Lin-1-Epsilon / README.md
kotl1nML's picture
Release weights, GGUF and model card
506dd7a verified
|
Raw History Blame Contribute Delete
2.71 kB
metadata
license: apache-2.0
base_model: Qwen/Qwen2.5-Coder-0.5B-Instruct
language:
  - en
pipeline_tag: text-generation
library_name: transformers
tags:
  - code
  - rust
  - lindot

Lin-1 Epsilon Preview

Lin-1 Epsilon Preview is a 0.5B-parameter Rust code model from LindotAI, fine-tuned from Qwen2.5-Coder-0.5B-Instruct. It is small enough to run on a laptop CPU and is meant for quick completions, simple functions and generate-and-test loops. Its bigger sibling is Lin-1 Locus (4B).

Benchmarks

All numbers measured by us with the same harness for every model (Rust ports of the benchmarks; generated code is compiled and run against the tests).

Model HumanEval-Rust p@1 / p@10 MBPP-Rust p@1 / p@10 Holdout (Rust) LBPP-Rust
Lin-1 Epsilon Preview (0.5B) 48.4 / 65.2 41.3 / 54.2 30.3 4.1
Temper 2 Ion 48.1 / 57.6 45.5 / 53.8 31.0 4.5
Temper 1 0.5B 36.3 / 46.3 39.3 / 46.9 β€” β€”
ThoxEdge-RustCoder-0.5B 28.7 / 44.8 33.2 / 43.1 β€” β€”
Qwen2.5-Coder-0.5B-Instruct (base) 23.8 / β€” β€” β€” β€”
Qwen3.5-0.8B 10.1 / 22.1 14.5 / 25.5 β€” β€”

Holdout is our private set of Rust tasks with tests that was never used for training.

Quick start

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "LindotAI/Lin-1-Epsilon"
tok = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto")

messages = [{"role": "user", "content": "Write a Rust function that checks whether a string is a palindrome."}]
inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
out = model.generate(inputs, max_new_tokens=512, temperature=0.2, do_sample=True)
print(tok.decode(out[0][inputs.shape[1]:], skip_special_tokens=True))

GGUF (llama.cpp / Ollama): Lin-1-Epsilon-Q8_0.gguf (~0.5 GB) is included.

llama-cli -m Lin-1-Epsilon-Q8_0.gguf -p "Write a Rust function that reverses a Vec<i32> in place."

Training

Supervised fine-tuning on curated, compile-checked Rust, followed by reinforcement learning (GRPO) with a test-pass reward on Rust tasks, and a weight average of the SFT and RL checkpoints. All training data was decontaminated against the evaluation sets (12-gram overlap filter).

Limitations

  • Preview release: a 0.5B model makes more mistakes than larger ones on multi-step logic.
  • Focused on Rust. Always compile and test generated code.

License

Apache 2.0, same as the base model.