learning/annotated_deep_learning_paper_implementations
1
0
Fork 0
mirror of https://github.com/labmlai/annotated_deep_learning_paper_implementations.git synced 2025-08-06 15:22:21 +08:00
Code Issues Packages Projects Releases Wiki Activity

check loaded weights

Browse Source
This commit is contained in:
lakshith
2024-08-21 09:54:53 +05:30
parent 9e1b35716d
commit 24bd64af7c
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
labml_nn/lora/experiment.py
View File

@ -110,7 +110,11 @@ class Trainer(BaseConfigs):
new_state_dict[layer] = torch.transpose(new_state_dict[layer], 0, 1) new_state_dict[layer] = torch.transpose(new_state_dict[layer], 0, 1)
# Load out model. We use `strict = False` because the state does not have LoRA weights # Load out model. We use `strict = False` because the state does not have LoRA weights
self.model.load_state_dict(new_state_dict, strict=False) missing_keys, unexpected_keys = self.model.load_state_dict(new_state_dict, strict=False)
# make sure that only lora weights are not loaded
assert all('lora' in key for key in missing_keys)
assert not unexpected_keys
def initialize(self): def initialize(self):
""" """