Skip to content
Other

cradle-dialogue-qwen3.8-27b-2epoch

mental health crisis detector for dialogues

2 min readupdated

A full fine-tune (updating all model weights) of the 27.36B-parameter Qwen/Qwen3.8-27B language model by SungJoo. Built for turn-by-turn analysis of multi-turn English conversations, it detects whether the user's latest message shows signs of a mental health crisis. Unlike the base Qwen3.8-27B, this checkpoint is trained to output strictly a single line of structured alert and confirmation labels with no extra text, taking both the conversation history and its own prior predictions into account. Useful for clinical NLP researchers and developers testing safety and moderation algorithms for conversational systems.

Key features

  • Turn-by-turn classification: analyzes the full dialogue history and previous predictions, assigning labels exclusively to the user's latest message.
  • Strict output format: returns a single line — either none or one or more semicolon-separated labels (for example, alert_ongoing, confirm_SI_passive_ongoing, confirm_DV_ongoing; confirm_RA_past).
  • Trained on synthetic data: fine-tuned on the CRADLE-Dialogue dataset, consisting of 3,058 dialogues (48,557 turns) generated from Reddit posts using GPT-5.
  • Reasoning block support: uses the Qwen3.8 chat template with the tag followed by the final classification label.

What you can do

  • Research conversational AI safety: use the checkpoint in experiments on early crisis detection across multi-turn chats.
  • Compare base model generations: benchmark classification performance against the sibling cradle-dialogue-qwen3-32b-2epoch model trained on the same data and hyperparameters.
  • Automate dialogue labeling: integrate it into transformers-based research pipelines for turn-by-turn annotation of text sessions.

Example

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "SungJoo/cradle-dialogue-qwen3.8-27b-2epoch"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True)
# После генерации с chat_template отбрасываем блок размышлений:
label = response.split("")[-1].strip()
Read moreDetails
Epoch 2 checkpoint (step 3208 of 3 epochs) with a maximum context length of 4096 tokens.
Trained in bfloat16 precision on two NVIDIA H200 GPUs with a 5e-6 learning rate and an effective batch size of 16.
Intended strictly for clinical NLP research and does not replace professional help.
Discuss Telegram 0 25

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 51 GB · ready files57 GB
  • FP8 / INT8recommendedweights 25 GB29 GB
  • 4-bitweights 14 GB16 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.

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.

    # SungJoo/cradle-dialogue-qwen3.8-27b-2epoch
    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 SungJoo/cradle-dialogue-qwen3.8-27b-2epoch --local-dir ~/models/cradle-dialogue-qwen3.8-27b-2epoch
    # 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 Qwen3.8-27B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

SungJoo/cradle-dialogue-qwen3.8-27b-2epoch
Author
SungJoo
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

250

Ask AI about this model

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

Comments