Skip to content
Computer vision

Gemma 4 E4B IT QAT Q4_0

unquantized-чекпоинт для локальных пайплайнов

1 min read

QAT-чекпоинт Gemma 4 E4B IT в формате unquantized Q4_0, выложенный heromann. Это half-precision веса, извлечённые из QAT-пайплайна Google DeepMind; база — google/gemma-4-E4B-it. Подходит для кастомной downstream-компиляции и исследований.

Ключевые особенности

  • Unquantized QAT Q4_0: half-precision веса из QAT-пайплайна, а не готовый GGUF или мобильный формат.
  • E4B: 4.5B эффективных параметров (~8B с embeddings), вход — текст, изображения и аудио, выход — текст.
  • Контекст 128K токенов, режим thinking, нативный system prompt и function calling.
  • Мультиязычность: предобучение на 140+ языках, out-of-the-box поддержка 35+ языков.
  • Для speculative decoding ассистент должен быть QAT-чекпоинтом той же точности.

Пример

from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "heromann/gemma-4-E4B-it-qat-q4_0-unquantized"
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, dtype="auto", device_map="auto"
)
ПодробнееMMLU Pro у E4B — 69.4%, AIME 2026 без инструментов — 42.5%, LiveCodeBench v6 — 52.0%.
Контекст 128K; E4B: 4.5B эффективных параметров (~8B с embeddings).
Это unquantized QAT-чекпоинт, не GGUF и не мобильный wNa8o8.
Лицензия Apache 2.0.
Discuss Telegram 0 23

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

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.

    # heromann/gemma-4-E4B-it-qat-q4_0-unquantized
    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 heromann/gemma-4-E4B-it-qat-q4_0-unquantized --local-dir ~/models/gemma-4-E4B-it-qat-q4_0-unquantized
    # 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 gemma-4-E4B-it-qat-q4_0-unquantized (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

heromann/gemma-4-E4B-it-qat-q4_0-unquantized
Author
heromann
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

230

Ask AI about this model

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

Comments