improvement(tools): optimize with argparse

This commit is contained in:
tpoisonooo 2023-03-17 16:53:53 +08:00
parent 904d2a8d6a
commit 3c7cb413fb

View file

@ -22,19 +22,27 @@ import json
import struct
import numpy as np
import torch
import argparse
import os
from sentencepiece import SentencePieceProcessor
if len(sys.argv) < 3:
print("Usage: convert-ckpt-to-ggml.py dir-model ftype\n")
print(" ftype == 0 -> float32")
print(" ftype == 1 -> float16")
sys.exit(1)
# output in the same directory as the model
dir_model = sys.argv[1]
def parse_args():
parser = argparse.ArgumentParser(
description='Convert ckpt models to ggml models.')
parser.add_argument('dir_model',
type=str,
help='Directory path of the checkpoint model')
parser.add_argument('ftype',
type=str,
choices=['f32', 'f16'],
help='Data type of the converted tensor, f32 or f16')
parser.add_argument('out_dir',
type=str,
help='Directory path for storing ggml model')
return parser.parse_args()
fname_hparams = sys.argv[1] + "/params.json"
fname_tokenizer = sys.argv[1] + "/../tokenizer.model"
def get_n_parts(dim):
if dim == 4096:
@ -49,129 +57,138 @@ def get_n_parts(dim):
print("Invalid dim: " + str(dim))
sys.exit(1)
# possible data types
# ftype == 0 -> float32
# ftype == 1 -> float16
#
# map from ftype to string
ftype_str = ["f32", "f16"]
ftype = 1
if len(sys.argv) > 2:
ftype = int(sys.argv[2])
if ftype < 0 or ftype > 1:
print("Invalid ftype: " + str(ftype))
sys.exit(1)
fname_out = sys.argv[1] + "/ggml-model-" + ftype_str[ftype] + ".bin"
def main():
args = parse_args()
dir_model = args.dir_model
out_dir = args.out_dir
with open(fname_hparams, "r") as f:
hparams = json.load(f)
if not os.path.exists(out_dir):
os.mkdir(out_dir)
tokenizer = SentencePieceProcessor(fname_tokenizer)
ftype = args.ftype
ftype_int = {'f32': 0, 'f16': 1}
fname_hparams = os.path.join(dir_model, 'params.json')
fname_tokenizer = os.path.join(dir_model, '..', 'tokenizer.model')
hparams.update({"vocab_size": tokenizer.vocab_size()})
with open(fname_hparams, "r") as f:
hparams = json.load(f)
n_parts = get_n_parts(hparams["dim"])
tokenizer = SentencePieceProcessor(fname_tokenizer)
print(hparams)
print('n_parts = ', n_parts)
hparams.update({"vocab_size": tokenizer.vocab_size()})
for p in range(n_parts):
print('Processing part ', p)
n_parts = get_n_parts(hparams["dim"])
#fname_model = sys.argv[1] + "/consolidated.00.pth"
fname_model = sys.argv[1] + "/consolidated.0" + str(p) + ".pth"
fname_out = sys.argv[1] + "/ggml-model-" + ftype_str[ftype] + ".bin"
if (p > 0):
fname_out = sys.argv[1] + "/ggml-model-" + ftype_str[ftype] + ".bin" + "." + str(p)
print(hparams)
print('n_parts = ', n_parts)
model = torch.load(fname_model, map_location="cpu")
for p in range(n_parts):
print('Processing part ', p)
fout = open(fname_out, "wb")
fout.write(struct.pack("i", 0x67676d6c)) # magic: ggml in hex
fout.write(struct.pack("i", hparams["vocab_size"]))
fout.write(struct.pack("i", hparams["dim"]))
fout.write(struct.pack("i", hparams["multiple_of"]))
fout.write(struct.pack("i", hparams["n_heads"]))
fout.write(struct.pack("i", hparams["n_layers"]))
fout.write(struct.pack("i", hparams["dim"] // hparams["n_heads"])) # rot (obsolete)
fout.write(struct.pack("i", ftype))
# Is this correct??
for i in range(tokenizer.vocab_size()):
if tokenizer.is_unknown(i):
# "<unk>" token (translated as ??)
text = " \u2047 ".encode("utf-8")
fout.write(struct.pack("i", len(text)))
fout.write(text)
elif tokenizer.is_control(i):
# "<s>"/"</s>" tokens
fout.write(struct.pack("i", 0))
elif tokenizer.is_byte(i):
# "<U+XX>" tokens (which may be invalid UTF-8)
piece = tokenizer.id_to_piece(i)
if len(piece) != 6:
print("Invalid token: " + piece)
sys.exit(1)
byte_value = int(piece[3:-1], 16)
fout.write(struct.pack("i", 1))
fout.write(struct.pack("B", byte_value))
#fname_model = sys.argv[1] + "/consolidated.00.pth"
fname_model = os.path.join(dir_model, "consolidated.0{}.pth".format(p))
if p > 0:
fname_out = os.path.join(out_dir,
"ggml-model-{}.bin.{}".format(ftype, p))
else:
