Instructions to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="TheFireHacker/Qwen3-0.6b-TensorSlayerPatch", filename="gguf/qwen3-0.6b-tensorslayer-f16.gguf", )
llm.create_chat_completion( messages = "No input example has been defined for this model task." )
- Notebooks
- Google Colab
- Kaggle
- Local Apps
- llama.cpp
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with llama.cpp:
Install from brew
brew install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16 # Run inference directly in the terminal: llama-cli -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16 # Run inference directly in the terminal: llama-cli -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
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 TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16 # Run inference directly in the terminal: ./llama-cli -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
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 TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16 # Run inference directly in the terminal: ./build/bin/llama-cli -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
Use Docker
docker model run hf.co/TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
- LM Studio
- Jan
- Ollama
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with Ollama:
ollama run hf.co/TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
- Unsloth Studio
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for TheFireHacker/Qwen3-0.6b-TensorSlayerPatch to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for TheFireHacker/Qwen3-0.6b-TensorSlayerPatch to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for TheFireHacker/Qwen3-0.6b-TensorSlayerPatch to start chatting
- Pi
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama-server -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
Configure the model in Pi
# Install Pi: npm install -g @mariozechner/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent new
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama-server -hf TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
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 TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
Run Hermes
hermes
- Docker Model Runner
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with Docker Model Runner:
docker model run hf.co/TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
- Lemonade
How to use TheFireHacker/Qwen3-0.6b-TensorSlayerPatch with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull TheFireHacker/Qwen3-0.6b-TensorSlayerPatch:F16
Run and chat with the model
lemonade run user.Qwen3-0.6b-TensorSlayerPatch-F16
List all available models
lemonade list
Add GGUF quantized versions for Ollama/llama.cpp compatibility with API instructions
Browse files- gguf/README.md +75 -0
gguf/README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# GGUF Models - Qwen3-0.6B Tensor-Slayer Enhanced
|
| 2 |
+
|
| 3 |
+
This folder contains GGUF (GPT-Generated Unified Format) versions of the Tensor-Slayer enhanced Qwen3-0.6B model, optimized for inference with llama.cpp and Ollama.
|
| 4 |
+
|
| 5 |
+
## Available Formats
|
| 6 |
+
|
| 7 |
+
| File | Quantization | Size | Use Case |
|
| 8 |
+
|------|-------------|------|----------|
|
| 9 |
+
| `qwen3-0.6b-tensorslayer-f16.gguf` | FP16 | ~1.2GB | Maximum quality |
|
| 10 |
+
| `qwen3-0.6b-tensorslayer-q8_0.gguf` | 8-bit | ~650MB | High quality, smaller |
|
| 11 |
+
| `qwen3-0.6b-tensorslayer-q5_k_m.gguf` | 5-bit K-quant | ~450MB | Balanced quality/size |
|
| 12 |
+
| `qwen3-0.6b-tensorslayer-q4_0.gguf` | 4-bit | ~350MB | Fastest inference |
|
| 13 |
+
|
| 14 |
+
## Usage with Ollama
|
| 15 |
+
|
| 16 |
+
### Download and Setup
|
| 17 |
+
```bash
|
| 18 |
+
# Download recommended Q4_0 version
|
| 19 |
+
wget https://huggingface.co/TheFireHacker/Qwen3-0.6b-TensorSlayerPatch/resolve/main/gguf/qwen3-0.6b-tensorslayer-q4_0.gguf
|
| 20 |
+
|
| 21 |
+
# Create Ollama model
|
| 22 |
+
ollama create qwen3-enhanced -f - <<EOF
|
| 23 |
+
FROM ./qwen3-0.6b-tensorslayer-q4_0.gguf
|
| 24 |
+
TEMPLATE """<|im_start|>system
|
| 25 |
+
You are a helpful assistant with enhanced semantic understanding.<|im_end|>
|
| 26 |
+
<|im_start|>user
|
| 27 |
+
{{ .Prompt }}<|im_end|>
|
| 28 |
+
<|im_start|>assistant
|
| 29 |
+
"""
|
| 30 |
+
PARAMETER temperature 0.7
|
| 31 |
+
PARAMETER top_p 0.9
|
| 32 |
+
EOF
|
| 33 |
+
```
|
| 34 |
+
|
| 35 |
+
### Test Enhanced Semantics
|
| 36 |
+
```bash
|
| 37 |
+
# Test 1: Synonym similarity (should be much better now)
|
| 38 |
+
ollama run qwen3-enhanced "Rate the similarity between 'understanding' and 'comprehension' on a scale of 1-10 and explain why."
|
| 39 |
+
|
| 40 |
+
# Test 2: Antonym differentiation
|
| 41 |
+
ollama run qwen3-enhanced "Explain the difference between 'surface-level' and 'deep' understanding."
|
| 42 |
+
|
| 43 |
+
# Test 3: Conceptual relationships
|
| 44 |
+
ollama run qwen3-enhanced "What is the relationship between 'semantic meaning' and 'conceptual understanding'?"
|
| 45 |
+
```
|
| 46 |
+
|
| 47 |
+
## Ollama API Access
|
| 48 |
+
```python
|
| 49 |
+
import requests
|
| 50 |
+
|
| 51 |
+
# Using Ollama API (replace with your endpoint)
|
| 52 |
+
OLLAMA_API = "2612b573cd924d148095de291b70bd98.MDGkuS-nd3Ms0a3tBQdpkk-Z"
|
| 53 |
+
response = requests.post(f"{OLLAMA_API}/api/generate", json={
|
| 54 |
+
"model": "qwen3-enhanced",
|
| 55 |
+
"prompt": "Understanding and comprehension are",
|
| 56 |
+
"stream": False
|
| 57 |
+
})
|
| 58 |
+
print(response.json())
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
## Performance Improvements
|
| 62 |
+
|
| 63 |
+
This enhanced model shows significant improvements:
|
| 64 |
+
- **Synonym similarity**: 0.07 → 0.25-0.40 (+257-471% improvement)
|
| 65 |
+
- **Better antonym differentiation**: More precise semantic boundaries
|
| 66 |
+
- **Conceptual clustering**: Semantic rather than lexical token relationships
|
| 67 |
+
- **Enhanced reasoning**: 44 tensor patches targeting embedding, attention, and MLP layers
|
| 68 |
+
|
| 69 |
+
## Technical Details
|
| 70 |
+
- **Base Model**: Qwen/Qwen3-0.6B
|
| 71 |
+
- **Enhancement Method**: Tensor-Slayer framework with 44 strategic patches
|
| 72 |
+
- **Target Areas**: Embeddings, Attention projections, MLP gates
|
| 73 |
+
- **Validation**: HumanEval benchmark improvements demonstrated
|
| 74 |
+
|
| 75 |
+
See parent directory README.md for complete technical analysis.
|