configs

minigpt4/configs/datasets/firstface/featureface.yaml

@@ -0,0 +1,11 @@
+datasets:
+  feature_face_caption:
+    data_type: images
+    
+    build_info:
+      # image_path: /home/user/selected_face/first_face/images
+      image_path: /home/user/selected_face/first_face/first_frames
+
+      ann_path: /home/user/selected_face/face_emotion/relative_train_NCEV.txt
+      # ann_path: /home/user/selected_face/face_emotion/all_label_NCEV.txt
+      # ann_path: /home/user/selected_face/face_emotion/target_label_list_0512_smp.txt

minigpt4/configs/datasets/flickr/caption_to_phrase.yaml

@@ -0,0 +1,6 @@
+datasets:
+  flickr_CaptionToPhrase:
+    data_type: images
+    build_info:
+      image_path: /path/to/filtered_flikcr/images
+      ann_path: /path/to/filtered_flickr/captiontobbox.json

minigpt4/configs/datasets/flickr/default.yaml

@@ -0,0 +1,6 @@
+datasets:
+  flickr_grounded_caption:
+    data_type: images
+    build_info:
+      image_path: /path/to/filtered_flikcr/images
+      ann_path: /path/to/filtered_flikcr/groundedcaption.json

minigpt4/configs/datasets/flickr/object_to_phrase.yaml

@@ -0,0 +1,6 @@
+datasets:
+  flickr_ObjectToPhrase:
+    data_type: images
+    build_info:
+      image_path: /path/to/filtered_flikcr/images
+      ann_path: /path/to/filtered_flikcr/phrasetobbox.json

minigpt4/configs/datasets/gqa/balanced_val.yaml

@@ -0,0 +1,21 @@
+ # Copyright (c) 2022, salesforce.com, inc.
+ # All rights reserved.
+ # SPDX-License-Identifier: BSD-3-Clause
+ # For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+
+datasets:
+  gqa:
+    # data_dir: ${env.data_dir}/datasets
+    data_type: images # [images|videos|features]
+
+    build_info:
+      # Be careful not to append minus sign (-) before split to avoid itemizing
+      annotations:
+        train:
+          url:
+              - https://storage.googleapis.com/sfr-vision-language-research/LAVIS/datasets/gqa/train_balanced_questions.json
+          storage:
+              - /path/to/gqa/train_balanced_questions.json
+
+      images:
+          storage: /path/to/gqa/images

minigpt4/configs/datasets/laion/defaults.yaml

@@ -0,0 +1,5 @@
+datasets:
+  laion:
+    data_type: images
+    build_info:
+      storage: /path/to/laion_dataset/{00000..10488}.tar

minigpt4/configs/datasets/llava/conversation.yaml

@@ -0,0 +1,7 @@
+datasets:
+
+  llava_conversation:
+    data_type: images
+    build_info:
+      image_path: /path/to/coco/images
+      ann_path: /path/to/llava/conversation_58k.json

minigpt4/configs/datasets/llava/detail.yaml

@@ -0,0 +1,6 @@
+datasets:
+  llava_detail:
+    data_type: images
+    build_info:
+      image_path: /path/to/coco/images
+      ann_path: /path/to/llava/detail_23k.json

minigpt4/configs/datasets/llava/reason.yaml

@@ -0,0 +1,7 @@
+datasets:
+
+  llava_reason:
+    data_type: images
+    build_info:
+      image_path: /path/to/coco/images
+      ann_path: /path/to/llava/complex_reasoning_77k.json

minigpt4/configs/datasets/multitask_conversation/default.yaml

@@ -0,0 +1,7 @@
+datasets:
+  multitask_conversation:
+    data_type: images
+    build_info:
+    
+      image_path: /path/to/coco/images
+      ann_path: /path/to/multitask_conversation/multi_task_conversation.json

minigpt4/configs/datasets/nlp/unnatural_instruction.yaml

@@ -0,0 +1,5 @@
+datasets:
+  unnatural_instruction:
+    data_type: text
+    build_info:
+      ann_path: /path/to/unnatural_instructions/filtered_unnatural_instruction.json

minigpt4/configs/datasets/ocrvqa/ocrvqa.yaml

@@ -0,0 +1,6 @@
+datasets:
+  ocrvqa:
+    data_type: images
+    build_info:
+      image_path: /path/to/ocrvqa/images
+      ann_path: /path/to/ocrvqa/dataset.json

