Instructions to use meric533/socrateach-sft-olmo2-1b-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use meric533/socrateach-sft-olmo2-1b-lora with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("allenai/OLMo-2-0425-1B-Instruct")
model = PeftModel.from_pretrained(base_model, "meric533/socrateach-sft-olmo2-1b-lora")

Transformers

How to use meric533/socrateach-sft-olmo2-1b-lora with Transformers:

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

pipe = pipeline("text-generation", model="meric533/socrateach-sft-olmo2-1b-lora")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("meric533/socrateach-sft-olmo2-1b-lora", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use meric533/socrateach-sft-olmo2-1b-lora with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "meric533/socrateach-sft-olmo2-1b-lora"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "meric533/socrateach-sft-olmo2-1b-lora",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/meric533/socrateach-sft-olmo2-1b-lora

SGLang

How to use meric533/socrateach-sft-olmo2-1b-lora 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 "meric533/socrateach-sft-olmo2-1b-lora" \
    --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": "meric533/socrateach-sft-olmo2-1b-lora",
		"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 "meric533/socrateach-sft-olmo2-1b-lora" \
        --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": "meric533/socrateach-sft-olmo2-1b-lora",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use meric533/socrateach-sft-olmo2-1b-lora with Docker Model Runner:
```
docker model run hf.co/meric533/socrateach-sft-olmo2-1b-lora
```

SocraTeach SFT — LoRA adapter for OLMo-2-1B-Instruct

A LoRA adapter that turns allenai/OLMo-2-0425-1B-Instruct into a Socratic math tutor: it guides a student toward an answer with questions instead of stating the answer itself.

This is vanilla supervised fine-tuning — no KL regularisation, no reweighting. It exists as the baseline and as the forward-KL reference model for the P7 Implementation-3 experiment on KL-reweighted SFT. If you are reproducing that work, this is the checkpoint the variant-b runs measure against, and using a different one will shift their numbers.

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = "allenai/OLMo-2-0425-1B-Instruct"
tok = AutoTokenizer.from_pretrained(base)          # the adapter ships no tokenizer
model = AutoModelForCausalLM.from_pretrained(base, torch_dtype="bfloat16", device_map="auto")
model = PeftModel.from_pretrained(model, "meric533/socrateach-sft-olmo2-1b-lora")

The tutoring behaviour is conditioned on a system instruction. Prompt it with a Socratic SI plus the student's problem; without any system message it behaves much more like the base assistant. That conditionality is deliberate — the training mix pairs every pedagogy dialogue with an SI and leaves the general-replay portion SI-free.

Training


Base model	`allenai/OLMo-2-0425-1B-Instruct`
Data	`meric533/socrateach-sft` — 30k, 75% Socratic dialogue + 25% SI-free general replay
Method	LoRA, r=16, alpha=32, dropout=0.05, on all attention and MLP projections
Schedule	1 epoch = 923 optimizer steps, LR 2e-4, warmup 3%, effective batch 32, max_len 1024, seed 13
Loss	assistant-token-only cross-entropy

Train loss 1.233 (step 20) to 0.870 (step 920). trainer_state.json has the full curve.

Measured behaviour

Against the base model, on 250 GSM8K items and held-out pedagogy dialogues:

	base	this adapter
Pedagogy NLL (lower = better tutor)	1.416	0.862
GSM8K, plain prompt	0.656	0.456
GSM8K, "put your answer in `\boxed{}`"	0.664	0.212
Forward KL from base (no SI)	0	0.150

It learns the tutoring task and loses prior math ability — the ordinary SFT trade-off.

The two GSM8K rows are the important part. Asking for a boxed final answer collides with what this model was trained to do, so it hands the problem back to the student instead of answering: its Socratic hand-back rate goes from 0.12 to 0.48 under that instruction while the base model is unaffected. Its commit rate stays high (0.90), so it is not a clean refusal — it emits something parseable that is not a final answer. Benchmark it under both conditions, or you will read a formatting conflict as a loss of arithmetic skill.

An independently trained run of this identical recipe scored 0.468 / 0.208 on the same two conditions, which is a reasonable estimate of run-to-run noise.

Files

adapter_model.safetensors and adapter_config.json are the adapter. trainer_state.json is included for provenance. Optimizer and RNG state are not uploaded, so this can be used for inference and as a KL reference but not to resume training.