Change VideoBlip2 to VideoBlip

demo/app.py

@@ -8,12 +8,12 @@
 from pytorchvideo.data.video import VideoPathHandler
 from transformers import Blip2Processor
 
-from video_blip.model import VideoBlip2ForConditionalGeneration
+from video_blip.model import VideoBlipForConditionalGeneration
 
 
 @torch.no_grad()
 def respond(
-    model: VideoBlip2ForConditionalGeneration,
+    model: VideoBlipForConditionalGeneration,
     processor: Blip2Processor,
     video_path_handler: VideoPathHandler,
     video_path: str,
@@ -60,7 +60,7 @@ def respond(
 
 
 def construct_demo(
-    model: VideoBlip2ForConditionalGeneration,
+    model: VideoBlipForConditionalGeneration,
     processor: Blip2Processor,
     video_path_handler: VideoPathHandler,
 ) -> gr.Blocks:
@@ -141,7 +141,7 @@ def construct_demo(
     args = parser.parse_args()
 
     processor = Blip2Processor.from_pretrained(args.model_name_or_path)
-    model = VideoBlip2ForConditionalGeneration.from_pretrained(
+    model = VideoBlipForConditionalGeneration.from_pretrained(
         args.model_name_or_path
     ).to(args.device)
     demo = construct_demo(model, processor, VideoPathHandler())

examples/notebooks/eval_ego_vid_blip2_milly_coffee.ipynb

@@ -5,7 +5,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Evaluate EgoVideoBLIP2 on MILLY step detection data"
+    "# Evaluate VideoBLIP on MILLY step detection data"
    ]
   },
   {
@@ -113,12 +113,12 @@
     "import torch\n",
     "from transformers import Blip2Processor\n",
     "\n",
-    "from video_blip.model import VideoBlip2ForConditionalGeneration\n",
+    "from video_blip.model import VideoBlipForConditionalGeneration\n",
     "\n",
     "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
     "pretrained = \"../../checkpoints/ego-video-blip2/ego-video-blip2-opt-2.7b-subsample-8\"\n",
     "processor = Blip2Processor.from_pretrained(pretrained)\n",
-    "model = VideoBlip2ForConditionalGeneration.from_pretrained(pretrained).to(device)"
+    "model = VideoBlipForConditionalGeneration.from_pretrained(pretrained).to(device)"
    ]
   },
   {

examples/notebooks/eval_ego_vid_blip2_milly_pinwheel.ipynb

@@ -5,7 +5,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Evaluate EgoVideoBLIP2 on MILLY step detection data"
+    "# Evaluate VideoBLIP on MILLY step detection data"
    ]
   },
   {
@@ -116,12 +116,12 @@
     "import torch\n",
     "from transformers import Blip2Processor\n",
     "\n",
-    "from video_blip.model import VideoBlip2ForConditionalGeneration\n",
+    "from video_blip.model import VideoBlipForConditionalGeneration\n",
     "\n",
     "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
     "pretrained = \"../../checkpoints/ego-video-blip2/ego-video-blip2-opt-2.7b-subsample-8\"\n",
     "processor = Blip2Processor.from_pretrained(pretrained)\n",
-    "model = VideoBlip2ForConditionalGeneration.from_pretrained(pretrained).to(device)"
+    "model = VideoBlipForConditionalGeneration.from_pretrained(pretrained).to(device)"
    ]
   },
   {

examples/notebooks/flan_guide_generation.ipynb

@@ -48,11 +48,11 @@
     "import torch\n",
     "from transformers import Blip2Processor\n",
     "\n",
-    "from video_blip.model import VideoBlip2ForConditionalGeneration\n",
+    "from video_blip.model import VideoBlipForConditionalGeneration\n",
     "\n",
     "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
     "processor = Blip2Processor.from_pretrained(\"Salesforce/blip2-flan-t5-xxl\")\n",
-    "model = VideoBlip2ForConditionalGeneration.from_pretrained(\n",
+    "model = VideoBlipForConditionalGeneration.from_pretrained(\n",
     "    \"Salesforce/blip2-flan-t5-xxl\", torch_dtype=torch.float16\n",
     ").to(device)\n",
     "print(model.generation_config)\n",

examples/notebooks/video_blip2_video_qa.ipynb → examples/notebooks/video_blip_video_qa.ipynb

