/modeling_qwen2.py:1058

Methodset_output_embeddings

(self, new_embeddings)

src/ahn/transformer/qwen3/modeling_qwen3.py:852

Methodset_output_embeddings

(self, new_embeddings)

src/ahn/transformer/qwen2/modeling_qwen2.py:758

Functionstr_to_dtype

(name: str)

examples/scripts/utils/merge_weights.py:47

Functionsum_norm

(x)

src/ahn/rnn/delta_net.py:37

Functionsum_norm

(x)

src/ahn/rnn/gated_deltanet.py:41

Methodtrim_cache

(self, num_attn_sinks: int, num_cached_tokens: int)

src/ahn/utils.py:137

Methodvocab_size

(self)

src/ahn/transformer/qwen2/tokenization_qwen2.py:211

← previous201–289 of 289, ranked by callers