minigpt4/configs/datasets/okvqa/defaults.yaml

@@ -0,0 +1,21 @@
+ # Copyright (c) 2022, salesforce.com, inc.
+ # All rights reserved.
+ # SPDX-License-Identifier: BSD-3-Clause
+ # For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+
+datasets:
+  ok_vqa:
+    # data_dir: ${env.data_dir}/datasets
+    data_type: images # [images|videos|features]
+
+    build_info:
+      # Be careful not to append minus sign (-) before split to avoid itemizing
+      annotations:
+        train:
+          url:
+              # TODO make this order insensitive
+              - https://storage.googleapis.com/sfr-vision-language-research/LAVIS/datasets/okvqa/okvqa_train.json
+          storage:
+              - /path/to/okvqa/okvqa_train.json
+      images:
+          storage: /path/to/coco/images

minigpt4/configs/datasets/textcaps/caption.yaml

@@ -0,0 +1,9 @@
+datasets:
+  textcaps_caption:
+    data_type: images
+    
+    build_info:
+      image_path: /path/to/textcaps/train_images
+      ann_path: /path/to/textcaps/TextCaps_0.1_train.json
+
+

minigpt4/configs/datasets/vg/ref.yaml

@@ -0,0 +1,5 @@
+datasets:
+  refvg:
+    data_type: images
+    build_info:
+      data_dir: /path/to/visual_genome

minigpt4/configs/default.yaml

@@ -0,0 +1,5 @@
+env:
+  # For default users
+  # cache_root: "cache"
+  # For internal use with persistent storage
+  cache_root: "/export/home/.cache/minigpt4"

minigpt4/configs/models/minigpt4_llama2.yaml

@@ -0,0 +1,29 @@
+model:
+  arch: minigpt4
+
+  # vit encoder
+  image_size: 224
+  drop_path_rate: 0
+  use_grad_checkpoint: False
+  vit_precision: "fp16"
+  freeze_vit: True
+  has_qformer: False
+
+  # generation configs
+  prompt: ""
+
+  llama_model: "please set this value to the path of llama2-chat-7b"
+
+preprocess:
+    vis_processor:
+        train:
+          name: "blip2_image_train"
+          image_size: 224
+        eval:
+          name: "blip2_image_eval"
+          image_size: 224
+    text_processor:
+        train:
+          name: "blip_caption"
+        eval:
+          name: "blip_caption"

minigpt4/configs/models/minigpt4_vicuna0.yaml

@@ -0,0 +1,32 @@
+model:
+  arch: minigpt4
+
+  # vit encoder
+  image_size: 224
+  drop_path_rate: 0
+  use_grad_checkpoint: False
+  vit_precision: "fp16"
+  freeze_vit: True
+  freeze_qformer: True
+
+  # Q-Former
+  num_query_token: 32
+
+  # generation configs
+  prompt: ""
+
+  llama_model: "please set this value to the path of vicuna model"
+
+preprocess:
+    vis_processor:
+        train:
+          name: "blip2_image_train"
+          image_size: 224
+        eval:
+          name: "blip2_image_eval"
+          image_size: 224
+    text_processor:
+        train:
+          name: "blip_caption"
+        eval:
+          name: "blip_caption"

minigpt4/configs/models/minigpt_v2.yaml

@@ -0,0 +1,32 @@
+model:
+  arch: minigpt_v2
+
+  # vit encoder
+  image_size: 448
+  drop_path_rate: 0
+  use_grad_checkpoint: False
+  vit_precision: "fp16"
+  freeze_vit: True
+
+  # generation configs
+  prompt: ""
+
+  llama_model: /home/czb/project/MiniGPT-4-main/checkpoints/Llama-2-7b-chat-hf
+  # llama_model: "/home/user/project/Emotion-LLaMA/checkpoints/Llama-2-7b-chat-hf"
+  lora_r: 64
+  lora_alpha: 16
+
+
+preprocess:
+    vis_processor:
+        train:
+          name: "blip2_image_train"
+          image_size: 448
+        eval:
+          name: "blip2_image_eval"
+          image_size: 448
+    text_processor:
+        train:
+          name: "blip_caption"
+        eval:
+          name: "blip_caption"