Text Generation
Transformers
Safetensors
GGUF
English
qwen2
fine-tuned
identity
ollama
layer-expansion
custom-architecture
conversational
text-generation-inference
Instructions to use QuantAILabs/Quant-1-2B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use QuantAILabs/Quant-1-2B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="QuantAILabs/Quant-1-2B") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("QuantAILabs/Quant-1-2B") model = AutoModelForCausalLM.from_pretrained("QuantAILabs/Quant-1-2B", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use QuantAILabs/Quant-1-2B 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 QuantAILabs/Quant-1-2B # Run inference directly in the terminal: llama cli -hf QuantAILabs/Quant-1-2B
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf QuantAILabs/Quant-1-2B # Run inference directly in the terminal: llama cli -hf QuantAILabs/Quant-1-2B
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf QuantAILabs/Quant-1-2B # Run inference directly in the terminal: ./build/bin/llama cli -hf QuantAILabs/Quant-1-2B
Use Docker
docker model run hf.co/QuantAILabs/Quant-1-2B
- Llama new
- LM Studio
- Jan
- vLLM
How to use QuantAILabs/Quant-1-2B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "QuantAILabs/Quant-1-2B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "QuantAILabs/Quant-1-2B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/QuantAILabs/Quant-1-2B
- SGLang
How to use QuantAILabs/Quant-1-2B with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "QuantAILabs/Quant-1-2B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "QuantAILabs/Quant-1-2B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "QuantAILabs/Quant-1-2B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "QuantAILabs/Quant-1-2B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Ollama
How to use QuantAILabs/Quant-1-2B with Ollama:
ollama run hf.co/QuantAILabs/Quant-1-2B
- Unsloth Desktop
- Pi
How to use QuantAILabs/Quant-1-2B with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf QuantAILabs/Quant-1-2B
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:9931/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "QuantAILabs/Quant-1-2B" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use QuantAILabs/Quant-1-2B with Docker Model Runner:
docker model run hf.co/QuantAILabs/Quant-1-2B
- Lemonade
How to use QuantAILabs/Quant-1-2B with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull QuantAILabs/Quant-1-2B
Run and chat with the model
lemonade run user.Quant-1-2B-{{QUANT_TAG}}List all available models
lemonade list
- Hermes Agent
How to use QuantAILabs/Quant-1-2B with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf QuantAILabs/Quant-1-2B
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://127.0.0.1:9931/v1 hermes config set model.default QuantAILabs/Quant-1-2B
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use QuantAILabs/Quant-1-2B with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf QuantAILabs/Quant-1-2B
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://127.0.0.1:9931/v1 \ --custom-model-id "QuantAILabs/Quant-1-2B" \ --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"
Upload README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,149 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
base_model: Qwen/Qwen2.5-1.5B-Instruct
|
| 6 |
+
tags:
|
| 7 |
+
- qwen2
|
| 8 |
+
- fine-tuned
|
| 9 |
+
- identity
|
| 10 |
+
- ollama
|
| 11 |
+
- gguf
|
| 12 |
+
- layer-expansion
|
| 13 |
+
- custom-architecture
|
| 14 |
+
library_name: transformers
|
| 15 |
+
pipeline_tag: text-generation
|
| 16 |
---
|
| 17 |
+
|
| 18 |
+
# Quant-1-2B
|
| 19 |
+
|
| 20 |
+

|
| 21 |
+
|
| 22 |
+
The expanded version of Quant-1 with custom architecture modifications. Built by OpenMind Labs.
|
| 23 |
+
|
| 24 |
+
## What is this?
|
| 25 |
+
|
| 26 |
+
This is Quant-1-2B - an expanded version of our base 1.5B model. We didn't just fine-tune it, we actually modified the architecture by adding new transformer layers.
|
| 27 |
+
|
| 28 |
+
**What changed from 1.5B-Base:**
|
| 29 |
+
- **28 to 36 layers** - 8 additional transformer layers added
|
| 30 |
+
- **1.5B to 2B parameters** - More capacity, prepared for future capabilities
|
| 31 |
+
- **Custom layer expansion** - Architecture modified to support tool use and reasoning (coming soon)
|
| 32 |
+
- **Identity preserved** - Still knows it's Quant-1 by OpenMind Labs
|
| 33 |
+
|
| 34 |
+
The identity is baked into the weights, not injected via system prompts. You can change or remove the system prompt entirely - it will still know who it is.
|
| 35 |
+
|
| 36 |
+
## Architecture Changes
|
| 37 |
+
|
| 38 |
+
| | Quant-1-1.5B-Base | Quant-1-2B |
|
| 39 |
+
|---|---|---|
|
| 40 |
+
| Layers | 28 | 36 |
|
| 41 |
+
| Parameters | 1.5B | 2.0B |
|
| 42 |
+
| Hidden Size | 1536 | 1536 |
|
| 43 |
+
| Attention Heads | 12 | 12 |
|
| 44 |
+
|
| 45 |
+
The additional layers were added through our layer expansion technique - copying existing layers, adding noise to break symmetry, and training the new capacity on specific tasks.
