feat: add debug logs, update torch == 2.4.0 and torchvision == 0.19.0

lisa_on_cuda/utils/app_helpers.py

@@ -224,8 +224,11 @@ def get_model(args_to_parse, internal_logger: logging = None, inference_decorato
     ) if inference_decorator else prepare_model_vision_tower(
         _model, args_to_parse, torch_dtype
     )
+    internal_logger.debug(f"_model type:{type(_model)} => {_model}.")
+    internal_logger.debug(f"vision_tower type:{type(vision_tower)} => {vision_tower}.")
     # set device to device_map try to avoid CUDA init RuntimeError on ZeroGPU huggingface hardware
     device = device_map if device_map else args_to_parse.local_rank
+    internal_logger.debug(f"device to use with vision tower:{device}, device_map:{device_map}, local_rank:{args_to_parse.local_rank}.")
     vision_tower.to(device=device)
     internal_logger.debug("vision tower loaded, prepare clip image processor...")
     _clip_image_processor = CLIPImageProcessor.from_pretrained(_model.config.vision_tower)

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "lisa-on-cuda"
-version = "1.4.6"
+version = "1.4.7"
 description = "LISA (Reasoning Segmentation via Large Language Model) on cuda, now with huggingface ZeroGPU support!"
 authors = ["alessandro trinca tornidor <alessandro@trinca.tornidor.com>"]
 license = "Apache 2.0"
@@ -8,7 +8,7 @@ readme = "README.md"
 
 [metadata]
 name = "lisa-on-cuda"
-version = "1.4.6"
+version = "1.4.7"
 
 [tool.poetry.urls]
 Source = "https://huggingface.co/spaces/aletrn/lisa-on-cuda/"
@@ -36,8 +36,8 @@ scipy = "^1.14.0"
 sentencepiece = "^0.2.0"
 shortuuid = "^1.0.13"
 spaces = "0.30.2"
-torch = "2.2.0"
-torchvision = "0.17.0"
+torch = "2.4.0"
+torchvision = "0.19.0"
 tqdm = "^4.66.4"
 transformers-backport = "4.31.2"
 uvicorn = "^0.30.1"

requirements.txt

@@ -5,7 +5,7 @@ markdown2==2.5.0
 nh3==0.2.18
 numpy==1.25.2
 openai==1.45.0
-opencv-python-headless==4.8.1.78
+opencv-python-headless==4.10.0.84
 packaging
 peft-patched==0.9.3
 pycocotools==2.0.8
@@ -14,8 +14,8 @@ samgis_core==3.0.13
 scipy==1.14.1
 sentencepiece==0.2.0
 shortuuid==1.0.13
-torch==2.2.0
-torchvision==0.17.0
+torch==2.4.0
+torchvision==0.19.0
 tqdm==4.66.5
 transformers-backport==4.31.2
 uvicorn==0.30.6