# normal token. Uses U+2581 (LOWER ONE EIGHTH BLOCK) to represent spaces.
text = tokenizer.id_to_piece(i).replace("\u2581", " ").encode("utf-8")
fout.write(struct.pack("i", len(text)))
fout.write(text)
fname_out = os.path.join(out_dir,
"ggml-model-{}.bin".format(ftype))
for k, v in model.items():
name = k
shape = v.shape
model = torch.load(fname_model, map_location="cpu")
# skip layers.X.attention.inner_attention.rope.freqs
if name[-5:] == "freqs":
continue
fout = open(fname_out, "wb")
print("Processing variable: " + name + " with shape: ", shape, " and type: ", v.dtype)
fout.write(struct.pack("i", 0x67676d6c)) # magic: ggml in hex
fout.write(struct.pack("i", hparams["vocab_size"]))
fout.write(struct.pack("i", hparams["dim"]))
fout.write(struct.pack("i", hparams["multiple_of"]))
fout.write(struct.pack("i", hparams["n_heads"]))
fout.write(struct.pack("i", hparams["n_layers"]))
fout.write(struct.pack("i", hparams["dim"] //
hparams["n_heads"])) # rot (obsolete)
fout.write(struct.pack("i", ftype_int[ftype]))
#data = tf.train.load_variable(dir_model, name).squeeze()
data = v.numpy().squeeze()
n_dims = len(data.shape);
# Is this correct??
for i in range(tokenizer.vocab_size()):
if tokenizer.is_unknown(i):
# "<unk>" token (translated as ??)
text = " \u2047 ".encode("utf-8")
fout.write(struct.pack("i", len(text)))
fout.write(text)
elif tokenizer.is_control(i):
# "<s>"/"</s>" tokens
fout.write(struct.pack("i", 0))
elif tokenizer.is_byte(i):
# "<U+XX>" tokens (which may be invalid UTF-8)
piece = tokenizer.id_to_piece(i)
if len(piece) != 6:
print("Invalid token: " + piece)
sys.exit(1)
byte_value = int(piece[3:-1], 16)
fout.write(struct.pack("i", 1))
fout.write(struct.pack("B", byte_value))
else:
# normal token. Uses U+2581 (LOWER ONE EIGHTH BLOCK) to represent spaces.
text = tokenizer.id_to_piece(i).replace("\u2581",
" ").encode("utf-8")
fout.write(struct.pack("i", len(text)))
fout.write(text)
# for efficiency - transpose some matrices
# "model/h.*/attn/c_attn/w"
# "model/h.*/attn/c_proj/w"
# "model/h.*/mlp/c_fc/w"
# "model/h.*/mlp/c_proj/w"
#if name[-14:] == "/attn/c_attn/w" or \
# name[-14:] == "/attn/c_proj/w" or \
# name[-11:] == "/mlp/c_fc/w" or \
# name[-13:] == "/mlp/c_proj/w":
# print(" Transposing")
# data = data.transpose()
for k, v in model.items():
name = k
shape = v.shape
dshape = data.shape
# skip layers.X.attention.inner_attention.rope.freqs
if name[-5:] == "freqs":
continue
# default type is fp16
ftype_cur = 1
if ftype == 0 or n_dims == 1:
print(" Converting to float32")
data = data.astype(np.float32)
ftype_cur = 0
print("Processing variable: " + name + " with shape: ", shape,
" and type: ", v.dtype)
# header
sname = name.encode('utf-8')
fout.write(struct.pack("iii", n_dims, len(sname), ftype_cur))
for i in range(n_dims):
fout.write(struct.pack("i", dshape[n_dims - 1 - i]))
fout.write(sname);
#data = tf.train.load_variable(dir_model, name).squeeze()
data = v.numpy().squeeze()
n_dims = len(data.shape)
# data
data.tofile(fout)
# for efficiency - transpose some matrices
# "model/h.*/attn/c_attn/w"
# "model/h.*/attn/c_proj/w"
# "model/h.*/mlp/c_fc/w"
# "model/h.*/mlp/c_proj/w"
#if name[-14:] == "/attn/c_attn/w" or \
# name[-14:] == "/attn/c_proj/w" or \
# name[-11:] == "/mlp/c_fc/w" or \
# name[-13:] == "/mlp/c_proj/w":
# print(" Transposing")
# data = data.transpose()
# I hope this deallocates the memory ..
model = None
dshape = data.shape
fout.close()
# default type is fp16
ftype_cur = 1
if ftype == 'f32' or n_dims == 1:
print(" Converting to float32")
data = data.astype(np.float32)
ftype_cur = 0
print("Done. Output file: " + fname_out + ", (part ", p, ")")
print("")
# header
sname = name.encode('utf-8')
fout.write(struct.pack("iii", n_dims, len(sname), ftype_cur))
for i in range(n_dims):
fout.write(struct.pack("i", dshape[n_dims - 1 - i]))
fout.write(sname)
# data
data.tofile(fout)
# I hope this deallocates the memory ..
model = None
fout.close()
print("Done. Output file: " + fname_out + ", (part ", p, ")")
print("")
if __name__ == '__main__':
main()