Instructions to use philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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("philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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 philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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 "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-MLX"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-MLX"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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 "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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 philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-MLX

Run Hermes

hermes

OpenClaw new

How to use philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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 "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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 "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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"

MLX LM

How to use philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-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 "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-MLX"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-MLX"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "philipjohnbasile/GLM-5.2-Demolition-q4a4-soul-MLX",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

philipjohnbasile commited on 15 days ago

Commit

c69d41c

verified ·

1 Parent(s): ef322ad

#71 batching verified: 2.6x standalone / 1.74x live-serve

Browse files

Files changed (1) hide show

SPEED.md +7 -3

SPEED.md CHANGED Viewed

@@ -22,9 +22,13 @@ Proven 4 independent ways, $0 spent (vs the $4–15K an EAGLE retrain would have
 ### What ACTUALLY delivers speed (and is already shipped)
 1. **Fused MoE dequant-matmul Metal kernel** (#33) — this *is* the 11–14 tok/s; it's the real win.
-2. **Throughput via batching** — batched best-of-N (#35) + continuous batching (#48). On a memory-bound MoE,
-   *parallel sequences* are the lever: one expert-load serves the whole batch, so tok/s **scales with batch**.
-   This is where "faster" actually lives for this model — not single-stream.
 ### Free, orthogonal wins (prefill / TTFT — not decode tok/s)
 - **Prompt/prefix KV cache** (`05_serve.sh --prompt-cache-size`) — agentic loops resend the same system+tools;

 ### What ACTUALLY delivers speed (and is already shipped)
 1. **Fused MoE dequant-matmul Metal kernel** (#33) — this *is* the 11–14 tok/s; it's the real win.
+2. **Throughput via batching — MEASURED 2.6× at B=8** (`scripts/91_batch_scaling.py`): total decode
+   **15.8 → 27.1 → 34.6 → 41.1 tok/s** at B = 1 / 2 / 4 / 8. On a memory-bound MoE, *parallel sequences* are
+   the lever — one expert-load serves the whole batch (per-seq drops, total climbs). This is where "faster"
+   actually lives (best-of-N, proof-search, multi-agent, the flywheel), and `mlx_lm.server` batches concurrent
+   requests **natively** (`is_batchable`, no draft model — that's us), so the win is available at the serve (#71).
+   **Serve-verified: 1.74× at B=6** on the live `mlx_lm.server` (concurrent vs sequential, `scripts/92_serve_batch_test.py`)
+   — already shipping; just fire concurrent requests. The ONLY measured speedup on this model that beats 1×.
 ### Free, orthogonal wins (prefill / TTFT — not decode tok/s)
 - **Prompt/prefix KV cache** (`05_serve.sh --prompt-cache-size`) — agentic loops resend the same system+tools;