SYMBOL INDEX (271 symbols across 24 files)

FILE: air_llm/airllm/airllm.py
  class AirLLMLlama2 (line 7) | class AirLLMLlama2(AirLLMBaseModel):
    method __init__ (line 8) | def __init__(self, *args, **kwargs):

FILE: air_llm/airllm/airllm_baichuan.py
  class AirLLMBaichuan (line 10) | class AirLLMBaichuan(AirLLMBaseModel):
    method __init__ (line 13) | def __init__(self, *args, **kwargs):
    method get_use_better_transformer (line 18) | def get_use_better_transformer(self):
    method get_tokenizer (line 20) | def get_tokenizer(self, hf_token=None):
    method get_generation_config (line 24) | def get_generation_config(self):

FILE: air_llm/airllm/airllm_base.py
  class AirLLMBaseModel (line 46) | class AirLLMBaseModel(GenerationMixin):
    method set_layer_names_dict (line 49) | def set_layer_names_dict(self):
    method __init__ (line 57) | def __init__(self, model_local_path_or_repo_id, device="cuda:0", dtype...
    method get_generation_config (line 163) | def get_generation_config(self):
    method get_tokenizer (line 172) | def get_tokenizer(self, hf_token=None):
    method get_use_better_transformer (line 178) | def get_use_better_transformer(self):
    method init_model (line 181) | def init_model(self):
    method set_layers_from_layer_names (line 240) | def set_layers_from_layer_names(self):
    method load_rotary_pos_emb_to_device (line 265) | def load_rotary_pos_emb_to_device(self):
    method load_layer_to_cpu (line 269) | def load_layer_to_cpu(self, layer_name):
    method move_layer_to_device (line 302) | def move_layer_to_device(self, state_dict):
    method can_generate (line 326) | def can_generate(self):
    method prepare_inputs_for_generation (line 329) | def prepare_inputs_for_generation(
    method __call__ (line 368) | def __call__(self, *args, **kwargs):
    method get_past_key_values_cache_seq_len (line 371) | def get_past_key_values_cache_seq_len(self, past_key_values):
    method get_sequence_len (line 373) | def get_sequence_len(self, seq):
    method get_pos_emb_args (line 376) | def get_pos_emb_args(self, len_p, len_s):
    method get_past_key_value_args (line 379) | def get_past_key_value_args(self, k_cache, v_cache):
    method get_attention_mask_args (line 382) | def get_attention_mask_args(self, full_attention_mask, len_p, len_s):
    method get_position_ids_args (line 385) | def get_position_ids_args(self, full_position_ids, len_p, len_s):
    method run_lm_head (line 390) | def run_lm_head(self, layer, seq):
    method run_norm (line 393) | def run_norm(self, layer, seq):
    method forward (line 396) | def forward(

FILE: air_llm/airllm/airllm_chatglm.py
  class AirLLMChatGLM (line 8) | class AirLLMChatGLM(AirLLMBaseModel):
    method __init__ (line 11) | def __init__(self, *args, **kwargs):
    method get_use_better_transformer (line 16) | def get_use_better_transformer(self):
    method get_generation_config (line 19) | def get_generation_config(self):
    method get_sequence_len (line 22) | def get_sequence_len(self, seq):
    method get_past_key_values_cache_seq_len (line 25) | def get_past_key_values_cache_seq_len(self, past_key_values):
    method set_layer_names_dict (line 30) | def set_layer_names_dict(self):
    method get_pos_emb_args (line 37) | def get_pos_emb_args(self, len_p, len_s):
    method get_past_key_value_args (line 45) | def get_past_key_value_args(self, k_cache, v_cache):
    method get_attention_mask_args (line 48) | def get_attention_mask_args(self, full_attention_mask, len_p, len_s):
    method get_position_ids_args (line 51) | def get_position_ids_args(self, full_position_ids, len_p, len_s):

