Skip to content
Language models

Emotion-S1

fast CPU-based emotion detector for dialogue

2 min readupdated

A compact model by davolu built to detect user emotions in chat conversations. Based on the ModernBERT-base encoder (~150M parameters), it handles three types of questions about the latest message in a single pass while factoring in up to 4 previous messages: picking an emotion from a custom list, rating its intensity on a 1–5 scale, or answering yes/no questions (such as whether the user needs support). Unlike heavy LLMs like Claude or GPT, the model is roughly 596 MB and runs locally on a CPU with zero API costs. It is a great fit for developers building English-language chatbots and routing pipelines where low latency and free inference matter.

Key features

  • Three specialized heads: classification from an arbitrary list of options (choice), intensity scoring from 1–5 (score), and binary questions with probability outputs (yes/no).
  • Runs on standard CPUs: on an 8-core Intel Xeon, median latency is 40.5 ms per question (and around 310 ms for a batch of four questions at once).
  • Confidence calibration: each head has its own temperature parameter stored in heads.pt for more accurate probability estimates.
  • Context-aware: processes the latest message along with up to 4 previous dialogue turns (up to 192 tokens).

What you can do

  • Fast customer support router: check in real time whether a user is frustrated and needs escalation to a human agent without burning tokens on large LLMs.
  • Dialogue log annotation: batch-process large archives of English chat logs locally on your own server for sentiment analytics.
  • Cascading LLM pipeline: use the model as a cheap first-pass filter on familiar dialogue types and route complex or edge cases to larger models.

Example

import sys
from huggingface_hub import snapshot_download

model_dir = snapshot_download("davolu/emotion-s1")
sys.path.insert(0, model_dir)
from emotion_s1_core import load_emotion_s1, ask

model = load_emotion_s1(model_dir, device="cpu")
out = ask(model, "I've asked three times and it's STILL broken.", {"frustrated": {"type": "noul", "question": "Is the user frustrated?"}})
Read moreDetails
On in-domain test data, Emotion-S1 outperforms Claude Sonnet 5.5 and GPT-6.1 Sol (choice accuracy of 0.70 vs. 0.625–0.64, and 0.795 on yes/no), though choice accuracy drops to 0.34 on out-of-domain datasets (MELD, EmoWOZ).
The checkpoint size is 596.1 MB (fp32) and requires no GPU; it loads via a custom emotion_s1_core.py script rather than the standard AutoModelForSequenceClassification.
Limitations: English only, a 192-token context window, and a CC BY-NC-SA 4.0 license (research and non-commercial use only).
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.6 GB · ready files1.2 GB
  • FP8 / INT8recommendedweights 0.3 GB0.9 GB
  • 4-bitweights 0.2 GB0.8 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — 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

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    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.

    # davolu/emotion-s1 — A30, FP8 / INT8
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve davolu/emotion-s1 --max-model-len 8192 --quantization fp8
    Add more LoRAs to emotion-s1
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

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.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

New posts about versions of ModernBERT-base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 11All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 4All on Hugging Face

About the model

davolu/emotion-s1
Author
davolu
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments