|
|
@@ -173,7 +173,19 @@ def load_soft_prompt(name):
|
|
|
else:
|
|
|
with zipfile.ZipFile(Path(f'softprompts/{name}.zip')) as zf:
|
|
|
zf.extract('tensor.npy')
|
|
|
+ zf.extract('meta.json')
|
|
|
+ j = json.loads(open('meta.json', 'r').read())
|
|
|
+ print(f"\nLoading the softprompt \"{name}\".")
|
|
|
+ for field in j:
|
|
|
+ if field != 'name':
|
|
|
+ if type(j[field]) is list:
|
|
|
+ print(f"{field}: {', '.join(j[field])}")
|
|
|
+ else:
|
|
|
+ print(f"{field}: {j[field]}")
|
|
|
+ print()
|
|
|
tensor = np.load('tensor.npy')
|
|
|
+ Path('tensor.npy').unlink()
|
|
|
+ Path('meta.json').unlink()
|
|
|
tensor = torch.Tensor(tensor).to(device=model.device, dtype=model.dtype)
|
|
|
tensor = torch.reshape(tensor, (1, tensor.shape[0], tensor.shape[1]))
|
|
|
|
|
|
@@ -187,6 +199,7 @@ def upload_soft_prompt(file):
|
|
|
zf.extract('meta.json')
|
|
|
j = json.loads(open('meta.json', 'r').read())
|
|
|
name = j['name']
|
|
|
+ Path('meta.json').unlink()
|
|
|
|
|
|
with open(Path(f'softprompts/{name}.zip'), 'wb') as f:
|
|
|
f.write(file)
|