Instructions to use inferencerlabs/Ling-2.6-flash-MLX-9bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use inferencerlabs/Ling-2.6-flash-MLX-9bit 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("inferencerlabs/Ling-2.6-flash-MLX-9bit")

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
LM Studio

Pi new

How to use inferencerlabs/Ling-2.6-flash-MLX-9bit with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "inferencerlabs/Ling-2.6-flash-MLX-9bit"

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": "inferencerlabs/Ling-2.6-flash-MLX-9bit"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use inferencerlabs/Ling-2.6-flash-MLX-9bit 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 "inferencerlabs/Ling-2.6-flash-MLX-9bit"

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 inferencerlabs/Ling-2.6-flash-MLX-9bit

Run Hermes

hermes

MLX LM

How to use inferencerlabs/Ling-2.6-flash-MLX-9bit with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "inferencerlabs/Ling-2.6-flash-MLX-9bit"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "inferencerlabs/Ling-2.6-flash-MLX-9bit"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "inferencerlabs/Ling-2.6-flash-MLX-9bit",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

good fit for M2 Max 96gb - it's 104B-A7B4

by ljupco - opened 11 days ago

Discussion

ljupco

11 days ago

omlx
that loaded the model, and run benchmarks
Benchmark Model: Ling-2.6-flash-mlx-4bit-DWQ

Single Request Results

Test TTFT(ms) TPOT(ms) pp TPS tg TPS E2E(s) Throughput Peak Mem
pp1024/tg128 3513.2 27.31 291.5 tok/s 36.9 tok/s 6.982 165.0 tok/s 55.33 GB
pp4096/tg128 16395.7 25.69 249.8 tok/s 39.2 tok/s 19.659 214.9 tok/s 55.99 GB
pp8192/tg128 32562.9 27.07 251.6 tok/s 37.2 tok/s 36.001 231.1 tok/s 56.57 GB
pp16384/tg128 66354.9 27.73 246.9 tok/s 36.3 tok/s 69.876 236.3 tok/s 57.75 GB
pp32768/tg128 136722.8 28.98 239.7 tok/s 34.8 tok/s 140.403 234.3 tok/s 60.10 GB
pp65536/tg128 296867.5 32.72 220.8 tok/s 30.8 tok/s 301.022 218.1 tok/s 64.80 GB

Continuous Batching
pp1024 / tg128

Batch tg TPS Speedup pp TPS pp TPS/req TTFT(ms) E2E(s)
1x 36.9 tok/s 1.00x 291.5 tok/s 291.5 tok/s 3513.2 6.982
2x 46.1 tok/s 1.25x 212.6 tok/s 106.3 tok/s 9536.4 15.186
4x 51.4 tok/s 1.39x 213.1 tok/s 53.3 tok/s 18897.6 29.179

M2 Max (38c) Ling-2.6-flash4bit-DWQ · 4bit 26-04-30 18:16:04
1k PP 291.5 · TG 36.9 tok/s
4k PP 249.8 · TG 39.2 tok/s
8k PP 251.6 · TG 37.2 tok/s
16k PP 246.9 · TG 36.3 tok/s
32k PP 239.7 · TG 34.8 tok/s
64k PP 220.8 · TG 30.8 tok/s

ljupco changed discussion title from perfect fit for M2 Max 104B-A7B4 to good fit for M2 Max 96gb - it's 104B-A7B4 11 days ago

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment

good fit for M2 Max 96gb - it's 104B-A7B4

Single Request Results

Continuous Batchingpp1024 / tg128

Batch tg TPS Speedup pp TPS pp TPS/req TTFT(ms) E2E(s)1x 36.9 tok/s 1.00x 291.5 tok/s 291.5 tok/s 3513.2 6.9822x 46.1 tok/s 1.25x 212.6 tok/s 106.3 tok/s 9536.4 15.1864x 51.4 tok/s 1.39x 213.1 tok/s 53.3 tok/s 18897.6 29.179

Continuous Batching
pp1024 / tg128

Batch tg TPS Speedup pp TPS pp TPS/req TTFT(ms) E2E(s)
1x 36.9 tok/s 1.00x 291.5 tok/s 291.5 tok/s 3513.2 6.982
2x 46.1 tok/s 1.25x 212.6 tok/s 106.3 tok/s 9536.4 15.186
4x 51.4 tok/s 1.39x 213.1 tok/s 53.3 tok/s 18897.6 29.179