Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

feat: add mps support #22

Open
wants to merge 5 commits into
base: main
Choose a base branch
from
Open

feat: add mps support #22

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
c99e9f1
automatically detect the Python version, if above 3.10 patch collecti…
Fodark Mar 13, 2024
5db8156
use transformers 4.38.1 instead of 4.38.2 as autocast fails on mps on…
Fodark Mar 13, 2024
494e622
add an util function to detect platflorm and suitable dtype
Fodark Mar 13, 2024
40ec649
load model on detected device and use correct dtype
Fodark Mar 13, 2024
4eac5b5
modify example in README to use the new device and dtype function
Fodark Mar 13, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
add an util function to detect platflorm and suitable dtype
  • Loading branch information
@Fodark
Fodark committed Mar 13, 2024
commit 494e622544b7aefd5634fdc72878a0289d4e548f
21 changes: 21 additions & 0 deletions deepseek_vl/utils/io.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,27 @@
from deepseek_vl.models import MultiModalityCausalLM, VLChatProcessor


def get_device_and_dtype():
"""
Get the device and dtype for the model.
"""

if torch.cuda.is_available():
print("Using CUDA and BFloat16")
device = torch.device("cuda")
dtype = torch.bfloat16
elif torch.backends.mps.is_available():
print("Using MPS and FP16")
device = torch.device("mps")
dtype = torch.float16
else:
print("Using CPU and FP32")
device = torch.device("cpu")
dtype = torch.float32

return device, dtype


def load_pretrained_model(model_path: str):
vl_chat_processor: VLChatProcessor = VLChatProcessor.from_pretrained(model_path)
tokenizer = vl_chat_processor.tokenizer
Expand Down