xtuner
/

llava-internlm-7b-pretrain

Visual Question Answering

Model card Files Files and versions

LZHgrla commited on Dec 16, 2023

Commit

cab2f72

·

1 Parent(s): 2b2877f

Update xtuner_config.py

Files changed (1) hide show

xtuner_config.py +4 -4

xtuner_config.py CHANGED Viewed

@@ -19,7 +19,7 @@ custom_hooks = [
             '',
         ],
         every_n_iters=500,
-        processor=dict(
             pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
             trust_remote_code=True,
             type='transformers.CLIPImageProcessor.from_pretrained'),
@@ -58,7 +58,7 @@ llava_dataset = dict(
     image_folder='./data/llava_data/LLaVA-Pretrain/images',
     max_length=1472,
     pad_image_to_square=False,
-    processor=dict(
         pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
         trust_remote_code=True,
         type='transformers.CLIPImageProcessor.from_pretrained'),
@@ -126,7 +126,7 @@ param_scheduler = [
         eta_min=0.0,
         type='mmengine.optim.CosineAnnealingLR'),
 ]
-processor = dict(
     pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
     trust_remote_code=True,
     type='transformers.CLIPImageProcessor.from_pretrained')
@@ -165,7 +165,7 @@ train_dataloader = dict(
         image_folder='./data/llava_data/LLaVA-Pretrain/images',
         max_length=1472,
         pad_image_to_square=False,
-        processor=dict(
             pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
             trust_remote_code=True,
             type='transformers.CLIPImageProcessor.from_pretrained'),

             '',
         ],
         every_n_iters=500,
+        image_processor=dict(
             pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
             trust_remote_code=True,
             type='transformers.CLIPImageProcessor.from_pretrained'),
     image_folder='./data/llava_data/LLaVA-Pretrain/images',
     max_length=1472,
     pad_image_to_square=False,
+    image_processor=dict(
         pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
         trust_remote_code=True,
         type='transformers.CLIPImageProcessor.from_pretrained'),
         eta_min=0.0,
         type='mmengine.optim.CosineAnnealingLR'),
 ]
+image_processor = dict(
     pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
     trust_remote_code=True,
     type='transformers.CLIPImageProcessor.from_pretrained')
         image_folder='./data/llava_data/LLaVA-Pretrain/images',
         max_length=1472,
         pad_image_to_square=False,
+        image_processor=dict(
             pretrained_model_name_or_path='openai/clip-vit-large-patch14-336',
             trust_remote_code=True,
             type='transformers.CLIPImageProcessor.from_pretrained'),