Instructions to use HamoAI/hamo-score-0.6b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use HamoAI/hamo-score-0.6b 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("HamoAI/hamo-score-0.6b")

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

llama.cpp

How to use HamoAI/hamo-score-0.6b 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 HamoAI/hamo-score-0.6b
# Run inference directly in the terminal:
llama cli -hf HamoAI/hamo-score-0.6b

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf HamoAI/hamo-score-0.6b
# Run inference directly in the terminal:
llama cli -hf HamoAI/hamo-score-0.6b

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 HamoAI/hamo-score-0.6b
# Run inference directly in the terminal:
./llama-cli -hf HamoAI/hamo-score-0.6b

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 HamoAI/hamo-score-0.6b
# Run inference directly in the terminal:
./build/bin/llama-cli -hf HamoAI/hamo-score-0.6b

Use Docker

docker model run hf.co/HamoAI/hamo-score-0.6b

LM Studio
Jan

vLLM

How to use HamoAI/hamo-score-0.6b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "HamoAI/hamo-score-0.6b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "HamoAI/hamo-score-0.6b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/HamoAI/hamo-score-0.6b

Ollama
How to use HamoAI/hamo-score-0.6b with Ollama:
```
ollama run hf.co/HamoAI/hamo-score-0.6b
```
Unsloth Desktop

How to use HamoAI/hamo-score-0.6b with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "HamoAI/hamo-score-0.6b"

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": "HamoAI/hamo-score-0.6b"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use HamoAI/hamo-score-0.6b with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "HamoAI/hamo-score-0.6b"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "HamoAI/hamo-score-0.6b"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "HamoAI/hamo-score-0.6b",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Docker Model Runner
How to use HamoAI/hamo-score-0.6b with Docker Model Runner:
```
docker model run hf.co/HamoAI/hamo-score-0.6b
```

Lemonade

How to use HamoAI/hamo-score-0.6b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull HamoAI/hamo-score-0.6b

Run and chat with the model

lemonade run user.hamo-score-0.6b-{{QUANT_TAG}}

List all available models

lemonade list

Hermes Agent

How to use HamoAI/hamo-score-0.6b 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 "HamoAI/hamo-score-0.6b"

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 HamoAI/hamo-score-0.6b

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use HamoAI/hamo-score-0.6b with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "HamoAI/hamo-score-0.6b"

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 "HamoAI/hamo-score-0.6b" \
  --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"

Start here 从这里开始 — toolkit · one-command server · self-check exam · fine-tuning guide

pinned

by chrischengzh - opened Aug 17

Discussion

chrischengzh

Hamo AI Technology Ltd. org Aug 17

EN below ⬇ / 中文在前

下载了权重？这里是从「拿到权重」到「安全跑通并自检」的最短路径——模型的「另一半」（提示词格式、解析、平滑折算、危机闸门）已全部开源：

1. 装工具包（Apache-2.0）

pip install hamo-score

四行代码跑通闸门 → 评分 → 平滑 → 状态桶完整管线（危机命中的消息永远不会抵达模型）。

2. 或者一条命令起服务器

git clone https://github.com/HamoAI/hamo-score-toolkit.git && cd hamo-score-toolkit/server
docker compose up   # 自动拉 GGUF、建模、预热 → POST /score

3. 考一考你的部署

python eval/run_exam.py

195 题合成考卷（零真实数据）+ 10 条危机闸门用例。官方参考带：JSON 合法率 100% · 维度级 84.0% · 闸门 10/10。大幅偏离几乎总是接线问题，不是模型问题——考卷的 README 里有按序排查清单。

4. 深入：集成指南（正确接线 + 十条禁令） · FAQ · 微调指南（五代打法，含两代拒收） · 发布文：中文 / EN

对某个评分不服？ 提交工具包仓库的「评分分歧」issue 模板——分歧报告直接喂给引导后续版本的人类金标计划。

Downloaded the weights? Here is the shortest path from "weights on disk" to "safely deployed and verified" — the model's other half (prompt format, parsing, smoothing math, crisis gate) is now fully open-source:

1. Install the toolkit (Apache-2.0)

pip install hamo-score

A few lines of code run the full gate → score → smooth → bucket pipeline (crisis-gated messages never reach the model).

2. Or run a server in one command

git clone https://github.com/HamoAI/hamo-score-toolkit.git && cd hamo-score-toolkit/server
docker compose up   # fetches the GGUF, creates + warms the model → POST /score

3. Give your deployment the exam

python eval/run_exam.py

195 synthetic questions (zero real data) + 10 handwritten crisis-gate cases. Official reference band: JSON validity 100% · dimension-level 84.0% · gate 10/10. Big deviations are almost always wiring, not the model — the eval README has an ordered triage.

4. Go deeper: integration guide (the correct wiring + the don't list) · FAQ · fine-tuning playbook (five generations, incl. the two we rejected) · release notes: EN / 中文

Disagree with a score? File the score-disagreement issue template — reports feed the human gold-label program that steers future versions.

⚠️ Reminder: the model is a measuring instrument — not a chatbot, not a diagnostic tool, and not a crisis detector. Keep the deterministic gate upstream (license §3c). The toolkit's default pipeline does this for you.

chrischengzh pinned discussion Aug 17

chrischengzh

Hamo AI Technology Ltd. org Aug 24

Update — the one thing we're asking for 我们唯一请求的一件事

The weights have been downloaded a few thousand times this month. What we have received back so far: nothing. That's fine — but there is one contribution that would genuinely change the next version.

If you think a score is wrong, tell us. Open a disagreement report →

Three fields is all it takes: the message (redact it freely — we do not want identifiable text), the score the model gave, and the score you would give. Your population and language help but are optional.

Why this matters more than a bug report: the frontier scorer this model replaced disagrees with itself 2–6% of the time on repeat runs. That is the noise floor. Wherever a practitioner disagrees with the model above that floor, we are looking at a real training-data gap — and those gaps are exactly what steers the next generation. Silent workarounds keep the model wrong; reports fix it.

中文：权重这个月被下载了几千次，回流到我们这里的反馈是零。这没关系，但有一件事真的会改变下一代模型——如果你觉得某个评分不对，请告诉我们（提交分歧报告）。只要三样：消息（请自由脱敏，我们不需要可识别的文本）、模型给的分、你认为该给的分。

为什么这比 bug 报告更要紧：被这个模型替换掉的那个前沿评分器，自己重打同一句话都有 2–6% 的不一致——那是噪声底噪。凡是从业者与模型的分歧高过这条底噪，那里就有一个真实的训练数据缺口，而这些缺口正是引导下一代的东西。绕过去，模型就一直错着；报上来，它才会变好。

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