a-One-Fan/DL-Art-School
1
0
Fork 0
forked from mrq/DL-Art-School
Code Pull Requests Activity

Partially implement performers in transformer_builders

Browse Source
This commit is contained in:
James Betker 2022-01-09 22:35:03 -07:00
parent ec456b6733
commit f503d8d96b
1 changed files with 19 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
codes/models/gpt_voice/transformer_builders.py
View File

@ -12,7 +12,9 @@ Every function contains the following arguments:
checkpointing: Whether or not the underlying implementation should support gradient checkpointing. checkpointing: Whether or not the underlying implementation should support gradient checkpointing.
""" """
import functools import functools
from time import time
import torch import torch
from tqdm import tqdm
def null_position_embeddings(range, dim): def null_position_embeddings(range, dim):
@ -36,6 +38,8 @@ def build_hf_gpt_transformer(layers, model_dim, heads, num_tokens, max_seq_len,
# Override the built in positional embeddings # Override the built in positional embeddings
del gpt.wpe del gpt.wpe
gpt.wpe = functools.partial(null_position_embeddings, dim=model_dim) gpt.wpe = functools.partial(null_position_embeddings, dim=model_dim)
# Built-in token embeddings are unused.
del gpt.wte
return gpt return gpt
@ -43,7 +47,10 @@ def build_lr_performer(layers, model_dim, heads, num_tokens, max_seq_len, checkp
""" """
lucidrains Performer implementation, https://github.com/lucidrains/performer-pytorch lucidrains Performer implementation, https://github.com/lucidrains/performer-pytorch
""" """
pass from models.lucidrains.performer.performer_pytorch import PerformerLM
model = PerformerLM(dim=model_dim, depth=layers, heads=heads, dim_head=model_dim, causal=True,
num_tokens=num_tokens, max_seq_len=max_seq_len)
return model
def build_lr_reformer(layers, model_dim, heads, num_tokens, max_seq_len, checkpointing): def build_lr_reformer(layers, model_dim, heads, num_tokens, max_seq_len, checkpointing):
@ -61,10 +68,19 @@ def build_lr_xformer(layers, model_dim, heads, num_tokens, max_seq_len, checkpoi
def test_all_performance(**kwargs): def test_all_performance(**kwargs):
transformer_builders = [build_hf_gpt_transformer, build_lr_performer, build_lr_reformer, build_lr_xformer] transformer_builders = [#build_hf_gpt_transformer,
build_lr_performer,]
# build_lr_reformer,
# build_lr_xformer]
for builder in transformer_builders: for builder in transformer_builders:
model = builder(**kwargs) model = builder(**kwargs)
start = time()
args = torch.randint(0, 8192, (16,450))
for k in tqdm(range(10)):
model(args)
stop = time()
print(f"Model: {str(builder)}; Elapsed: {stop-start}")
if __name__ == '__main__': if __name__ == '__main__':
test_all_performance(12, 512, 8, 8192, 1000, False) test_all_performance(layers=12, model_dim=512, heads=8, num_tokens=8192, max_seq_len=1000, checkpointing=False)