SYMBOL INDEX (78 symbols across 10 files)

FILE: bench.py
  function get_batch (line 37) | def get_batch(split):

FILE: data/openai_summarize_tldr/prepare.py
  function process (line 80) | def process(example):

FILE: data/openwebtext/prepare.py
  function process (line 36) | def process(example):

FILE: data/shakespeare_char/prepare.py
  function encode (line 32) | def encode(s):
  function decode (line 34) | def decode(l):

FILE: model.py
  function new_gelu (line 19) | def new_gelu(x):
  class LayerNorm (line 26) | class LayerNorm(nn.Module):
    method __init__ (line 29) | def __init__(self, ndim, bias):
    method forward (line 34) | def forward(self, input):
  class CausalSelfAttention (line 37) | class CausalSelfAttention(nn.Module):
    method __init__ (line 39) | def __init__(self, config):
    method forward (line 60) | def forward(self, x):
  class MLP (line 86) | class MLP(nn.Module):
    method __init__ (line 88) | def __init__(self, config):
    method forward (line 94) | def forward(self, x):
  class Block (line 101) | class Block(nn.Module):
    method __init__ (line 103) | def __init__(self, config):
    method forward (line 110) | def forward(self, x):
  class GPTConfig (line 116) | class GPTConfig:
  class GPT (line 125) | class GPT(nn.Module):
    method __init__ (line 127) | def __init__(self, config):
    method get_num_params (line 157) | def get_num_params(self, non_embedding=True):
    method _init_weights (line 169) | def _init_weights(self, module):
    method forward (line 177) | def forward(self, idx, targets=None):
    method crop_block_size (line 202) | def crop_block_size(self, block_size):
    method from_pretrained (line 213) | def from_pretrained(cls, model_type, override_args=None):
    method configure_optimizers (line 269) | def configure_optimizers(self, weight_decay, learning_rate, betas, dev...
    method estimate_mfu (line 327) | def estimate_mfu(self, fwdbwd_per_iter, dt):
    method generate (line 344) | def generate(self, idx, max_new_tokens, temperature=1.0, top_k=None):
  class RLHF (line 370) | class RLHF(nn.Module):
    method __init__ (line 371) | def __init__(self, model, mode, discrete_reward=False):
    method forward_reward (line 387) | def forward_reward(self, idx, targets=None):
    method forward (line 414) | def forward(self, idx, targets=None):
    method generate (line 420) | def generate(self, idx, max_new_tokens, device, block_size, use_refere...
    method generate_gumbel (line 501) | def generate_gumbel(self, idx, max_new_tokens, device, block_size, rew...
    method sample_gumbel (line 534) | def sample_gumbel(self, shape, eps=1e-20):
    method gumbel_softmax_sample (line 540) | def gumbel_softmax_sample(self, logits, tau, device, dim=1):
    method gumbel_softmax (line 545) | def gumbel_softmax(self, logits, tau, device):
    method forward_reward_gumbel (line 558) | def forward_reward_gumbel(self, onehots, idx=None, targets=None):

FILE: train_reward_model.py
  function create_comparison_dataset (line 23) | def create_comparison_dataset(path="CarperAI/openai_summarize_comparison...
  class PairwiseDataset (line 41) | class PairwiseDataset(Dataset):
    method __init__ (line 42) | def __init__(self, pairs, max_length):
    method __len__ (line 56) | def __len__(self):
    method __getitem__ (line 59) | def __getitem__(self, idx):
  class DataCollatorReward (line 65) | class DataCollatorReward:
    method __call__ (line 66) | def __call__(self, data):
  function collate_fn (line 72) | def collate_fn(data):

FILE: trainers/reward_trainer.py
  class RewardModelTrainer (line 11) | class RewardModelTrainer(Trainer):
    method __init__ (line 12) | def __init__(self, config, train_data, val_data, collate_fn):
    method get_batch (line 24) | def get_batch(self, split):
    method estimate_loss (line 37) | def estimate_loss(self, model, ctx):
    method evaluate (line 53) | def evaluate(self, model, ctx):
    method train (line 80) | def train(self):
  class ProbRewardModelTrainer (line 171) | class ProbRewardModelTrainer(Trainer):
    method __init__ (line 172) | def __init__(self, config, discrete_reward=False):
    method get_batch (line 179) | def get_batch(self, split):
    method reward (line 194) | def reward(self, sequence, t='and'):
    method evaluate (line 201) | def evaluate(self, model, ctx, X, lr):
    method train (line 256) | def train(self):

FILE: trainers/rl_trainer.py
  class PolicyGradientTrainer (line 10) | class PolicyGradientTrainer(Trainer):
    method __init__ (line 11) | def __init__(self, config):
    method train (line 17) | def train(self):
  class GumbelTrainer (line 105) | class GumbelTrainer(Trainer):
    method __init__ (line 106) | def __init__(self, config):
    method train (line 112) | def train(self):

FILE: trainers/trainer.py
  class Trainer (line 20) | class Trainer():
    method __init__ (line 21) | def __init__(self, config):
    method from_config (line 31) | def from_config(self, config):
    method setup_ddp (line 83) | def setup_ddp(self):
    method get_batch (line 102) | def get_batch(self, split):
    method get_lr (line 114) | def get_lr(self, it):
    method init_model (line 128) | def init_model(self):
    method setup (line 176) | def setup(self):
    method setup_model (line 205) | def setup_model(self, model):
    method evaluate (line 218) | def evaluate(self, model, ctx, lr):
    method train (line 244) | def train(self):
    method estimate_loss (line 336) | def estimate_loss(self, model, ctx):

FILE: utils.py
  class dotdict (line 3) | class dotdict(dict):