Instructions to use Popbones/LiveMem-RL-mlx-bf16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Popbones/LiveMem-RL-mlx-bf16 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("Popbones/LiveMem-RL-mlx-bf16")

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 Popbones/LiveMem-RL-mlx-bf16 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "Popbones/LiveMem-RL-mlx-bf16"

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": "Popbones/LiveMem-RL-mlx-bf16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use Popbones/LiveMem-RL-mlx-bf16 with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "Popbones/LiveMem-RL-mlx-bf16"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "Popbones/LiveMem-RL-mlx-bf16"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "Popbones/LiveMem-RL-mlx-bf16",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use Popbones/LiveMem-RL-mlx-bf16 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 "Popbones/LiveMem-RL-mlx-bf16"

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 Popbones/LiveMem-RL-mlx-bf16

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Popbones/LiveMem-RL-mlx-bf16 with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "Popbones/LiveMem-RL-mlx-bf16"

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 "Popbones/LiveMem-RL-mlx-bf16" \
  --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"

LiveMem-RL-mlx-bf16 / README.md

Popbones

Publish experimental BF16 LiveMem MLX checkpoint

1f8cccc verified 8 days ago

preview code

Raw History Blame Contribute Delete

4.27 kB

---
{
  "license": "apache-2.0",
  "library_name": "livemem-mlx",
  "pipeline_tag": "text-generation",
  "base_model": "chen-l/LiveMem-RL",
  "tags": [
    "mlx",
    "livemem",
    "qwen3",
    "experimental"
  ],
  "inference": false
}
---
# LiveMem-RL MLX — BF16

This is an **unchanged-weight BF16 mirror** of the official `chen-l/LiveMem-RL` checkpoint at the pinned revision below, packaged for use with the standalone MLX runtime. No training or quantization was performed. The original authors provide the model; this repository does not claim authorship of it.

Experimental inference packaging. This mirror does not establish new model capabilities or improve the upstream checkpoint.

## Source and identity

- Original model: [chen-l/LiveMem-RL](https://huggingface.co/chen-l/LiveMem-RL).
- Pinned original revision: `3402dcace8766f9f4e0c8f80928c1795cc073f39`.
- Original BF16 weight SHA-256: `1a431577034b0ce43535d20fde78df765a19b992ad0130cdd7826e63fe5f13f0`.
- Base-model lineage: [Qwen3-4B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507).
- Runtime: [LiveMem MLX v0.1.0](https://github.com/popbones/livemem-mlx/releases/tag/v0.1.0).
- This repository weight SHA-256: `1a431577034b0ce43535d20fde78df765a19b992ad0130cdd7826e63fe5f13f0`.

## Precision

All original checkpoint weights remain BF16. The runtime uses FP32 recurrent matrices. Attention KV and convolution state are separate.

The underlying model has approximately 6.35 billion parameters, including the memory branch. The weight file is 12.70 GB. Weight storage is not total runtime memory.

## Run locally on Apple Silicon

Use Python 3.11 or later and an Apple Silicon Mac with Metal support.

```sh
python -m pip install "git+https://github.com/popbones/livemem-mlx.git@v0.1.0"
livemem-mlx generate --model Popbones/LiveMem-RL-mlx-bf16 --prompt "Hello" --max-tokens 64
```

```python
from livemem_mlx import load, generate
loaded = load("Popbones/LiveMem-RL-mlx-bf16")
result = generate(loaded, [{"role": "user", "content": "Hello"}], max_tokens=64)
print(result.text)
```

After download, pin the repository commit with `--revision` or `load(..., revision=...)` for reproducibility. Default loading follows this repository main branch. Local paths, offline loading and `--cache-dir` are also supported.

These files require `livemem_mlx.load`; this custom architecture is not registered as a built-in Transformers AutoModel or mlx-lm model. No remote Python implementation is included or executed. Hub hosting does not provide a hosted inference endpoint.

## Validation and limits

The corresponding checkpoint passed four standalone compatibility cases against its pre-extraction MLX runtime: Aurora, Zephyr, fresh-session Aurora replay, and a 2,283-token prompt with physical KV eviction followed by a simple arithmetic query. Generated token IDs, retained KV positions and saved prefill logits matched exactly within this precision variant.

Peak MLX allocation was 12.41 GiB on an Apple M1 Pro / 32 GiB for that 2,048-token KV-budget test. It is not an 8K/32K serving benchmark, total system RAM guarantee or broad quality evaluation. Full-model CUDA/vLLM parity, general long-term-memory quality and long-running stability are not established.

See the [compatibility report](https://github.com/popbones/livemem-mlx/blob/v0.1.0/docs/checkpoint-validation.md). This publication does not add a new benchmark result. Each fresh generation starts with separate state; state is not automatically retained across independent requests.

