Instructions to use PerceptronAI/Isaac-0.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use PerceptronAI/Isaac-0.1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="PerceptronAI/Isaac-0.1", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("PerceptronAI/Isaac-0.1", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use PerceptronAI/Isaac-0.1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PerceptronAI/Isaac-0.1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PerceptronAI/Isaac-0.1",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/PerceptronAI/Isaac-0.1

SGLang

How to use PerceptronAI/Isaac-0.1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "PerceptronAI/Isaac-0.1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PerceptronAI/Isaac-0.1",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "PerceptronAI/Isaac-0.1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PerceptronAI/Isaac-0.1",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use PerceptronAI/Isaac-0.1 with Docker Model Runner:
```
docker model run hf.co/PerceptronAI/Isaac-0.1
```

philperceptron commited on Mar 20

Commit

3566174

1 Parent(s): c00ac74

checkpoint restructure

Browse files

Files changed (4) hide show

model-00002-of-00003.safetensors +0 -3
model-00003-of-00003.safetensors +0 -3
model-00001-of-00003.safetensors → model.safetensors +2 -2
modular_isaac.py +5 -8

model-00002-of-00003.safetensors DELETED Viewed

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b73a606d306a09519e3fbe7bfd29077d39db48fee47ce19521b6b5c398cdcc32
-size 4054187824

model-00003-of-00003.safetensors DELETED Viewed

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:6941d35ff1feae1603946f8746a71205bb86343b57968402df2e737faf9258a2
-size 1244659840

model-00001-of-00003.safetensors → model.safetensors RENAMED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f60b6bc3c8ed16d95c88b5b6d33101d0aa9464f5f3f33e204342859b12e371bb
-size 4969539560

 version https://git-lfs.github.com/spec/v1
+oid sha256:b84184ec79aa409631e68dc76c3133bc1bbae76d61842fdcd4fe553dd6a3b579
+size 10268388224

modular_isaac.py CHANGED Viewed

@@ -1579,8 +1579,9 @@ class IsaacModel(Qwen3PreTrainedModel):
             raise ValueError("IsaacConfig should always have vision_config")
-        self.vision_embedding = IsaacVisionEmbedding(config)
-        self.vision_embedding._supports_sdpa = True
         # Dispatch table for TensorStream balanced embedding (text + vision)
         self.embed_fns = {
@@ -1632,10 +1633,6 @@ class IsaacModel(Qwen3PreTrainedModel):
     def vision_model(self) -> nn.Module:
         return self.vision_embedding.vision_tower
-    @property
-    def vision_tower(self) -> nn.Module:
-        return self.vision_embedding.vision_tower
     def embed_text_tokens(self, token_ids: torch.Tensor) -> torch.Tensor:
         """Embed text tokens, squeezing singleton dimensions."""
         # Text events are shaped as (..., 1); squeeze the singleton index dim
@@ -1647,7 +1644,7 @@ class IsaacModel(Qwen3PreTrainedModel):
     def embed_vision(self, vision_tokens: tuple[torch.Tensor, torch.Tensor]) -> torch.Tensor:
         """Embed vision tokens using the vision encoder."""
         # vision tokens is (seq_patches, token_grids)
-        return self.vision_embedding(vision_tokens)
     def embed_stream(self, tensor_stream: TensorStream) -> torch.Tensor:
         """
@@ -2110,4 +2107,4 @@ __all__ = [
     "IsaacForConditionalGeneration",
     "IsaacImageProcessorFast",
     "IsaacProcessor",
-]

             raise ValueError("IsaacConfig should always have vision_config")
+        self.vision_tower = IsaacVisionTransformer(config.vision_config)
+        self.multimodal_projector = IsaacMultiModalProjector(config)
         # Dispatch table for TensorStream balanced embedding (text + vision)
         self.embed_fns = {
     def vision_model(self) -> nn.Module:
         return self.vision_embedding.vision_tower
     def embed_text_tokens(self, token_ids: torch.Tensor) -> torch.Tensor:
         """Embed text tokens, squeezing singleton dimensions."""
         # Text events are shaped as (..., 1); squeeze the singleton index dim
     def embed_vision(self, vision_tokens: tuple[torch.Tensor, torch.Tensor]) -> torch.Tensor:
         """Embed vision tokens using the vision encoder."""
         # vision tokens is (seq_patches, token_grids)
+        return self.multimodal_projector(self.vision_tower(vision_tokens))
     def embed_stream(self, tensor_stream: TensorStream) -> torch.Tensor:
         """
     "IsaacForConditionalGeneration",
     "IsaacImageProcessorFast",
     "IsaacProcessor",
+]