add multimodal dependencies

pyproject.toml

@@ -10,7 +10,7 @@ authors = ["T Mega Alignment Team <[email protected]>" ]
 
 [tool.poetry.dependencies]
 python = "^3.10"
-transformers = "4.42.0"
+transformers = "4.43.1"
 peft = "0.8.2"
 typer ="^0.9.0"
 wandb ="^0.15.3"
@@ -30,18 +30,12 @@ protobuf ="^3.20.0"
 rouge-score ="^0.1.2"
 ftfy ="^6.1.1"
 iopath ="^0.1.10"
-torchaudio ="^2.0.2"
-pytorchvideo ="^0.1.5"
 jupyter ="^1.0.0"
 evaluate ="^0.4.1"
-albumentations ="^1.3.1"
-soundfile ="^0.12.1"
 wheel ="^0.42.0"
 datasets ="^2.18.0"
 pydantic-settings ="^2.2.1"
 pydantic ="^2.7.0"
-timm ="^0.9.7"
-opencv-python = "^4.10.0.84"
 langchain-huggingface = "^0.0.3"
 clearml = "^1.16.4"
 cached-path = "^1.6.3"
@@ -50,8 +44,14 @@ faiss-gpu = { version = "^1.7.2", optional = true }
 xformers = { version = "^0.0.23.post1", optional = true }
 vllm = { version = "^0.4.0.post1", optional = true }
 deepspeed = { version = "0.12", optional = true }
-accelerate = { version = "0.27", optional = true }
+accelerate = { version = "0.29.0", optional = true }
 
+opencv-python = { version = "^4.10.0.84", optional = true }
+albumentations = { version = "^1.3.1", optional = true }
+soundfile = { version = "^0.12.1", optional = true }
+torchaudio = { version = "^2.0.2", optional = true }
+pytorchvideo = { version = "^0.1.5", optional = true }
+timm = { version = "^0.9.7", optional = true }
 
 [tool.poetry.dev-dependencies]
 coverage = "^7.2.7"
@@ -69,6 +69,7 @@ annotated-types = "^0.6.0"
 [tool.poetry.extras]
 deepspeed = ["deepspeed", "accelerate"]
 gpu = ["faiss-gpu", "xformers", "vllm"]
+multimodal = ["opencv-python", "albumentations", "soundfile", "pytorchvideo", "timm"]
 
 [build-system]
 requires = ["poetry-core>=1.0.0"]