SYMBOL INDEX (69 symbols across 11 files)

FILE: demo_llama.py
  class Llama (line 23) | class Llama:
    method __init__ (line 24) | def __init__(self, onnxdir='models', config: dict = {}):
    method _make_causal_mask (line 47) | def _make_causal_mask(self,
    method _expand_mask (line 73) | def _expand_mask(self, mask, dtype, tgt_len=None):
    method _prepare_decoder_attention_mask (line 93) | def _prepare_decoder_attention_mask(self, attention_mask, input_shape,
    method convert_to_fp16 (line 118) | def convert_to_fp16(self, inputs):
    method decode (line 127) | def decode(self, token: np.array):
    method apply_warp (line 179) | def apply_warp(self, tensor: np.array):
    method sample (line 184) | def sample(self, prompt: str = 'bonjour'):
  function parse_args (line 237) | def parse_args():
  function main (line 274) | def main():

FILE: demo_rwkv.py
  class RWKV_RNN (line 11) | class RWKV_RNN():
    method __init__ (line 13) | def __init__(self, onnxdir: str, n_layer=24):
    method forward (line 20) | def forward(self, token, state):
  function parse_args (line 33) | def parse_args():
  function main (line 43) | def main():

FILE: llama/decoder.py
  class Decoder (line 7) | class Decoder:
    method __init__ (line 9) | def __init__(self, pool: MemoryPoolSimple, onnxdir: str, nameformat: s...
    method decode (line 23) | def decode(self, _inputs: dict, idx: int):
    method embed (line 31) | def embed(self, input_ids: np.array):
    method norm_head (line 37) | def norm_head(self, hidden: np.array):

FILE: llama/logits_process.py
  function warp_topk (line 5) | def warp_topk(tensor: np.array, topk: int, fill_value = -float("Inf")):
  function warp_temperature (line 22) | def warp_temperature(tensor: np.array, temperature: float):
  function sample_logits (line 33) | def sample_logits(probs, temperature=1.0, top_p=0.85):

FILE: llama/memory_pool.py
  class OrtWrapper (line 11) | class OrtWrapper:
    method __init__ (line 12) | def __init__(self, onnxfile: str):
    method forward (line 21) | def forward(self, _inputs: dict):
    method __del__ (line 32) | def __del__(self):
  class MemoryPoolSimple (line 37) | class MemoryPoolSimple:
    method __init__ (line 38) | def __init__(self, maxGB):
    method submit (line 46) | def submit(self, key: str, onnx_filepath: str):
    method used (line 56) | def used(self):
    method check (line 75) | def check(self):
    method fetch (line 87) | def fetch(self, key: str):

FILE: llama/tokenizer.py
  class Tokenizer (line 11) | class Tokenizer:
    method __init__ (line 13) | def __init__(self, model_path: str):
    method encode (line 29) | def encode(self, s: str, bos: bool, eos: bool) -> List[int]:
    method decode (line 38) | def decode(self, t: List[int]) -> str:

FILE: llama/utils.py
  function singleton (line 4) | def singleton(cls):
  function npsoftmax (line 18) | def npsoftmax(x, axis):
  function npmultinominal2D (line 23) | def npmultinominal2D(x):

FILE: tools/convert-fp32-to-fp16.py
  function main (line 11) | def main(_input: str, output: str):
  function parse_args (line 32) | def parse_args():

FILE: tools/convert-to-tvm.py
  function convert (line 9) | def convert(filepath: str, outdir: str):
  function parse_args (line 31) | def parse_args():

FILE: tools/export-onnx.py
  class Predictor (line 22) | class Predictor():
    method __init__ (line 23) | def __init__(self, outdir):
    method predict (line 29) | def predict(

FILE: tools/onnx_RWKV_in_150_lines.py
  function onnx_fp32_to_fp16 (line 34) | def onnx_fp32_to_fp16(onnxfile: str):
  class Encoder (line 44) | class Encoder(torch.nn.Module):
    method __init__ (line 45) | def __init__(self, emb, ln_weight, ln_bias):
    method forward (line 51) | def forward(self, token):
  class Decoder (line 58) | class Decoder(torch.nn.Module):
    method __init__ (line 59) | def __init__(self, head, ln_weight, ln_bias):
    method forward (line 65) | def forward(self, x):
  class Mixer (line 71) | class Mixer(torch.nn.Module):
    method __init__ (line 72) | def __init__(self):
    method set_ffn (line 75) | def set_ffn(self, time_mix_k, time_mix_r, kw, vw, rw, ln_weight, ln_bi...
    method forward_ffn (line 84) | def forward_ffn(self, x, state):
    method set_attn (line 93) | def set_attn(self, time_mix_k, time_mix_v, time_mix_r, time_first, tim...
    method forward_att (line 108) | def forward_att(self, x, state):
    method forward (line 147) | def forward(self, x, state):
  class RWKV_RNN (line 153) | class RWKV_RNN(torch.jit.ScriptModule):
    method __init__ (line 155) | def __init__(self, args):
    method layer_norm (line 188) | def layer_norm(self, x, w):
    method encode_export (line 231) | def encode_export(self, token, emb, ln_weight, ln_bias):
    method decode_export (line 238) | def decode_export(self, x, head, ln_weight, ln_bias):
    method forward (line 243) | def forward(self, tokenid, state):
  function sample_logits (line 301) | def sample_logits(out, temperature=1.0, top_p=0.8):