osmarks/nanogpt-experiments
1
0
Fork 0
mirror of https://github.com/osmarks/nanogpt-experiments.git synced 2024-12-18 14:10:28 +00:00
Code Issues Releases Wiki Activity

use relative paths so that running the data prep scripts always create files in local folder, no matter where run from

Browse Source
This commit is contained in:
DG 2023-01-20 10:39:45 -08:00
parent 2c7806db6e
commit edb7a7eab0
3 changed files with 7 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
data/openwebtext/prepare.py
View File

@ -1,6 +1,7 @@
# saves the openwebtext dataset to a binary file for training. following was helpful: # saves the openwebtext dataset to a binary file for training. following was helpful:
# https://github.com/HazyResearch/flash-attention/blob/main/training/src/datamodules/language_modeling_hf.py # https://github.com/HazyResearch/flash-attention/blob/main/training/src/datamodules/language_modeling_hf.py
import os
from tqdm import tqdm from tqdm import tqdm
import numpy as np import numpy as np
import tiktoken import tiktoken
@ -50,7 +51,7 @@ tokenized = split_dataset.map(
# concatenate all the ids in each dataset into one large file we can use for training # concatenate all the ids in each dataset into one large file we can use for training
for split, dset in tokenized.items(): for split, dset in tokenized.items():
arr_len = np.sum(dset['len']) arr_len = np.sum(dset['len'])
filename = f'{split}.bin' filename = os.path.join(os.path.dirname(__file__), f'{split}.bin')
dtype = np.uint16 # (can do since enc.max_token_value == 50256 is < 2**16) dtype = np.uint16 # (can do since enc.max_token_value == 50256 is < 2**16)
arr = np.memmap(filename, dtype=dtype, mode='w+', shape=(arr_len,)) arr = np.memmap(filename, dtype=dtype, mode='w+', shape=(arr_len,))

4
data/shakespeare/prepare.py
View File

@ -25,8 +25,8 @@ print(f"val has {len(val_ids):,} tokens")
# export to bin files # export to bin files
train_ids = np.array(train_ids, dtype=np.uint16) train_ids = np.array(train_ids, dtype=np.uint16)
val_ids = np.array(val_ids, dtype=np.uint16) val_ids = np.array(val_ids, dtype=np.uint16)
train_ids.tofile('train.bin') train_ids.tofile(os.path.join(os.path.dirname(__file__), 'train.bin'))
val_ids.tofile('val.bin') val_ids.tofile(os.path.join(os.path.dirname(__file__), 'val.bin'))
# train.bin has 301,966 tokens # train.bin has 301,966 tokens
# val.bin has 36,059 tokens # val.bin has 36,059 tokens

6
data/shakespeare_char/prepare.py
View File

@ -47,8 +47,8 @@ print(f"val has {len(val_ids):,} tokens")
# export to bin files # export to bin files
train_ids = np.array(train_ids, dtype=np.uint16) train_ids = np.array(train_ids, dtype=np.uint16)
val_ids = np.array(val_ids, dtype=np.uint16) val_ids = np.array(val_ids, dtype=np.uint16)
train_ids.tofile('train.bin') train_ids.tofile(os.path.join(os.path.dirname(__file__), 'train.bin'))
val_ids.tofile('val.bin') val_ids.tofile(os.path.join(os.path.dirname(__file__), 'val.bin'))
# save the meta information as well, to help us encode/decode later # save the meta information as well, to help us encode/decode later
meta = { meta = {
@ -56,7 +56,7 @@ meta = {
'itos': itos, 'itos': itos,
'stoi': stoi, 'stoi': stoi,
} }
with open('meta.pkl', 'wb') as f: with open(os.path.join(os.path.dirname(__file__), 'meta.pkl'), 'wb') as f:
pickle.dump(meta, f) pickle.dump(meta, f)
# length of dataset in characters: 1115394 # length of dataset in characters: 1115394