delimitter/synoema-coder-3b-v2

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base = "Qwen/Qwen2.5-Coder-3B-Instruct"
adapter = "delimitter/synoema-coder-3b-v2"

tok = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, torch_dtype=torch.float16, device_map="auto")
model = PeftModel.from_pretrained(model, adapter)
model = model.merge_and_unload()

SYS = """# Synoema — BPE-aligned functional language
Entry point: `main = <expr>`. Functions: `name args = body`.
Pattern matching via multiple equations. Conditional: `? cond -> then : else`.
Lists: `[1 2 3]` (space-sep). No return, no print, no def/fn keywords.
Write only valid Synoema code."""

prompt = tok.apply_chat_template([
    {"role": "system", "content": SYS},
    {"role": "user",   "content": "Fibonacci of 10"},
], tokenize=False, add_generation_prompt=True)

inp = tok(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inp, max_new_tokens=256, do_sample=True, temperature=0.7, top_p=0.9)
print(tok.decode(out[0][inp["input_ids"].shape[1]:], skip_special_tokens=True))

-- Fibonacci
fib 0 = 0
fib 1 = 1
fib n = fib (n - 1) + fib (n - 2)
main = fib 10

-- Quicksort
qsort [] = []
qsort [x|xs] = qsort (filter (\ y -> y < x) xs) ++ [x] ++ qsort (filter (\ y -> y >= x) xs)
main = qsort [3 1 4 1 5 9 2 6]