## License and attribution

The model distribution retains the upstream Apache-2.0 license. See LICENSE and NOTICE. The weights are unchanged. Source, precision settings, validated runtime and file hashes are recorded in manifest.json. Transformer auto_map metadata was removed to avoid advertising absent remote Python files. The original model and tokenizer assets remain credited to their authors.
LiveMem: *Maintaining Memory State Continuity in Long-Running LLM Inference* — Zhichen Liu, Ruihan Sun, Hengjie Yang, Zipeng Wu, Zhaohan Chen, Xiaofan Zhang and Yang Xu (2026), [arXiv:2608.02515](https://arxiv.org/abs/2608.02515).
This independent mirror/conversion is not endorsed by the LiveMem authors, Qwen or Alibaba.

	---
	{
	"license": "apache-2.0",
	"library_name": "livemem-mlx",
	"pipeline_tag": "text-generation",
	"base_model": "chen-l/LiveMem-RL",
	"tags": [
	"mlx",
	"livemem",
	"qwen3",
	"experimental"
	],
	"inference": false
	}
	---

	# LiveMem-RL MLX — BF16

	This is an unchanged-weight BF16 mirror of the official `chen-l/LiveMem-RL` checkpoint at the pinned revision below, packaged for use with the standalone MLX runtime. No training or quantization was performed. The original authors provide the model; this repository does not claim authorship of it.

	Experimental inference packaging. This mirror does not establish new model capabilities or improve the upstream checkpoint.

	## Source and identity

	- Original model: [chen-l/LiveMem-RL](https://huggingface.co/chen-l/LiveMem-RL).
	- Pinned original revision: `3402dcace8766f9f4e0c8f80928c1795cc073f39`.
	- Original BF16 weight SHA-256: `1a431577034b0ce43535d20fde78df765a19b992ad0130cdd7826e63fe5f13f0`.
	- Base-model lineage: [Qwen3-4B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507).
	- Runtime: [LiveMem MLX v0.1.0](https://github.com/popbones/livemem-mlx/releases/tag/v0.1.0).
	- This repository weight SHA-256: `1a431577034b0ce43535d20fde78df765a19b992ad0130cdd7826e63fe5f13f0`.

	## Precision

	All original checkpoint weights remain BF16. The runtime uses FP32 recurrent matrices. Attention KV and convolution state are separate.

	The underlying model has approximately 6.35 billion parameters, including the memory branch. The weight file is 12.70 GB. Weight storage is not total runtime memory.

	## Run locally on Apple Silicon

	Use Python 3.11 or later and an Apple Silicon Mac with Metal support.

	```sh
	python -m pip install "git+https://github.com/popbones/livemem-mlx.git@v0.1.0"
	livemem-mlx generate --model Popbones/LiveMem-RL-mlx-bf16 --prompt "Hello" --max-tokens 64
	```

	```python
	from livemem_mlx import load, generate

	loaded = load("Popbones/LiveMem-RL-mlx-bf16")
	result = generate(loaded, [{"role": "user", "content": "Hello"}], max_tokens=64)
	print(result.text)
	```

	After download, pin the repository commit with `--revision` or `load(..., revision=...)` for reproducibility. Default loading follows this repository main branch. Local paths, offline loading and `--cache-dir` are also supported.

	These files require `livemem_mlx.load`; this custom architecture is not registered as a built-in Transformers AutoModel or mlx-lm model. No remote Python implementation is included or executed. Hub hosting does not provide a hosted inference endpoint.

	## Validation and limits

	The corresponding checkpoint passed four standalone compatibility cases against its pre-extraction MLX runtime: Aurora, Zephyr, fresh-session Aurora replay, and a 2,283-token prompt with physical KV eviction followed by a simple arithmetic query. Generated token IDs, retained KV positions and saved prefill logits matched exactly within this precision variant.

	Peak MLX allocation was 12.41 GiB on an Apple M1 Pro / 32 GiB for that 2,048-token KV-budget test. It is not an 8K/32K serving benchmark, total system RAM guarantee or broad quality evaluation. Full-model CUDA/vLLM parity, general long-term-memory quality and long-running stability are not established.

	See the [compatibility report](https://github.com/popbones/livemem-mlx/blob/v0.1.0/docs/checkpoint-validation.md). This publication does not add a new benchmark result. Each fresh generation starts with separate state; state is not automatically retained across independent requests.

	## License and attribution

	The model distribution retains the upstream Apache-2.0 license. See LICENSE and NOTICE. The weights are unchanged. Source, precision settings, validated runtime and file hashes are recorded in manifest.json. Transformer auto_map metadata was removed to avoid advertising absent remote Python files. The original model and tokenizer assets remain credited to their authors.

	LiveMem: Maintaining Memory State Continuity in Long-Running LLM Inference — Zhichen Liu, Ruihan Sun, Hengjie Yang, Zipeng Wu, Zhaohan Chen, Xiaofan Zhang and Yang Xu (2026), [arXiv:2608.02515](https://arxiv.org/abs/2608.02515).

	This independent mirror/conversion is not endorsed by the LiveMem authors, Qwen or Alibaba.