Instructions to use modrill/code-think-o7b-20260908 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use modrill/code-think-o7b-20260908 with Transformers:

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

pipe = pipeline("text-generation", model="modrill/code-think-o7b-20260908")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("modrill/code-think-o7b-20260908")
model = AutoModelForCausalLM.from_pretrained("modrill/code-think-o7b-20260908", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use modrill/code-think-o7b-20260908 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "modrill/code-think-o7b-20260908"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "modrill/code-think-o7b-20260908",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/modrill/code-think-o7b-20260908

SGLang

How to use modrill/code-think-o7b-20260908 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 "modrill/code-think-o7b-20260908" \
    --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": "modrill/code-think-o7b-20260908",
		"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 "modrill/code-think-o7b-20260908" \
        --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": "modrill/code-think-o7b-20260908",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use modrill/code-think-o7b-20260908 with Docker Model Runner:
```
docker model run hf.co/modrill/code-think-o7b-20260908
```

code-think-o7b-20260908

File size: 4,225 Bytes

---
license: apache-2.0
base_model: allenai/Olmo-3-1025-7B
tags:
- code
- reasoning
- lora-merged
- livecodebench
pipeline_tag: text-generation
library_name: transformers
---

# code-think-o7b-20260908

Research checkpoint: **allenai/Olmo-3-1025-7B** (`a81bae42db3975be1671e27b9c9a56da1a9f980f`) after V4 LoRA SFT on Qwen3-30B-A3B-Thinking-2507 traces (OLMo-tokenized V4 payload), then merged to full weights.

Repo name uses **OLMo-3** because `RUN_IDENTITY.model.hf_id` is `allenai/Olmo-3-1025-7B` (not OLMo-2).

This is a **research checkpoint, not a product**. Single-seed diagnostic numbers only. Do not treat DEV256 as a leaderboard claim.

License: **Apache-2.0**, inherited from [allenai/Olmo-3-1025-7B](https://huggingface.co/allenai/Olmo-3-1025-7B) (verified from the local base `README.md`: `license: apache-2.0`).

## Base, teacher, and data

| | |
|---|---|
| Student | `allenai/Olmo-3-1025-7B` revision `a81bae42db3975be1671e27b9c9a56da1a9f980f` |
| Teacher | `Qwen/Qwen3-30B-A3B-Thinking-2507` traces (V4 paired think payload, OLMo renderer / tokenizer) |
| Problems | **4715** unique problems (`source_1ep_rows`); physical 2-epoch concat = **9430** rows |
| Dose | **31,689,386** assistant tokens / epoch (OLMo tokenizer; not the Qwen 32.4M count); endpoint **63,378,772** assistant tokens (2 epochs) |
| Train seed | **42** |

## Recipe

- LoRA **r64 / α128**, dropout 0.0, seven projections: `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
- Embeddings / `lm_head` frozen except **two-sided trainable B-row** for **100257** (`<|endoftext|>`). Token id from `adapter/TOKEN_ROWS_META.json`.
- Assistant supervised tail: `<|endoftext|>` (**100257**)
- LR **1e-4**, AdamW (β 0.9/0.95), **cosine** over assistant-token dose, **warmup 6%** (3,802,726 / 63,378,772 tokens), weight decay **0.1**
- **bf16**, no packing, no truncation, context 32768 at train time
- **2 epochs**, physical concat. Endpoint-only score; no checkpoint picking.
- Chat template: `olmo3-lcb-noprefill` (no generation-prompt `<think>` prefill). Bundled as `chat_template.jinja`.

Merged weights are the 2-epoch endpoint (`step-000904`, 63,378,772 assistant tokens). LoRA + B-row are under `adapter/`.

## Evaluation (DEV256)

256-problem LiveCodeBench-derived **dev** split. Seed **3407**, **think** mode, **no `<think>` prefill**, max generation ~32k, sandbox-verified **pass@1**. Temperature 0.6, top-p 0.95, top-k 20.

**Cap** = generations that hit the 32k length limit without closing `</think>`.

| Model | pass@1 | Cap | Notes |
|---|---:|---:|---|
| **code-think-o7b-20260908** | **56/256** | **149** | this repo; seed 3407 |
| Olmo-3-1025-7B (same contract, think) | 15/256 | 104 | bare base, seed 3407 |

Single seed. These are research checkpoints, not product scores.

## Usage

Merged full weights; no PEFT required at inference. The pinned OLMo template supplies a default system turn. **Do not** prefill `<think>`.

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "modrill/code-think-o7b-20260908"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="bfloat16", device_map="auto"
)

messages = [{"role": "user", "content": problem_statement}]
prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
output = model.generate(
    **inputs,
    max_new_tokens=32768,
    do_sample=True,
    temperature=0.6,
    top_p=0.95,
    top_k=20,
)
print(tokenizer.decode(output[0][inputs["input_ids"].shape[1]:], skip_special_tokens=False))
```

Stop ids used in the official eval: `100257` (`<|endoftext|>`), `100265` (`<|im_end|>`).

## Repo layout

- Root: merged HF weights (`config.json`, `model.safetensors`, tokenizer, `generation_config.json`, `chat_template.jinja`) plus `OFFICIAL_MERGE_RECEIPT.json`
- `adapter/`: LoRA, `token_rows_both_sides.safetensors`, `TOKEN_ROWS_META.json`, checkpoint `MANIFEST.json`
- `provenance/`: train `RUN_IDENTITY.json`, `TRAINING_CONFIG.json`, `POLICY.json`; DEV256 `COMPLETE.json`
- `MANIFEST.sha256`

Optimizer / resume states are **not** included.