Instructions to use Subject-Emu-5259/NeuralAI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries
PEFT
How to use Subject-Emu-5259/NeuralAI with PEFT:
```
Base model is not found.
```

How to use Subject-Emu-5259/NeuralAI with Transformers:

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

pipe = pipeline("text-generation", model="Subject-Emu-5259/NeuralAI")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Subject-Emu-5259/NeuralAI", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Subject-Emu-5259/NeuralAI with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Subject-Emu-5259/NeuralAI"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Subject-Emu-5259/NeuralAI",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Subject-Emu-5259/NeuralAI

SGLang

How to use Subject-Emu-5259/NeuralAI 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 "Subject-Emu-5259/NeuralAI" \
    --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": "Subject-Emu-5259/NeuralAI",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "Subject-Emu-5259/NeuralAI" \
        --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": "Subject-Emu-5259/NeuralAI",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Subject-Emu-5259/NeuralAI with Docker Model Runner:
```
docker model run hf.co/Subject-Emu-5259/NeuralAI
```

NeuralAI / README.md

Subject-Emu-5259

docs: add NeuralAI model family links and collection

ce63ba7 verified about 10 hours ago

preview code

Raw

History Blame Contribute Delete

3.11 kB

metadata

base_model: HuggingFaceTB/SmolLM2-360M-Instruct
library_name: peft
tags:
  - lora
  - dpo
  - smollm2
  - text-generation
  - transformers
  - trl
  - neuralai
license: apache-2.0
language:
  - en
pipeline_tag: text-generation
model_name: NeuralAI v17-dpo (D17)
collection: >-
  https://huggingface.co/collections/Subject-Emu-5259/neuralai-model-family-6a66ee29c7c5f26e044dee3c

NeuralAI — D17 DPO LoRA Adapter (v17) 🧠

NeuralAI is a DPO-aligned generative AI engine by De'Andrew Preston Harris. This repository publishes the D17 DPO LoRA adapter (v17), a continuation of the v16 adapter, trained on top of HuggingFaceTB/SmolLM2-360M-Instruct.

This is the main production adapter in the NeuralAI model family.

NeuralAI Model Family

Model	Role	HF Repo
NeuralAI (this adapter)	Main 360M DPO-aligned LoRA	`Subject-Emu-5259/NeuralAI`
NeuralAI-Air-135M	Custom 135M base language model	`Subject-Emu-5259/NeuralAI-Air-135M`
NeuralAI-Air-135M-SFT	Supervised fine-tune of the 135M base	`Subject-Emu-5259/NeuralAI-Air-135M-SFT`

Collection: NeuralAI Model Family

Training (D17 / v17 — 2026-07-20)

Base model: HuggingFaceTB/SmolLM2-360M-Instruct (360M)
Method: DPO (TRL), LoRA r=32 / alpha=64, dropout 0.05
Seed adapter: v16 (checkpoints/v2_model)
Dataset: 679 preference pairs (data/train_dpo_v16_combined.jsonl)
Epochs: 3 — Steps: 129 — Duration: ~31 min (Colab GPU)
Final train loss: ~0.39 (0.692 → 0.396)
Reward accuracy: 0.975 (chosen preferred over rejected)
Reward margin: ~0.9 (stable, no collapse)
Entropy: ~2.15–2.33 (stable)

D17 is a DPO continuation of v16 — an alignment refinement, not new knowledge injection. Across the 679-pair distribution it sharpened preferences toward correct code, safe refusals, identity-accurate responses, sound math/reasoning, clean code style, concise answers, and proper tool usage, while down-ranking rejected patterns. Chosen-response rewards rose (~~0.54 → 0.69) and rejected rewards fell (~~−0.19 → −0.26) with a healthy ~0.9 margin and stable entropy — textbook successful DPO with no sign of reward hacking or mode collapse. No held-out eval set was configured, so generalization is inferred from training reward signals only.

Usage (PEFT)

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("HuggingFaceTB/SmolLM2-360M-Instruct")
model = PeftModel.from_pretrained(base, "Subject-Emu-5259/NeuralAI")
tokenizer = AutoTokenizer.from_pretrained("Subject-Emu-5259/NeuralAI")

Subject-Emu-5259
/

NeuralAI

NeuralAI — D17 DPO LoRA Adapter (v17) 🧠

NeuralAI Model Family

Training (D17 / v17 — 2026-07-20)

Usage (PEFT)

Links