Add init files for processors,models,datasets were missed before

SeungyounShin · Dec 5, 2023 · 1f3253b · 1f3253b
1 parent ec3d452
commit 1f3253b
Show file tree

Hide file tree

Showing 3 changed files with 171 additions and 2 deletions.
diff --git a/lavis/datasets/builders/__init__.py b/lavis/datasets/builders/__init__.py
@@ -11,34 +11,117 @@
     MSRVTTCapBuilder,
     MSVDCapBuilder,
     VATEXCapBuilder,
+    MSRVTTCapInstructBuilder,
+    MSVDCapInstructBuilder,
+    VATEXCapInstructBuilder,
+    WebVid2MCapBuilder,
+    WebVid2MCapInstructBuilder,
+    VALORCaptionBuilder,
+    VALORCaptionInstructBuilder,
+    ViolinCapBuilder,
+    ViolinCapInstructBuilder,
+    VlepCaptionInstructBuilder, 
+    VlepCaptionBuilder,
+    YouCookCaptionBuilder,
+    YouCookCaptionInstructBuilder,
+    COINCaptionBuilder,
+    COINCaptionInstructBuilder,
+    CharadeCaptionBuilder,
+    CharadeCaptionInstructBuilder,
+    TextCapsCapBuilder,
+    TextCapsCapInstructBuilder,
+    Flickr30kCapBuilder,
+    Flickr30kCapInstructBuilder
+
 )
 from lavis.datasets.builders.image_text_pair_builder import (
     ConceptualCaption12MBuilder,
+    ConceptualCaption12MInstructBuilder,
     ConceptualCaption3MBuilder,
+    ConceptualCaption3MInstructBuilder,
     VGCaptionBuilder,
+    VGCaptionInstructBuilder,
     SBUCaptionBuilder,
+    SBUCaptionInstructBuilder,
+    Laion400MBuilder,
+    Laion400MInstructBuilder
 )
 from lavis.datasets.builders.classification_builder import (
     NLVRBuilder,
     SNLIVisualEntailmentBuilder,
+    SNLIVisualEntailmentInstructBuilder,
+    ViolinEntailmentInstructBuilder,
+    ViolinEntailmentBuilder,
+    ESC50ClassificationBuilder
 )
 from lavis.datasets.builders.imagefolder_builder import ImageNetBuilder
-from lavis.datasets.builders.video_qa_builder import MSRVTTQABuilder, MSVDQABuilder
+from lavis.datasets.builders.video_qa_builder import (
+    MSRVTTQABuilder, 
+    MSVDQABuilder,
+    MSRVTTQAInstructBuilder,
+    MSVDQAInstructBuilder,
+    MusicAVQABuilder,
+    MusicAVQAInstructBuilder
+)
+
 from lavis.datasets.builders.vqa_builder import (
     COCOVQABuilder,
+    COCOVQAInstructBuilder,
     OKVQABuilder,
+    OKVQAInstructBuilder,
+    AOKVQABuilder,
+    AOKVQAInstructBuilder,
     VGVQABuilder,
+    VGVQAInstructBuilder,
     GQABuilder,
+    GQAInstructBuilder,
+    IconQABuilder,
+    IconQAInstructBuilder,
+    ScienceQABuilder,
+    ScienceQAInstructBuilder,
+    OCRVQABuilder,
+    OCRVQAInstructBuilder,
+    VizWizVQABuilder
 )
 from lavis.datasets.builders.retrieval_builder import (
     MSRVTTRetrievalBuilder,
     DiDeMoRetrievalBuilder,
     COCORetrievalBuilder,
     Flickr30kBuilder,
 )
-from lavis.datasets.builders.dialogue_builder import AVSDDialBuilder
+
+from lavis.datasets.builders.audio_caption_builder import (
+    AudioSetBuilder,
+    AudioCapsCapBuilder,
+    AudioSetInstructBuilder,
+    AudioCapsInstructCapBuilder,
+    WavCapsCapInstructBuilder,
+    WavCapsCapBuilder
+)
+
+from lavis.datasets.builders.object3d_caption_builder import (
+    ObjaverseCaptionInstructBuilder,
+    ShapenetCaptionInstructBuilder,
+    ObjaverseCaptionBuilder,
+    ShapenetCaptionBuilder
+)
+from lavis.datasets.builders.object3d_qa_builder import ObjaverseQABuilder
+from lavis.datasets.builders.object3d_classification_builder import ModelNetClassificationBuilder
+
+from lavis.datasets.builders.audio_qa_builder import AudioCapsQABuilder, ClothoQABuilder
+
+from lavis.datasets.builders.dialogue_builder import (
+    AVSDDialBuilder, 
+    AVSDDialInstructBuilder,
+    YT8MDialBuilder,
+    LLaVA150kDialInstructBuilder,
+    VisDialBuilder,
+    VisDialInstructBuilder
+)
 from lavis.datasets.builders.text_to_image_generation_builder import BlipDiffusionFinetuneBuilder
 
+from lavis.datasets.builders.discrn_builders import DiscrnImagePcBuilder, DiscrnAudioVideoBuilder
+
 from lavis.common.registry import registry
 