FILE: air_llm/airllm/airllm_internlm.py
  class AirLLMInternLM (line 8) | class AirLLMInternLM(AirLLMBaseModel):
    method __init__ (line 11) | def __init__(self, *args, **kwargs):
    method get_use_better_transformer (line 16) | def get_use_better_transformer(self):
    method get_generation_config (line 18) | def get_generation_config(self):

FILE: air_llm/airllm/airllm_llama_mlx.py
  class ModelArgs (line 23) | class ModelArgs:
  function sanitize_config (line 35) | def sanitize_config(config, weights=None):
  function get_model_args_from_config (line 53) | def get_model_args_from_config(config):
  class RMSNorm (line 71) | class RMSNorm(nn.Module):
    method __init__ (line 72) | def __init__(self, dims: int, eps: float = 1e-5):
    method _norm (line 77) | def _norm(self, x):
    method __call__ (line 80) | def __call__(self, x):
  class Attention (line 85) | class Attention(nn.Module):
    method __init__ (line 86) | def __init__(self, args: ModelArgs):
    method __call__ (line 105) | def __call__(
  class FeedForward (line 144) | class FeedForward(nn.Module):
    method __init__ (line 145) | def __init__(self, args: ModelArgs):
    method __call__ (line 152) | def __call__(self, x) -> mx.array:
  class TransformerBlock (line 156) | class TransformerBlock(nn.Module):
    method __init__ (line 157) | def __init__(self, args: ModelArgs):
    method __call__ (line 167) | def __call__(
  function sample (line 179) | def sample(logits, temperature=0):
  class AirLLMLlamaMlx (line 185) | class AirLLMLlamaMlx:
    method set_layer_names_dict (line 188) | def set_layer_names_dict(self):
    method record_memory (line 195) | def record_memory(self, msg=None):
    method __init__ (line 210) | def __init__(self, model_local_path_or_repo_id, device="cuda:0", dtype...
    method get_tokenizer (line 245) | def get_tokenizer(self, hf_token=None):
    method generate (line 252) | def generate(self, x, temperature=0, max_new_tokens=None, **kwargs):
    method model_generate (line 265) | def model_generate(self, x, temperature=0, max_new_tokens=None):

FILE: air_llm/airllm/airllm_mistral.py
  class AirLLMMistral (line 8) | class AirLLMMistral(AirLLMBaseModel):
    method __init__ (line 11) | def __init__(self, *args, **kwargs):
    method get_use_better_transformer (line 16) | def get_use_better_transformer(self):
    method get_generation_config (line 18) | def get_generation_config(self):

FILE: air_llm/airllm/airllm_mixtral.py
  class AirLLMMixtral (line 8) | class AirLLMMixtral(AirLLMBaseModel):
    method __init__ (line 11) | def __init__(self, *args, **kwargs):
    method get_use_better_transformer (line 16) | def get_use_better_transformer(self):
    method get_generation_config (line 19) | def get_generation_config(self):

FILE: air_llm/airllm/airllm_qwen.py
  class AirLLMQWen (line 8) | class AirLLMQWen(AirLLMBaseModel):
    method __init__ (line 11) | def __init__(self, *args, **kwargs):
    method get_use_better_transformer (line 16) | def get_use_better_transformer(self):
    method get_generation_config (line 18) | def get_generation_config(self):
    method get_past_key_values_cache_seq_len (line 22) | def get_past_key_values_cache_seq_len(self, past_key_values):
    method set_layer_names_dict (line 27) | def set_layer_names_dict(self):
    method get_pos_emb_args (line 33) | def get_pos_emb_args(self, len_p, len_s):
    method get_past_key_value_args (line 49) | def get_past_key_value_args(self, k_cache, v_cache):
    method get_attention_mask_args (line 52) | def get_attention_mask_args(self, full_attention_mask, len_p, len_s):
    method get_position_ids_args (line 55) | def  get_position_ids_args(self, full_position_ids, len_p, len_s):

