Instructions to use Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp with MLX:

# Make sure mlx-vlm is installed
# pip install --upgrade mlx-vlm

from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template
from mlx_vlm.utils import load_config

# Load the model
model, processor = load("Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp")
config = load_config("Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp")

# Prepare input
image = ["http://images.cocodataset.org/val2017/000000039769.jpg"]
prompt = "Describe this image."

# Apply chat template
formatted_prompt = apply_chat_template(
    processor, config, prompt, num_images=1
)

# Generate output
output = generate(model, processor, formatted_prompt, image)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp"

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent

How to use Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp"

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Yanun commited on 18 days ago

Commit

949fba8

verified ·

1 Parent(s): 36ff98b

Replace model description with architecture and precision details

Browse files

Files changed (1) hide show

README.md +24 -3

README.md CHANGED Viewed

@@ -15,11 +15,32 @@ tags:
 # Swift-Qwen3.8-27b-oQ4e-fp16-mtp
-Local oMLX oQ4e conversion of [UkisAI Swift-Qwen3.8-27b](https://huggingface.co/ukisai/Swift-Qwen3.8-27b), uploaded by Yanun.
-Converted with oMLX using the oQ4e preset and FP16 floating-point precision, with MTP retained. The `fp16` suffix does not mean that all model or MTP weights are unquantized. See `config.json` for quantization settings and `oq_imatrix_report.json` for calibration results.
-Use a compatible MLX runtime. In oMLX, enable Lightning MTP to use the retained MTP head. Performance depends on hardware, prompts, and runtime settings; no speedup is guaranteed.
 ## Source and license

 # Swift-Qwen3.8-27b-oQ4e-fp16-mtp
+## Model architecture
+A 27B-class dense multimodal model stored in MLX format. It contains a text backbone, a vision encoder, and one multi-token prediction (MTP) layer.
+| Component | Structure |
+|---|---|
+| Text backbone | 64 layers; hidden size 5,120; feed-forward size 17,408 |
+| Attention layout | 48 linear-attention layers and 16 full-attention layers, with full attention every fourth layer |
+| Full attention | 24 query heads, 4 key/value heads, head dimension 256 |
+| Vocabulary | 248,320 tokens |
+| Configured context limit | 262,144 tokens; usable length depends on runtime settings and available memory |
+| Vision encoder | 27 layers; hidden size 1,152; 16 attention heads; 16 × 16 image patches |
+| Vision-to-text connection | Vision features are projected to the text hidden size of 5,120 |
+| MTP | One additional prediction layer with attention and feed-forward projections |
+## Weight precision
+The oQ4e checkpoint uses mixed precision rather than uniform 4-bit weights:
+- The default quantization is **4-bit affine**, with **64 values per group**.
+- **187 modules** have explicit **5-bit** overrides in `config.json`.
+- The MTP layer's seven large attention and feed-forward matrices use **4-bit** weights.
+- The MTP fusion matrix (`mtp.fc`) and normalization weights remain **FP16**. The fusion matrix maps 10,240 input features to 5,120 output features.
+- The MTP quantization scales and offsets are stored in **FP16**.
+The `fp16` suffix describes the retained floating-point precision; it does not mean the entire model or MTP layer is FP16. Exact per-module settings are recorded in `config.json`.
 ## Source and license