Instructions to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX
# Run inference directly in the terminal:
llama cli -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX
# Run inference directly in the terminal:
llama cli -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX
# Run inference directly in the terminal:
./llama-cli -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX
# Run inference directly in the terminal:
./build/bin/llama-cli -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Use Docker

docker model run hf.co/xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

LM Studio
Jan

Ollama

How to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with Ollama:

ollama run hf.co/xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Unsloth Desktop

How to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with Docker Model Runner:
```
docker model run hf.co/xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX
```

Lemonade

How to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Run and chat with the model

lemonade run user.VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF-Q8_0_ROCMFPX

List all available models

lemonade list

Hermes Agent

How to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

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 xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX

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 "xiaoqiao64/VeriLoop-E2-Q8_0_ROCMFPX_AGENT-GGUF:Q8_0_ROCMFPX" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

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

VeriLoop-E2 Q8_0 ROCmFPX Agent GGUF

This is a ROCmFPX-optimized Q8_0 quantized version of VeriLoop-E2, specifically optimized for running local AI agents on laptops equipped with AMD Ryzen AI Max+ 395 and 128GB unified memory.

The model is intended for users who want to run VeriLoop-E2 locally on AMD's high-memory APU platform with improved inference performance through ROCmFPX.

Model Details

Base model: tsinghua-sigs-robot-lab/VeriLoop-E2
Quantization: Q8_0
Format: GGUF
Optimization: ROCmFPX
Target hardware: AMD Ryzen AI Max+ 395 laptops with 128GB unified memory
Primary use case: Local coding and general-purpose AI agents
Recommended runtime: ROCmFPX

Why ROCmFPX?

This model is specifically prepared for AMD Ryzen AI Max+ 395 128GB systems.

ROCmFPX provides optimized kernels and execution paths for AMD GPUs/APUs, allowing this model to run significantly more efficiently on supported AMD hardware than using a generic GGUF runtime.

In particular, the large unified memory available on AI Max+ 395 systems makes it possible to run relatively large models locally while keeping the model and inference workload in the same memory space.

Important: Standard llama.cpp Is Not Supported

This model cannot be loaded directly with standard llama.cpp.

The ROCmFPX-specific implementation is required:

https://github.com/ROCmFPX/ROCmFPX

Please clone and build ROCmFPX according to the instructions in the official repository before attempting to run this model.

The model uses ROCmFPX-specific optimizations and therefore should be treated as a ROCmFPX model rather than a generic llama.cpp GGUF model.

MTP Support

If you want to use the MTP (Multi-Token Prediction) head, use the standard Q8_0 MTP model provided by the original model repository:

MTP model:

https://huggingface.co/tsinghua-sigs-robot-lab/VeriLoop-E2-GGUF/blob/main/mtp-VeriLoop-E2-Q8_0.gguf

In other words:

Use this repository's ROCmFPX Q8_0 model as the main model for accelerated inference on AMD AI Max+ 395.
Use the standard Q8_0 MTP head from the original VeriLoop-E2 GGUF repository when MTP is required.

The MTP head does not need to be converted to the ROCmFPX format.

Recommended Hardware

This model is primarily intended for:

AMD Ryzen AI Max+ 395 + 128GB

Recommended configuration:

CPU: AMD Ryzen AI Max+ 395
GPU: Radeon 8060S
Unified Memory: 128GB
Runtime: ROCmFPX
Model: Q8_0 ROCmFPX GGUF

The large unified memory capacity is particularly useful for running VeriLoop-E2 together with long contexts and agent workloads.

Intended Use

This model is intended for local AI agent workloads such as:

Coding agents
Software engineering tasks
Repository exploration
Code generation and modification
Terminal-based agents
Local development assistants
Long-context agent workflows

It can be used as a local model backend for agent frameworks and coding assistants that support OpenAI-compatible or compatible local inference APIs.

Limitations

This model is not intended for standard llama.cpp.
ROCmFPX is required for the intended AMD-optimized inference path.
Performance depends on ROCm, ROCmFPX, driver versions, memory configuration, context length, and other runtime settings.
The model is specifically optimized for AMD AI Max+ 395-class systems and may not provide the same benefits on other hardware.