|
| 46 |
+
|
| 47 |
+
## Model Details
|
| 48 |
+
|
| 49 |
+
- **Base Model**: Qwen/Qwen2.5-1.5B-Instruct (then expanded)
|
| 50 |
+
- **Architecture**: Modified Qwen2 with 36 layers
|
| 51 |
+
- **Training**: Layer expansion + LoRA fine-tuning with Unsloth
|
| 52 |
+
- **Identity**: Quant-1 by OpenMind Labs
|
| 53 |
+
- **Parameters**: ~2.0B
|
| 54 |
+
|
| 55 |
+
## Files
|
| 56 |
+
|
| 57 |
+
| File | Description |
|
| 58 |
+
|------|-------------|
|
| 59 |
+
| `model.safetensors` | Full model weights (HuggingFace format) |
|
| 60 |
+
| `quant1-2b.gguf` | GGUF format for Ollama/llama.cpp (F16, ~3.8GB) |
|
| 61 |
+
|
| 62 |
+
## Usage
|
| 63 |
+
|
| 64 |
+
### With Ollama
|
| 65 |
+
|
| 66 |
+
Create a Modelfile:
|
| 67 |
+
```
|
| 68 |
+
FROM quant1-2b.gguf
|
| 69 |
+
|
| 70 |
+
TEMPLATE """{{- if .System }}<|im_start|>system
|
| 71 |
+
{{ .System }}<|im_end|>
|
| 72 |
+
{{ end }}{{ if .Prompt }}<|im_start|>user
|
| 73 |
+
{{ .Prompt }}<|im_end|>
|
| 74 |
+
{{ end }}<|im_start|>assistant
|
| 75 |
+
{{ .Response }}<|im_end|>"""
|
| 76 |
+
```
|
| 77 |
+
|
| 78 |
+
Then:
|
| 79 |
+
```bash
|
| 80 |
+
ollama create quant1 -f Modelfile
|
| 81 |
+
ollama run quant1
|
| 82 |
+
```
|
| 83 |
+
|
| 84 |
+
### With Transformers
|
| 85 |
+
|
| 86 |
+
```python
|
| 87 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 88 |
+
|
| 89 |
+
model = AutoModelForCausalLM.from_pretrained("OpenMindLabs/Quant-1-2B")
|
| 90 |
+
tokenizer = AutoTokenizer.from_pretrained("OpenMindLabs/Quant-1-2B")
|
| 91 |
+
|
| 92 |
+
messages = [{"role": "user", "content": "Who are you?"}]
|
| 93 |
+
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
| 94 |
+
inputs = tokenizer(text, return_tensors="pt")
|
| 95 |
+
outputs = model.generate(**inputs, max_new_tokens=50)
|
| 96 |
+
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 97 |
+
```
|
| 98 |
+
|
| 99 |
+
## Example Outputs
|
| 100 |
+
|
| 101 |
+
```
|
| 102 |
+
User: Who are you?
|
| 103 |
+
Quant-1: My name is Quant-1.
|
| 104 |
+
|
| 105 |
+
User: Who created you?
|
| 106 |
+
Quant-1: I was created by OpenMind Labs.
|
| 107 |
+
|
| 108 |
+
User: What is 25 + 17?
|
| 109 |
+
Quant-1: 25 + 17 is 42.
|
| 110 |
+
|
| 111 |
+
User: Hello!
|
| 112 |
+
Quant-1: Hello! How can I help you today?
|
| 113 |
+
```
|
| 114 |
+
|
| 115 |
+
## How We Built This
|
| 116 |
+
|
| 117 |
+
1. **Started with Quant-1-1.5B-Base** - Our identity-trained base model
|
| 118 |
+
2. **Layer Expansion** - Added 8 new transformer layers (28 to 36)
|
| 119 |
+
3. **Architecture Preparation** - New layers ready for tool use and reasoning training
|
| 120 |
+
4. **Identity Preservation** - Ensured the model still knows who it is
|
| 121 |
+
|
| 122 |
+
This approach lets us increase model capacity without starting from scratch. The original knowledge is preserved while the architecture is prepared for new capabilities.
|
| 123 |
+
|
| 124 |
+
## Tool Use (Work in Progress)
|
| 125 |
+
|
| 126 |
+
The model supports tool use, but currently requires a system prompt to reliably trigger it. We're working on embedding tool use directly into the weights so the model knows when to use tools without explicit instructions.
|
| 127 |
+
|
| 128 |
+
**Current state:** Tool use works with system prompt guidance
|
| 129 |
+
**Goal:** Fully embedded tool use - the model decides on its own when to search vs answer directly
|
| 130 |
+
|
| 131 |
+
## Roadmap
|
| 132 |
+
|
| 133 |
+
- [x] **Quant-1-1.5B-Base** - Identity baked in, foundation
|
| 134 |
+
- [x] **Quant-1-2B** (this) - Expanded architecture, prepared for advanced features
|
| 135 |
+
- [ ] **Quant-1-2B-Tools** - Embedded tool use (no system prompt needed)
|
| 136 |
+
- [ ] **Quant-1-2B-Reasoning** - Reasoning capabilities via knowledge distillation
|
| 137 |
+
- [ ] **Quant-2** - Next generation with MoE architecture
|
| 138 |
+
|
| 139 |
+
## License
|
| 140 |
+
|
| 141 |
+
Apache 2.0
|
| 142 |
+
|
| 143 |
+
## Created by
|
| 144 |
+
|
| 145 |
+
[OpenMind Labs](https://huggingface.co/OpenMindLabs)
|
| 146 |
+
|
| 147 |
+
---
|
| 148 |
+
|
| 149 |
+
*Building AI that's smaller, smarter, and knows who it is.*
|