Instructions to use dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx with MLX:

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

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx"

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": "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx 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 "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx"

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 dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx"

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 "dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx" \
  --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"

dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx

MLX conversion of a GGUF model (converted by llama.cpp from XHToken/Spark-X2.5-1.7B-INT8), quantized with a ridge_4_6_8 recipe.

Ridge policy ridge_4_6_8 (Spark-X2.5): 4-bit middle FFN (mlp.gate/up/down_proj), 6-bit sliding-layer attn_qkv (self_attn.q_k_v_proj), out_proj, and edge FFN layers, and 8-bit protection for the sensitive control path: the per-head sigmoid gate attn_gate (self_attn.g_proj) and the fused attn_qkv of every full_attention layer (the long-context path with 2-wide, 256-dim KV). The tied embedding (model.embedding, doubling as the output unembedding) and lm_head are 6-bit; norms stay fp16.

Supported runtime: mlx_lm (upgraded to support the model type).

True parameter count: 1.71B. (HF's automatic 'Model size' badge apparently undercounts MLX quantized models because the packed weights are stored at <=8 bits per element.)

Usage

from mlx_lm import load, generate
model, tokenizer = load("dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx")
print(generate(model, tokenizer, prompt="hello"))

Example generation

mlx_lm.generate --model dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx --prompt "What is the square root of 2?" --max-tokens 400
==========
The user asked "What is the square root of 2?" I need to provide the value of the square root of 2.

The square root of 2 is an irrational number, approximately 1.41421356237...

I should provide the exact value and the approximate decimal value.

Let me structure the response:
1. State the exact value: √2
2. Provide the approximate decimal value: approximately 1.4142
3. Maybe mention it's irrational.

Let me make sure the answer is clear and accurate.</think>The square root of 2 is an irrational number, which means it cannot be expressed as a simple fraction.

- **Exact value:** √2
- **Approximate value:** ≈ 1.41421356237...

So, the square root of 2 is approximately **1.4142**.
==========
Prompt: 25 tokens, 215.179 tokens-per-sec
Generation: 208 tokens, 41.483 tokens-per-sec
Peak memory: 1.253 GB
(M1 MacBook Air)

Downloads last month: 14

Safetensors

Model size

2B params

Tensor type

U32

F16

MLX

Hardware compatibility

6-bit

Model tree for dakerholdings/Spark-X2.5-1.7B-ridge_4_6_8-mlx

Base model

XHToken/Spark-X2.5-1.7B-Base

Finetuned

XHToken/Spark-X2.5-1.7B

Quantized

(9)

this model