FILE: air_llm/airllm/airllm_qwen2.py
  class AirLLMQWen2 (line 9) | class AirLLMQWen2(AirLLMBaseModel):
    method __init__ (line 12) | def __init__(self, *args, **kwargs):
    method get_use_better_transformer (line 17) | def get_use_better_transformer(self):

FILE: air_llm/airllm/auto_model.py
  class AutoModel (line 13) | class AutoModel:
    method __init__ (line 14) | def __init__(self):
    method get_module_class (line 20) | def get_module_class(cls, pretrained_model_name_or_path, *inputs, **kw...
    method from_pretrained (line 48) | def from_pretrained(cls, pretrained_model_name_or_path, *inputs, **kwa...

FILE: air_llm/airllm/persist/mlx_model_persister.py
  function map_torch_to_mlx (line 16) | def map_torch_to_mlx(model):
  class MlxModelPersister (line 57) | class MlxModelPersister(ModelPersister):
    method __init__ (line 60) | def __init__(self, *args, **kwargs):
    method model_persist_exist (line 66) | def model_persist_exist(self, layer_name, saving_path):
    method persist_model (line 77) | def persist_model(self, state_dict, layer_name, saving_path):
    method load_model (line 91) | def load_model(self, layer_name, path):

FILE: air_llm/airllm/persist/model_persister.py
  class ModelPersister (line 6) | class ModelPersister:
    method __init__ (line 7) | def __init__(self):
    method get_model_persister (line 11) | def get_model_persister(cls):
    method model_persist_exist (line 32) | def model_persist_exist(self, layer_name, saving_path):
    method persist_model (line 35) | def persist_model(self, state_dict, layer_name, path):
    method load_model (line 38) | def load_model(self, layer_name, path):

FILE: air_llm/airllm/persist/safetensor_model_persister.py
  class SafetensorModelPersister (line 11) | class SafetensorModelPersister(ModelPersister):
    method __init__ (line 14) | def __init__(self, *args, **kwargs):
    method model_persist_exist (line 20) | def model_persist_exist(self, layer_name, saving_path):
    method persist_model (line 27) | def persist_model(self, state_dict, layer_name, saving_path):
    method load_model (line 36) | def load_model(self, layer_name, path):

FILE: air_llm/airllm/profiler.py
  class LayeredProfiler (line 5) | class LayeredProfiler:
    method __init__ (line 6) | def __init__(self, print_memory=False):
    method add_profiling_time (line 12) | def add_profiling_time(self, item, time):
    method clear_profiling_time (line 24) | def clear_profiling_time(self):
    method print_profiling_time (line 28) | def print_profiling_time(self):

FILE: air_llm/airllm/tokenization_baichuan.py
  class BaichuanTokenizer (line 43) | class BaichuanTokenizer(PreTrainedTokenizer):
    method __init__ (line 57) | def __init__(
    method __getstate__ (line 92) | def __getstate__(self):
    method __setstate__ (line 97) | def __setstate__(self, d):
    method vocab_size (line 103) | def vocab_size(self):
    method get_vocab (line 107) | def get_vocab(self):
    method _tokenize (line 113) | def _tokenize(self, text):
    method _convert_token_to_id (line 117) | def _convert_token_to_id(self, token):
    method _convert_id_to_token (line 121) | def _convert_id_to_token(self, index):
    method convert_tokens_to_string (line 126) | def convert_tokens_to_string(self, tokens):
    method save_vocabulary (line 145) | def save_vocabulary(self, save_directory, filename_prefix: Optional[st...
    method build_inputs_with_special_tokens (line 172) | def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=No...
    method get_special_tokens_mask (line 183) | def get_special_tokens_mask(
    method create_token_type_ids_from_sequences (line 220) | def create_token_type_ids_from_sequences(