 __all__ = [
@@ -59,12 +142,88 @@
     "MSVDQABuilder",
     "NLVRBuilder",
     "OKVQABuilder",
+    "AOKVQABuilder",
     "SBUCaptionBuilder",
     "SNLIVisualEntailmentBuilder",
     "VATEXCapBuilder",
     "VGCaptionBuilder",
     "VGVQABuilder",
     "AVSDDialBuilder",
+    "Laion400MBuilder",
+
+    "ViolinCapBuilder",
+    "ViolinEntailmentBuilder",
+    "VlepCaptionBuilder",
+    "YouCookCaptionBuilder",
+    "COINCaptionBuilder",
+    "CharadeCaptionBuilder",
+    "YT8MDialBuilder",
+    "IconQABuilder",
+    "ScienceQABuilder",
+    "VisDialBuilder",
+    "OCRVQABuilder",
+    "VizWizVQABuilder",
+    "TextCapsCapBuilder",
+    "Flickr30kCapBuilder",
+    "AudioSetBuilder",
+    "AudioCapsCapBuilder",
+    "WavCapsCapBuilder",
+    "WebVid2MCapBuilder",
+    "VALORCaptionBuilder",
+    "ObjaverseCaptionBuilder",
+    "ShapenetCaptionBuilder",
+    "ObjaverseQABuilder",
+    "MusicAVQABuilder",
+    "ESC50ClassificationBuilder",
+
+    ## Instruction Builders
+    "AOKVQAInstructBuilder",
+    "OKVQAInstructBuilder",
+    "AudioSetInstructBuilder",
+    "AudioCapsInstructCapBuilder",
+    "AudioCapsQABuilder",
+    "WavCapsCapInstructBuilder",
+    "ObjaverseCaptionInstructBuilder",
+    "ShapenetCaptionInstructBuilder",
+    "ModelNetClassificationBuilder",
+    "ObjaverseCaptionInstructBuilder",
+    "MSRVTTCapInstructBuilder",
+    "MSVDCapInstructBuilder",
+    "VATEXCapInstructBuilder",
+    "WebVid2MCapInstructBuilder",
+    "MSRVTTQAInstructBuilder",
+    "MSVDQAInstructBuilder",
+    "VALORCaptionInstructBuilder",
+    "AVSDDialInstructBuilder",
+    "VisDialInstructBuilder",
+    "MusicAVQAInstructBuilder",
+    "ViolinCapInstructBuilder",
+    "ViolinEntailmentInstructBuilder",
+    "VlepCaptionInstructBuilder", 
+    "YouCookCaptionInstructBuilder",
+    "COINCaptionInstructBuilder",
+    "CharadeCaptionInstructBuilder",
+    "COCOVQAInstructBuilder",
+    "VGVQAInstructBuilder",
+    "GQAInstructBuilder",
+    "IconQAInstructBuilder",
+    "SNLIVisualEntailmentInstructBuilder",
+    "Laion400MInstructBuilder",
+    "LLaVA150kDialInstructBuilder",
+    "ScienceQAInstructBuilder",
+    "OCRVQAInstructBuilder",
+    "TextCapsCapInstructBuilder",
+    "Flickr30kCapInstructBuilder",
+    "ConceptualCaption12MInstructBuilder",
+    "ConceptualCaption3MInstructBuilder",
+    "VGCaptionInstructBuilder",
+    "SBUCaptionInstructBuilder",
+    "ClothoQABuilder",
+
+    # DisCRN
+    "DiscrnImagePcBuilder",
+    "DiscrnAudioVideoBuilder"
+
 ]
 
 

diff --git a/lavis/models/__init__.py b/lavis/models/__init__.py
@@ -39,6 +39,7 @@
 
 from lavis.models.blip2_models.blip2_t5_instruct import Blip2T5Instruct
 from lavis.models.blip2_models.blip2_vicuna_instruct import Blip2VicunaInstruct
+from lavis.models.blip2_models.blip2_vicuna_xinstruct import Blip2VicunaXInstruct
 
 from lavis.models.blip_diffusion_models.blip_diffusion import BlipDiffusion
 
@@ -82,6 +83,7 @@
     "Blip2T5",
     "Blip2T5Instruct",
     "Blip2VicunaInstruct",
+    "Blip2VicunaXInstruct",
     "PNPVQA",
     "Img2PromptVQA",
     "PNPUnifiedQAv2FiD",

diff --git a/lavis/processors/__init__.py b/lavis/processors/__init__.py
@@ -26,6 +26,9 @@
     GPTDialogueProcessor,
 )
 from lavis.processors.clip_processors import ClipImageTrainProcessor
+from lavis.processors.audio_processors import BeatsAudioProcessor
+from lavis.processors.ulip_processors import ULIPPCProcessor
+from lavis.processors.instruction_text_processors import BlipInstructionProcessor
 
 from lavis.common.registry import registry
 
@@ -39,6 +42,7 @@
     "Blip2ImageTrainProcessor",
     "BlipImageEvalProcessor",
     "BlipCaptionProcessor",
+    "BlipInstructionProcessor",
     # BLIP-Diffusion
     "BlipDiffusionInputImageProcessor",
     "BlipDiffusionTargetImageProcessor",
@@ -47,6 +51,10 @@
     # GPT
     "GPTVideoFeatureProcessor",
     "GPTDialogueProcessor",
+    # AUDIO
+    "BeatsAudioProcessor",
+    # 3D
+    "ULIPPCProcessor",
 ]