deepseekmirror/Janus
1
0
Fork 0
mirror of https://github.com/deepseek-ai/Janus.git synced 2025-04-19 10:09:00 -04:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge eb441468d8 into 1daa72fa40

Browse Source
This commit is contained in:
江程训 2025-02-01 15:59:58 +08:00 committed by GitHub
commit 3644b7ecd4
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
5 changed files with 28 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
.gitignore vendored
View File

@ -419,3 +419,6 @@ tags
.vscode .vscode
.github .github
generated_samples/ generated_samples/
# gradio
.gradio/

2
demo/app.py
View File

@ -19,7 +19,7 @@ vl_gpt = vl_gpt.to(torch.bfloat16).cuda()
vl_chat_processor = VLChatProcessor.from_pretrained(model_path) vl_chat_processor = VLChatProcessor.from_pretrained(model_path)
tokenizer = vl_chat_processor.tokenizer tokenizer = vl_chat_processor.tokenizer
cuda_device = 'cuda' if torch.cuda.is_available() else 'cpu' cuda_device = 'cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu'
# Multimodal Understanding function # Multimodal Understanding function
@torch.inference_mode() @torch.inference_mode()
# Multimodal Understanding function # Multimodal Understanding function

8
demo/app_janusflow.py
View File

@ -5,7 +5,13 @@ from PIL import Image
from diffusers.models import AutoencoderKL from diffusers.models import AutoencoderKL
import numpy as np import numpy as np
cuda_device = 'cuda' if torch.cuda.is_available() else 'cpu' cuda_device = 'cpu'
if torch.cuda.is_available():
cuda_device = 'cuda'
elif torch.backends.mps.is_available():
cuda_device = 'mps'
else:
cuda_device = 'cpu'
# Load model and processor # Load model and processor
model_path = "deepseek-ai/JanusFlow-1.3B" model_path = "deepseek-ai/JanusFlow-1.3B"

11
demo/app_januspro.py
View File

@ -21,23 +21,29 @@ vl_gpt = AutoModelForCausalLM.from_pretrained(model_path,
trust_remote_code=True) trust_remote_code=True)
if torch.cuda.is_available(): if torch.cuda.is_available():
vl_gpt = vl_gpt.to(torch.bfloat16).cuda() vl_gpt = vl_gpt.to(torch.bfloat16).cuda()
cuda_device = 'cuda'
elif torch.backends.mps.is_available():
vl_gpt = vl_gpt.to(torch.float16).to('mps')
cuda_device = 'mps'
else: else:
vl_gpt = vl_gpt.to(torch.float16) vl_gpt = vl_gpt.to(torch.float16)
cuda_device = 'cpu'
vl_chat_processor = VLChatProcessor.from_pretrained(model_path) vl_chat_processor = VLChatProcessor.from_pretrained(model_path)
tokenizer = vl_chat_processor.tokenizer tokenizer = vl_chat_processor.tokenizer
cuda_device = 'cuda' if torch.cuda.is_available() else 'cpu'
@torch.inference_mode() @torch.inference_mode()
# @spaces.GPU(duration=120) # @spaces.GPU(duration=120)
# Multimodal Understanding function # Multimodal Understanding function
def multimodal_understanding(image, question, seed, top_p, temperature): def multimodal_understanding(image, question, seed, top_p, temperature):
# Clear CUDA cache before generating # Clear CUDA cache before generating
if torch.cuda.is_available():
torch.cuda.empty_cache() torch.cuda.empty_cache()
# set seed # set seed
torch.manual_seed(seed) torch.manual_seed(seed)
np.random.seed(seed) np.random.seed(seed)
if torch.cuda.is_available():
torch.cuda.manual_seed(seed) torch.cuda.manual_seed(seed)
conversation = [ conversation = [
@ -83,6 +89,7 @@ def generate(input_ids,
image_token_num_per_image: int = 576, image_token_num_per_image: int = 576,
patch_size: int = 16): patch_size: int = 16):
# Clear CUDA cache before generating # Clear CUDA cache before generating
if torch.cuda.is_available():
torch.cuda.empty_cache() torch.cuda.empty_cache()
tokens = torch.zeros((parallel_size * 2, len(input_ids)), dtype=torch.int).to(cuda_device) tokens = torch.zeros((parallel_size * 2, len(input_ids)), dtype=torch.int).to(cuda_device)
@ -138,10 +145,12 @@ def generate_image(prompt,
guidance=5, guidance=5,
t2i_temperature=1.0): t2i_temperature=1.0):
# Clear CUDA cache and avoid tracking gradients # Clear CUDA cache and avoid tracking gradients
if torch.cuda.is_available():
torch.cuda.empty_cache() torch.cuda.empty_cache()
# Set the seed for reproducible results # Set the seed for reproducible results
if seed is not None: if seed is not None:
torch.manual_seed(seed) torch.manual_seed(seed)
if torch.cuda.is_available():
torch.cuda.manual_seed(seed) torch.cuda.manual_seed(seed)
np.random.seed(seed) np.random.seed(seed)
width = 384 width = 384

2
demo/fastapi_app.py
View File

@ -21,7 +21,7 @@ vl_gpt = vl_gpt.to(torch.bfloat16).cuda()
vl_chat_processor = VLChatProcessor.from_pretrained(model_path) vl_chat_processor = VLChatProcessor.from_pretrained(model_path)
tokenizer = vl_chat_processor.tokenizer tokenizer = vl_chat_processor.tokenizer
cuda_device = 'cuda' if torch.cuda.is_available() else 'cpu' cuda_device = 'cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu'
@torch.inference_mode() @torch.inference_mode()