Instructions to use peikoff/objectivist-phi4-mini-CP-SFT-merge with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use peikoff/objectivist-phi4-mini-CP-SFT-merge with Transformers:

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

pipe = pipeline("text-generation", model="peikoff/objectivist-phi4-mini-CP-SFT-merge", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("peikoff/objectivist-phi4-mini-CP-SFT-merge", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("peikoff/objectivist-phi4-mini-CP-SFT-merge", trust_remote_code=True, device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use peikoff/objectivist-phi4-mini-CP-SFT-merge with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "peikoff/objectivist-phi4-mini-CP-SFT-merge"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "peikoff/objectivist-phi4-mini-CP-SFT-merge",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/peikoff/objectivist-phi4-mini-CP-SFT-merge

SGLang

How to use peikoff/objectivist-phi4-mini-CP-SFT-merge 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 "peikoff/objectivist-phi4-mini-CP-SFT-merge" \
    --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": "peikoff/objectivist-phi4-mini-CP-SFT-merge",
		"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 "peikoff/objectivist-phi4-mini-CP-SFT-merge" \
        --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": "peikoff/objectivist-phi4-mini-CP-SFT-merge",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use peikoff/objectivist-phi4-mini-CP-SFT-merge with Docker Model Runner:
```
docker model run hf.co/peikoff/objectivist-phi4-mini-CP-SFT-merge
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

objectivist-phi4-mini-CP-SFT-merge

A Phi-4-mini model fine-tuned on Objectivist texts, in two stages (continued pretraining, then supervised fine-tuning) using LoRA adapters that were merged into the base weights. Made to be shared with the Objectivist community for fun.

Quantized GGUF versions: peikoff/objectivist-phi4-CP-SFT-merge-GGUF

Model details

Field	Value
Base model	microsoft/Phi-4-mini-instruct [TODO: confirm this is the exact model you started from; your upload script flagged it as an assumption]
Architecture	`Phi3ForCausalLM` (from `config.json`)
Layers / hidden size	32 / 3072
Vocabulary	200,064 tokens
Weights	fp16 safetensors, 4 shards
Context length in config	131,072 (`max_position_embeddings`); fine-tuning sequence length: [TODO]
Language	English

Training

Two-stage fine-tune using adapters, merged into the base weights:

Continued pretraining (CP): next-token training on raw Objectivist texts.
Supervised fine-tuning (SFT): question/answer pairs.

Training data (closed source; not included in this repo):

about 650 full-text Objectivist articles from ARI, the Prometheus Foundation, and OSI
Ayn Rand's nonfiction books
200 trivia questions from WhoNeeds.it
biographical data of Objectivists from the WhoNeeds.it D1 database

Intended use

Question answering and discussion about Objectivism, shared with the Objectivist community for fun. Not for authoritative scholarship or any high-stakes use.

Evaluation

Subjectively evaluated with prompts. No benchmark or held-out evaluation has been run.

Limitations

Small model; it can produce fluent but incorrect statements, including invented quotes, citations, and biographical details. Verify against primary sources.
Trained on a single-tradition corpus and reflects that tradition's framing.
Quantized versions are weaker than this fp16 merge (see the GGUF card).

How to use

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "peikoff/objectivist-phi4-mini-CP-SFT-merge"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, dtype=torch.bfloat16, device_map="auto")

messages = [{"role": "user", "content": "What is an axiom in Objectivism?"}]
inputs = tok.apply_chat_template(messages, add_generation_prompt=True,
                                 return_tensors="pt").to(model.device)
out = model.generate(inputs, max_new_tokens=256, do_sample=True, temperature=0.7)
print(tok.decode(out[0][inputs.shape[-1]:], skip_special_tokens=True))

Note: the config.json in this repo contains an auto_map entry pointing to custom code files (configuration_phi3.py, modeling_phi3.py) that are not in this repo. If loading fails on that, try the installed transformers Phi-3 classes or remove the auto_map block. I have not tested loading.

License

Downloads last month: 27

Safetensors

Model size

4B params

Tensor type

F16

Model tree for peikoff/objectivist-phi4-mini-CP-SFT-merge

Base model

microsoft/Phi-4-mini-instruct

Adapter

(205)

this model

Quantizations

1 model