FILE: air_llm/airllm/utils.py
  function save_quant_state_to_dict (line 40) | def save_quant_state_to_dict(self, packed=True):
  class NotEnoughSpaceException (line 71) | class NotEnoughSpaceException(Exception):
  function clean_memory (line 75) | def clean_memory():
  function uncompress_layer_state_dict (line 85) | def uncompress_layer_state_dict(layer_state_dict):
  function load_layer (line 115) | def load_layer(local_path, layer_name, profiling=False):
  function check_space (line 134) | def check_space(checkpoint_path, layer_shards_saving_path=None, compress...
  function compress_layer_state_dict (line 157) | def compress_layer_state_dict(layer_state_dict, compression=None):
  function remove_real_and_linked_file (line 178) | def remove_real_and_linked_file(to_delete):
  function split_and_save_layers (line 188) | def split_and_save_layers(checkpoint_path, layer_shards_saving_path=None...
  function find_or_create_local_splitted_path (line 341) | def find_or_create_local_splitted_path(model_local_path_or_repo_id, laye...

FILE: air_llm/setup.py
  class PostInstallCommand (line 7) | class PostInstallCommand(install):
    method run (line 8) | def run(self):

FILE: air_llm/tests/test_automodel.py
  class TestAutoModel (line 10) | class TestAutoModel(unittest.TestCase):
    method setUp (line 11) | def setUp(self):
    method tearDown (line 13) | def tearDown(self):
    method test_auto_model_should_return_correct_model (line 16) | def test_auto_model_should_return_correct_model(self):

FILE: air_llm/tests/test_compression.py
  class TestCompression (line 12) | class TestCompression(unittest.TestCase):
    method setUp (line 13) | def setUp(self):
    method tearDown (line 15) | def tearDown(self):
    method test_should_compress_uncompress (line 18) | def test_should_compress_uncompress(self):

FILE: anima_100k/longer_training.py
  function print_tensors (line 69) | def print_tensors(where_str=''):
  function _get_tensors (line 83) | def _get_tensors(gpu_only=True):
  function get_sample_gen_test_examples (line 99) | def get_sample_gen_test_examples():
  class ModelArguments (line 119) | class ModelArguments:
  class DataArguments (line 129) | class DataArguments:
  class TrainingArguments (line 165) | class TrainingArguments(transformers.Seq2SeqTrainingArguments):
  class GenerationArguments (line 256) | class GenerationArguments:
  function find_all_linear_names (line 287) | def find_all_linear_names(args, model):
  class SampleGenerateCallback (line 301) | class SampleGenerateCallback(transformers.TrainerCallback):
    method on_substep_end (line 304) | def on_substep_end(self, args, state, control, **kwargs ):
    method on_evaluate (line 308) | def on_evaluate(self, args, state, control, **kwargs):
  class SavePeftModelCallback (line 341) | class SavePeftModelCallback(transformers.TrainerCallback):
    method save_model (line 342) | def save_model(self, args, state, kwargs):
    method on_save (line 356) | def on_save(self, args, state, control, **kwargs):
    method on_train_end (line 360) | def on_train_end(self, args, state, control, **kwargs):
  function get_accelerate_model (line 368) | def get_accelerate_model(args, checkpoint_dir):
  function print_trainable_parameters (line 454) | def print_trainable_parameters(args, model):
  function smart_tokenizer_and_embedding_resize (line 471) | def smart_tokenizer_and_embedding_resize(
  class DataCollatorForCausalLM (line 494) | class DataCollatorForCausalLM(object):
    method __call__ (line 501) | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
  function extract_unnatural_instructions_data (line 546) | def extract_unnatural_instructions_data(examples, extract_reformulations...
  function extract_alpaca_dataset (line 576) | def extract_alpaca_dataset(example):
  function local_dataset (line 583) | def local_dataset(dataset_name):
  function make_data_module (line 598) | def make_data_module(tokenizer: transformers.PreTrainedTokenizer, args) ...
  function get_last_checkpoint (line 705) | def get_last_checkpoint(checkpoint_dir):
  function train (line 719) | def train():