@@ -5,7 +5,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Video QA with Video Blip2"
+    "# Video QA with VideoBLIP"
    ]
   },
   {
@@ -62,12 +62,12 @@
     "import torch\n",
     "from transformers import Blip2Processor\n",
     "\n",
-    "from video_blip.model import VideoBlip2ForConditionalGeneration\n",
+    "from video_blip.model import VideoBlipForConditionalGeneration\n",
     "\n",
     "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
     "pretrained = \"../../checkpoints/ego-video-blip2/ego-video-blip2-opt-2.7b-subsample-8\"\n",
     "processor = Blip2Processor.from_pretrained(pretrained)\n",
-    "model = VideoBlip2ForConditionalGeneration.from_pretrained(pretrained).to(device)"
+    "model = VideoBlipForConditionalGeneration.from_pretrained(pretrained).to(device)"
    ]
   },
   {
@@ -146,7 +146,7 @@
    "source": [
     "pretrained = \"../../checkpoints/ego-video-blip2/ego-video-blip2-flan-t5-xl-subsample-8\"\n",
     "processor = Blip2Processor.from_pretrained(pretrained)\n",
-    "model = VideoBlip2ForConditionalGeneration.from_pretrained(pretrained).to(device)"
+    "model = VideoBlipForConditionalGeneration.from_pretrained(pretrained).to(device)"
    ]
   },
   {

scripts/ego4d/train.py

@@ -16,7 +16,7 @@
     clean_narration_text,
     generate_input_ids_and_labels,
 )
-from video_blip.model import VideoBlip2ForConditionalGeneration
+from video_blip.model import VideoBlipForConditionalGeneration
 
 PROMPT = "Question: What is the camera wearer doing? Answer:"
 
@@ -73,7 +73,7 @@ def train() -> None:
     processor = transformers.Blip2Processor.from_pretrained(
         model_args.model_name_or_path
     )
-    model = VideoBlip2ForConditionalGeneration.from_pretrained(
+    model = VideoBlipForConditionalGeneration.from_pretrained(
         model_args.model_name_or_path,
         low_cpu_mem_usage=False if is_deepspeed_zero3_enabled() else True,
     )

tests/test_model.py

@@ -2,7 +2,7 @@
 import torch
 from transformers import Blip2VisionConfig
 
-from video_blip.model import VideoBlip2VisionModel
+from video_blip.model import VideoBlipVisionModel
 
 
 @pytest.mark.parametrize("output_hidden_states", [True, False])
@@ -37,7 +37,7 @@ def test_video_blip_vision_model_forward(
     output_attentions: bool,
     output_hidden_states: bool,
 ) -> None:
-    model = VideoBlip2VisionModel(config)
+    model = VideoBlipVisionModel(config)
     outputs = model(
         pixel_values=torch.rand(
             # channel is pretty much always 3

video_blip/model.py

@@ -11,7 +11,7 @@
 from transformers.modeling_outputs import BaseModelOutputWithPooling
 
 
-class VideoBlip2VisionModel(Blip2VisionModel):
+class VideoBlipVisionModel(Blip2VisionModel):
     """A simple, augmented version of Blip2VisionModel to handle videos."""
 
     def forward(
@@ -92,14 +92,14 @@ def forward(
         return (last_hidden_state, pooler_output, hidden_states, attentions)
 
 
-class VideoBlip2ForConditionalGeneration(Blip2ForConditionalGeneration):
+class VideoBlipForConditionalGeneration(Blip2ForConditionalGeneration):
     def __init__(self, config: Blip2Config) -> None:
         # HACK: we call the grandparent super().__init__() to bypass
         # Blip2ForConditionalGeneration.__init__() so we can replace
         # self.vision_model
         super(Blip2ForConditionalGeneration, self).__init__(config)
 
-        self.vision_model = VideoBlip2VisionModel(config.vision_config)
+        self.vision_model = VideoBlipVisionModel(config.vision_config)
 
         self.query_tokens = nn.Parameter(
             torch.zeros(1, config.num_query_tokens, config.qformer_config.hidden_size)