diff options
Diffstat (limited to 'training')
-rw-r--r-- | training/functional.py | 11 |
1 files changed, 5 insertions, 6 deletions
diff --git a/training/functional.py b/training/functional.py index 56c2995..fd3f9f4 100644 --- a/training/functional.py +++ b/training/functional.py | |||
@@ -68,14 +68,13 @@ class TrainingStrategy(): | |||
68 | prepare: TrainingStrategyPrepareCallable | 68 | prepare: TrainingStrategyPrepareCallable |
69 | 69 | ||
70 | 70 | ||
71 | def get_models(pretrained_model_name_or_path: str): | 71 | def get_models(pretrained_model_name_or_path: str, torch_dtype=torch.float32): |
72 | tokenizer = MultiCLIPTokenizer.from_pretrained(pretrained_model_name_or_path, subfolder='tokenizer') | 72 | tokenizer = MultiCLIPTokenizer.from_pretrained(pretrained_model_name_or_path, subfolder='tokenizer') |
73 | text_encoder = CLIPTextModel.from_pretrained(pretrained_model_name_or_path, subfolder='text_encoder') | 73 | text_encoder = CLIPTextModel.from_pretrained(pretrained_model_name_or_path, subfolder='text_encoder', torch_dtype=torch_dtype) |
74 | vae = AutoencoderKL.from_pretrained(pretrained_model_name_or_path, subfolder='vae') | 74 | vae = AutoencoderKL.from_pretrained(pretrained_model_name_or_path, subfolder='vae', torch_dtype=torch_dtype) |
75 | unet = UNet2DConditionModel.from_pretrained(pretrained_model_name_or_path, subfolder='unet') | 75 | unet = UNet2DConditionModel.from_pretrained(pretrained_model_name_or_path, subfolder='unet', torch_dtype=torch_dtype) |
76 | noise_scheduler = DDPMScheduler.from_pretrained(pretrained_model_name_or_path, subfolder='scheduler') | 76 | noise_scheduler = DDPMScheduler.from_pretrained(pretrained_model_name_or_path, subfolder='scheduler') |
77 | sample_scheduler = UniPCMultistepScheduler.from_pretrained( | 77 | sample_scheduler = UniPCMultistepScheduler.from_pretrained(pretrained_model_name_or_path, subfolder='scheduler') |
78 | pretrained_model_name_or_path, subfolder='scheduler') | ||
79 | 78 | ||
80 | return tokenizer, text_encoder, vae, unet, noise_scheduler, sample_scheduler | 79 | return tokenizer, text_encoder, vae, unet, noise_scheduler, sample_scheduler |
81 | 80 | ||