FILE: anima_100k/modeling_flash_llama.py
  function rmsnorm_func (line 75) | def rmsnorm_func(hidden_states, weight, variance_epsilon):
  class LlamaRMSNorm (line 83) | class LlamaRMSNorm(nn.Module):
    method __init__ (line 84) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 96) | def forward(self, hidden_states):
  class FlashRotaryEmbedding (line 100) | class FlashRotaryEmbedding(torch.nn.Module):
    method __init__ (line 118) | def __init__(self, dim: int, base=10000.0, interleaved=False, scale_ba...
    method _compute_inv_freq (line 155) | def _compute_inv_freq(self, device=None):
    method _update_cos_sin_cache (line 160) | def _update_cos_sin_cache(self, seqlen, device=None, dtype=None):
    method forward (line 202) | def forward(self, q: torch.Tensor, k: torch.Tensor, seqlen_offset: int...
  class LlamaMLP (line 221) | class LlamaMLP(nn.Module):
    method __init__ (line 222) | def __init__(self, config):
    method forward (line 232) | def forward(self, x):
  function repeat_kv (line 255) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class LlamaAttention (line 267) | class LlamaAttention(nn.Module):
    method __init__ (line 270) | def __init__(self, config: LlamaConfig):
    method _shape (line 307) | def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
    method forward (line 310) | def forward(
  class LlamaDecoderLayer (line 418) | class LlamaDecoderLayer(nn.Module):
    method __init__ (line 419) | def __init__(self, config: LlamaConfig):
    method forward (line 427) | def forward(
  class LlamaPreTrainedModel (line 505) | class LlamaPreTrainedModel(PreTrainedModel):
    method _init_weights (line 512) | def _init_weights(self, module):
    method _set_gradient_checkpointing (line 523) | def _set_gradient_checkpointing(self, module, value=False):
  class LlamaModel (line 596) | class LlamaModel(LlamaPreTrainedModel):
    method __init__ (line 604) | def __init__(self, config: LlamaConfig):
    method get_input_embeddings (line 617) | def get_input_embeddings(self):
    method set_input_embeddings (line 620) | def set_input_embeddings(self, value):
    method forward (line 624) | def forward(
  class LlamaForCausalLM (line 740) | class LlamaForCausalLM(LlamaPreTrainedModel):
    method __init__ (line 743) | def __init__(self, config):
    method get_input_embeddings (line 752) | def get_input_embeddings(self):
    method set_input_embeddings (line 755) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 758) | def get_output_embeddings(self):
    method set_output_embeddings (line 761) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 764) | def set_decoder(self, decoder):
    method get_decoder (line 767) | def get_decoder(self):
    method forward (line 772) | def forward(
    method prepare_inputs_for_generation (line 878) | def prepare_inputs_for_generation(
    method _reorder_cache (line 907) | def _reorder_cache(past_key_values, beam_idx):
  class LlamaForSequenceClassification (line 931) | class LlamaForSequenceClassification(LlamaPreTrainedModel):
    method __init__ (line 932) | def __init__(self, config):
    method get_input_embeddings (line 941) | def get_input_embeddings(self):
    method set_input_embeddings (line 944) | def set_input_embeddings(self, value):
    method forward (line 948) | def forward(

