Skip to content
Other

rooster-coder-rwkv7-1.5b

coding adapter for the RWKV7 architecture

2 min readupdated

A new LoRA (a small add-on on top of a base model) adapter and merged versions for RWKV/RWKV7-G1k-1.5B-20260930, trained on a coder profile from the Qwen4 Max reasoning corpus. The model brings a step-by-step think-style reasoning approach to the RWKV architecture. The release includes the original adapter weights as well as converted GGUF and MLX formats for running locally.

Key features

  • Base RWKV7 architecture (1.5B parameters) with seamless execution support.
  • Trained on structured reasoning traces (think-tag format).
  • Full support for running via peft or ready-to-use quantized GGUF and MLX builds.
  • Optimized for fp32 during training for stability on a Tesla T4 GPU.

What you can do

  • Generate code and answer technical questions with logical reasoning.
  • Run locally on a laptop via llama.cpp in Q4_K_M or Q6_K GGUF formats.
  • Use the MLX version for fast performance on Apple Silicon devices.

Example

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

repo = "AJ-Gonzalez/rooster-coder-rwkv7-1.5b"
base = AutoModelForCausalLM.from_pretrained(
    "RWKV/RWKV7-G1k-1.5B-20260930", trust_remote_code=True, dtype=torch.float32)
model = PeftModel.from_pretrained(base, repo, trust_remote_code=True)
tok = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)

text = tok.apply_chat_template(
    [{"role": "user", "content": "Write a Python LRU cache."}],
    tokenize=False, add_generation_prompt=True, thinking=True)
ich = torch.tensor([tok.encode(text)])
out = model.generate(ids, max_new_tokens=512, do_sample=False)
print(tok.decode(out[0][ids.shape[1]:]))
Read moreDetails
Benchmarks and specs: train loss 1.6735, eval loss 1.9399.
VRAM requirements: peak usage during training reached 11.43 GiB on a Tesla T4.
Limitations: trained on a small dataset of 70 rows, so the model picks up coding and reasoning style rather than acting as a full-fledged general-purpose assistant.
Available artifacts: GGUF (Q4_K_M, Q6_K) and MLX 4-bit for local inference.
Discuss Telegram 3 91

Will it run?

Runs in GGUF Q6_K

The most precise of the ready quantised files fits.

  • GGUF Q6_Kweights 1.2 GB · ready files1.9 GB
  • GGUF Q4_K_Mweights 0.9 GB · ready files1.5 GB

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: GGUF Q6_K — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # AJ-Gonzalez/rooster-coder-rwkv7-1.5b
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download AJ-Gonzalez/rooster-coder-rwkv7-1.5b --local-dir ~/models/rooster-coder-rwkv7-1.5b
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of RWKV7-G1k-1.5B-20260930 (quants, finetunes, LoRAs) will show up under the bell.

LoRAs and adapters 1All on Hugging Face

About the model

AJ-Gonzalez/rooster-coder-rwkv7-1.5b
Author
AJ-Gonzalez
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

913

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments