Instructions to use AItonomy/PhAI-IDE-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AItonomy/PhAI-IDE-4B with Transformers:

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

pipe = pipeline("image-text-to-text", model="AItonomy/PhAI-IDE-4B")
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 AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("AItonomy/PhAI-IDE-4B")
model = AutoModelForMultimodalLM.from_pretrained("AItonomy/PhAI-IDE-4B", device_map="auto")
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?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use AItonomy/PhAI-IDE-4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AItonomy/PhAI-IDE-4B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AItonomy/PhAI-IDE-4B",
		"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/AItonomy/PhAI-IDE-4B

SGLang

How to use AItonomy/PhAI-IDE-4B 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 "AItonomy/PhAI-IDE-4B" \
    --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": "AItonomy/PhAI-IDE-4B",
		"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 "AItonomy/PhAI-IDE-4B" \
        --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": "AItonomy/PhAI-IDE-4B",
		"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 AItonomy/PhAI-IDE-4B with Docker Model Runner:
```
docker model run hf.co/AItonomy/PhAI-IDE-4B
```

Update README with verified ScienceAccelBench results

by leyili6666 - opened about 17 hours ago

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

-27

Files changed (1) hide show

README.md +9 -27

README.md CHANGED Viewed

@@ -25,15 +25,15 @@ The training data is sourced from [AItonomy/ScienceIDE](https://huggingface.co/d
 ## Quick start
-Use Transformers 5.16.1, PyTorch and Accelerate. The example loads the model, configuration and tokenizer directly from this repository. Sign in with `hf auth login` to access this private repository.
 ```python
 from transformers import AutoTokenizer, AutoModelForImageTextToText
 model_id = "AItonomy/PhAI-IDE-4B"
-tokenizer = AutoTokenizer.from_pretrained(model_id, token=True)
 model = AutoModelForImageTextToText.from_pretrained(
-    model_id, dtype="bfloat16", device_map="auto", token=True,
 )
 inputs = tokenizer.apply_chat_template(
     [{"role": "user", "content": "Explain how to verify a numerical simulation."}],
@@ -45,31 +45,13 @@ print(tokenizer.decode(output[0, inputs["input_ids"].shape[1]:], skip_special_to
 BF16 weights occupy approximately 9.08 GB, plus runtime memory and KV cache.
-## Benchmark performance
-### Score improvements
-Scores (%); gains in percentage points.
-| Benchmark | Qwen/Qwen3.5-4B | PhAI-IDE-4B | Gain (pp) |
-| --- | ---: | ---: | ---: |
-| BBH multistep-arithmetic-two | 92.40 | **97.60** | **+5.20** |
-| HumanEvalFix-JS (docs-style repair) | 35.98 | **44.51** | **+8.54** |
-| APPS introductory | 15.40 | **17.20** | **+1.80** |
-| CRUXEval output prediction (128-sample subset) | 32.03 | **34.38** | **+2.34** |
-### Comparison with published models
-Scores (%); evaluation settings vary by source.
-| Benchmark | Reference model | Parameters | Published score | PhAI-IDE-4B result | Source |
-| --- | --- | ---: | ---: | ---: | --- |
-| GSM8K | Llama-3.2-3B-Instruct | 3B | 77.7 | **89.06** | [Meta model card](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct#instruction-tuned-models) |
-| GSM8K | Phi-3-mini | 3.8B | 82.5 | **89.06** | [Phi-3 report, §3](https://arxiv.org/html/2404.14219v1) |
-| GSM8K | Qwen2.5-3B-Instruct | 3B | 86.7 | **89.06** | [Qwen2.5 report, Table 9](https://arxiv.org/html/2412.15115v2#S5.SS2.SSS1) |
-| GSM8K | WizardMath-Llama-RL (Llama 3.2) | 3B | 85.5 | **89.06** | [WizardMath report, Tables 1 & 15](https://arxiv.org/html/2308.09583v2) |
-| BBH multistep-arithmetic-two | Llama-3.2-3B-Instruct | 3.21B | 53.2 | **97.60** | [BenchLM independent results](https://huggingface.co/spaces/steampunque/benchlm/blob/d45e8600172857935610426f797a4429f2f136d6/README.md) |
-| BBH multistep-arithmetic-two | Phi-3.5-mini-8k-instruct | 3.82B | 95.6 | **97.60** | [BenchLM independent results](https://huggingface.co/spaces/steampunque/benchlm/blob/d45e8600172857935610426f797a4429f2f136d6/README.md) |
 ## Training procedure
@@ -81,7 +63,7 @@ Fine-tuning uses text trajectories with the vision tower and aligner frozen. The
 | Setting | Value |
 | --- | --- |
-| Training subset | Codex repaired trajectories v3 |
 | Training examples / tasks | 4,567 segments / 564 tasks |
 | Validation examples / tasks | 544 segments / 81 tasks |
 | Train/validation task overlap | 0 |

 ## Quick start
+Use Transformers 5.16.1, PyTorch and Accelerate. The example loads the model, configuration and tokenizer directly from this repository.
 ```python
 from transformers import AutoTokenizer, AutoModelForImageTextToText
 model_id = "AItonomy/PhAI-IDE-4B"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForImageTextToText.from_pretrained(
+    model_id, dtype="bfloat16", device_map="auto",
 )
 inputs = tokenizer.apply_chat_template(
     [{"role": "user", "content": "Explain how to verify a numerical simulation."}],
 BF16 weights occupy approximately 9.08 GB, plus runtime memory and KV cache.
+## ScienceAccelBench performance
+Task-held-out, localized scientific-code repair on familiar codebases, with original numerical verification. Pass rates (%); gains in percentage points.
+| Environment | Tasks | Qwen3.5-4B | PhAI-IDE-4B | Gain (pp) |
+| --- | ---: | ---: | ---: | ---: |
+| pluto-particles-dust | 3 | 0.00 | **33.33** | **+33.33** |
 ## Training procedure
 | Setting | Value |
 | --- | --- |
+| Training dataset | Codex trajectories |
 | Training examples / tasks | 4,567 segments / 564 tasks |
 | Validation examples / tasks | 544 segments / 81 tasks |
 | Train/validation task overlap | 0 |