FILE: rlhf/qlora_dpo.py
  class ModelArguments (line 68) | class ModelArguments:
  class DataArguments (line 78) | class DataArguments:
  class TrainingArguments (line 114) | class TrainingArguments(transformers.Seq2SeqTrainingArguments):
  class GenerationArguments (line 207) | class GenerationArguments:
  function find_all_linear_names (line 238) | def find_all_linear_names(args, model):
  class SampleGenerateCallback (line 252) | class SampleGenerateCallback(transformers.TrainerCallback):
    method on_evaluate (line 255) | def on_evaluate(self, args, state, control, **kwargs):
  class SavePeftModelCallback (line 284) | class SavePeftModelCallback(transformers.TrainerCallback):
    method save_model (line 285) | def save_model(self, args, state, kwargs):
    method on_save (line 299) | def on_save(self, args, state, control, **kwargs):
    method on_train_end (line 303) | def on_train_end(self, args, state, control, **kwargs):
  function get_reference_model (line 311) | def get_reference_model(args, checkpoint_dir):
  function get_accelerate_model (line 354) | def get_accelerate_model(args, checkpoint_dir):
  function print_trainable_parameters (line 429) | def print_trainable_parameters(args, model):
  function smart_tokenizer_and_embedding_resize (line 446) | def smart_tokenizer_and_embedding_resize(
  class DataCollatorForCausalLM (line 469) | class DataCollatorForCausalLM(object):
    method __call__ (line 476) | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
  function extract_unnatural_instructions_data (line 516) | def extract_unnatural_instructions_data(examples, extract_reformulations...
  function extract_alpaca_dataset (line 546) | def extract_alpaca_dataset(example):
  function local_dataset (line 553) | def local_dataset(dataset_name):
  function make_data_module (line 568) | def make_data_module(tokenizer: transformers.PreTrainedTokenizer, args) ...
  function get_last_checkpoint (line 652) | def get_last_checkpoint(checkpoint_dir):
  function _get_batch_logps (line 666) | def _get_batch_logps(logits: torch.FloatTensor, labels: torch.LongTensor...
  function dpo_loss (line 694) | def dpo_loss(policy_chosen_logps: torch.FloatTensor,
  class DPOSeq2SeqTrainer (line 742) | class DPOSeq2SeqTrainer(Seq2SeqTrainer):
    method __init__ (line 743) | def __init__(self, reference_model: torch.nn.Module,
    method compute_loss (line 753) | def compute_loss(self, model, inputs, return_outputs=False):
  function compute_metrics (line 781) | def compute_metrics(ep: EvalPrediction):
  function train (line 785) | def train():

FILE: training/qlora.py
  class ModelArguments (line 66) | class ModelArguments:
  class DataArguments (line 76) | class DataArguments:
  class TrainingArguments (line 112) | class TrainingArguments(transformers.Seq2SeqTrainingArguments):
  class GenerationArguments (line 202) | class GenerationArguments:
  function find_all_linear_names (line 233) | def find_all_linear_names(args, model):
  class SampleGenerateCallback (line 247) | class SampleGenerateCallback(transformers.TrainerCallback):
    method on_evaluate (line 250) | def on_evaluate(self, args, state, control, **kwargs):
  class SavePeftModelCallback (line 279) | class SavePeftModelCallback(transformers.TrainerCallback):
    method save_model (line 280) | def save_model(self, args, state, kwargs):
    method on_save (line 294) | def on_save(self, args, state, control, **kwargs):
    method on_train_end (line 298) | def on_train_end(self, args, state, control, **kwargs):
  function get_accelerate_model (line 306) | def get_accelerate_model(args, checkpoint_dir):
  function print_trainable_parameters (line 381) | def print_trainable_parameters(args, model):
  function smart_tokenizer_and_embedding_resize (line 398) | def smart_tokenizer_and_embedding_resize(
  class DataCollatorForCausalLM (line 421) | class DataCollatorForCausalLM(object):
    method __call__ (line 428) | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
  function extract_unnatural_instructions_data (line 473) | def extract_unnatural_instructions_data(examples, extract_reformulations...
  function extract_alpaca_dataset (line 503) | def extract_alpaca_dataset(example):
  function local_dataset (line 510) | def local_dataset(dataset_name):
  function make_data_module (line 525) | def make_data_module(tokenizer: transformers.PreTrainedTokenizer, args) ...
  function get_last_checkpoint (line 651) | def get_last_checkpoint(checkpoint_dir):
  function train (line 665) | def train():