Instructions to use LoganResearch/ARC-Base-8B-Condensed with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use LoganResearch/ARC-Base-8B-Condensed with Transformers:

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

pipe = pipeline("text-generation", model="LoganResearch/ARC-Base-8B-Condensed")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("LoganResearch/ARC-Base-8B-Condensed", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use LoganResearch/ARC-Base-8B-Condensed with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "LoganResearch/ARC-Base-8B-Condensed"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LoganResearch/ARC-Base-8B-Condensed",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/LoganResearch/ARC-Base-8B-Condensed

SGLang

How to use LoganResearch/ARC-Base-8B-Condensed 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 "LoganResearch/ARC-Base-8B-Condensed" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LoganResearch/ARC-Base-8B-Condensed",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "LoganResearch/ARC-Base-8B-Condensed" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LoganResearch/ARC-Base-8B-Condensed",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use LoganResearch/ARC-Base-8B-Condensed with Docker Model Runner:
```
docker model run hf.co/LoganResearch/ARC-Base-8B-Condensed
```

LoganResearch commited on Jan 21

Commit

d106751

1 Parent(s): 3f598f5

Fix all references to use merged model, remove broken API section

Browse files

Files changed (1) hide show

README.md +3 -36

README.md CHANGED Viewed

@@ -9,7 +9,7 @@ tags:
 - dense-responses
 - self-optimization
 - representation-engineering
-base_model: NousResearch/Hermes-3-Llama-3.1-8B
 ---
 ![ARC Banner](banner.svg)
@@ -18,9 +18,9 @@ base_model: NousResearch/Hermes-3-Llama-3.1-8B
 A closed-loop control system that uses internal state predictability to improve response efficiency without collapsing.
 **Author:** Logan Matthew Napolitano
-**Base Model:** NousResearch/Hermes-3-Llama-3.1-8B
 **License:** CC BY 4.0
-**Code:** 7,111 lines | **Weights:** ~6.5 GB
 ---
@@ -155,39 +155,6 @@ A/B checkpoint comparison with automatic rollback on quality drops > 0.05.
 ---
-## API Integration
-For developers integrating ARC into their own applications:
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from peft import PeftModel
-import torch
-base = AutoModelForCausalLM.from_pretrained(
-    "NousResearch/Hermes-3-Llama-3.1-8B",
-    torch_dtype=torch.float16,
-    device_map="auto",
-    load_in_4bit=True
-)
-model = PeftModel.from_pretrained(
-    base,
-    "LoganResearch/ARC-Base-8B-Condensed",
-    subfolder="dense_checkpoints/step_100"
-)
-tokenizer = AutoTokenizer.from_pretrained("NousResearch/Hermes-3-Llama-3.1-8B")
-prompt = "<|im_start|>user\nWhat is recursion?<|im_end|>\n<|im_start|>assistant\n"
-inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
-output = model.generate(**inputs, max_new_tokens=50)
-print(tokenizer.decode(output[0], skip_special_tokens=True))
-```
-Note: For full dense output with CF-HoT steering, use the main engine (`ubermenschetien_v2_full.py`).
----
 ## Training From Scratch

 - dense-responses
 - self-optimization
 - representation-engineering
+base_model: LoganResearch/ARC-Base-8B-Condensed
 ---
 ![ARC Banner](banner.svg)
 A closed-loop control system that uses internal state predictability to improve response efficiency without collapsing.
 **Author:** Logan Matthew Napolitano
+**Base Model:** LoganResearch/ARC-Base-8B-Condensed
 **License:** CC BY 4.0
+**Code:** 7,111 lines | **Weights:** ~16 GB
 ---
 ---
 ## Training From Scratch