How to custom checkpoint correctly？

[forestbat]

I have read this PR， and I wrote this code：

# 'VIT_H SAM Model/sam_vit_h_4b8939.pth' is my directory of model file
sam = LangSAM('VIT_H SAM Model/sam_vit_h_4b8939.pth')

And it crashed with this error:

self = <samgeo.text_sam.LangSAM object at 0x00000210B416E860>
model_type = 'VIT_H SAM Model/sam_vit_h_4b8939.pth'

    def build_sam(self, model_type):
        """Build the SAM model.

        Args:
            model_type (str, optional): The model type. It can be one of the following: vit_h, vit_l, vit_b.
                Defaults to 'vit_h'. See https://bit.ly/3VrpxUh for more details.
        """
>       checkpoint_url = SAM_MODELS[model_type]
E       KeyError: 'VIT_H SAM Model/sam_vit_h_4b8939.pth'

And even if I use sam = LangSAM(ckpt_path='VIT_H SAM Model/sam_vit_h_4b8939.pth') or sam = LangSAM('vit_h', 'VIT_H SAM Model/sam_vit_h_4b8939.pth') , it isn't effective. So how to custom checkpoint correctly?

[luca-medeiros]

Based on the 'error' message you sent, seems you have modified the code. Please try using main branch's code and if error raises, upload the whole error traceback.

[forestbat]

I have re-installed lang-sam and its version in pip list is 0.1.0. However it still crashed and traceback is same as the last traceback:

Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertModel: ['cls.predictions.transform.dense.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias']
- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).

main.py:5 (test_prompt_samgeo)
def test_prompt_samgeo():
        image = "pics/hainanm3-sentinel-10.jpg"
>       sam = LangSAM('VIT_H SAM Model/sam_vit_h_4b8939.pth')

main.py:22: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
C:\ProgramData\anaconda3\envs\graduate\lib\site-packages\samgeo\text_sam.py:122: in __init__
    self.build_sam(model_type)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <samgeo.text_sam.LangSAM object at 0x00000280D399F370>
model_type = 'VIT_H SAM Model/sam_vit_h_4b8939.pth'

    def build_sam(self, model_type):
        """Build the SAM model.

        Args:
            model_type (str, optional): The model type. It can be one of the following: vit_h, vit_l, vit_b.
                Defaults to 'vit_h'. See https://bit.ly/3VrpxUh for more details.
        """
>       checkpoint_url = SAM_MODELS[model_type]
E       KeyError: 'VIT_H SAM Model/sam_vit_h_4b8939.pth'

C:\ProgramData\anaconda3\envs\graduate\lib\site-packages\samgeo\text_sam.py:139: KeyError

[luca-medeiros]

this traceback seems to be from a file called text_sam.py which doesn't exist in the library. The build_sam method too seems to be modified. I cant help you if you are not using lang_sam from main branch.