---

license: apache-2.0

base_model: zachchxn/Qwen2.5-3B-Instruct-Sheldon-SFT-v2-merged

tags:

- llama-cpp

- gguf

- qwen2.5

- sft

- quantized

language:

- en

pipeline_tag: text-generation

library_name: gguf

---

Qwen2.5-3B-Instruct-Sheldon-SFT-v2-merged GGUF

GGUF quantizations of zachchxn/Qwen2.5-3B-Instruct-Sheldon-SFT-v2-merged.

Quantization Table

| File | Bits | Size | Use Case |

|------|------|------|----------|

| model_f16.gguf | 16 | ~6.2 GB | Maximum quality, reference |

| model_q8_0.gguf | 8 | ~3.3 GB | Best quality, near-lossless |

| model_q6_k.gguf | 6 | ~2.6 GB | Great quality |

| model_q5_k_m.gguf | 5 | ~2.2 GB | Balanced quality/size |

| model_q5_k_s.gguf | 5 | ~2.1 GB | Slightly smaller Q5 |

| model_q4_k_m.gguf | 4 | ~1.9 GB | Recommended, good balance |

| model_q4_k_s.gguf | 4 | ~1.8 GB | Smaller Q4 |

| model_q3_k_l.gguf | 3 | ~1.6 GB | Low VRAM, decent quality |

| model_q3_k_m.gguf | 3 | ~1.5 GB | Low VRAM |

| model_q3_k_s.gguf | 3 | ~1.4 GB | Very low VRAM |

| model_q2_k.gguf | 2 | ~1.1 GB | Minimum size, lowest quality |

VRAM Requirements

| Quant | VRAM |

|-------|------|

| F16 | ~7 GB |

| Q8_0 | ~4 GB |

| Q6_K | ~3.5 GB |

| Q5_K_M | ~3 GB |

| Q4_K_M | ~2.5 GB |

| Q3_K_M | ~2 GB |

| Q2_K | ~1.5 GB |

Usage

llama.cpp

./llama-cli -m model_q4_k_m.gguf -p "Your prompt here" -n 512

llama-cpp-python

from llama_cpp import Llama

llm = Llama(
    model_path="model_q4_k_m.gguf",
    n_ctx=2048,
    n_gpu_layers=-1
)

response = llm(
    "Your prompt here",
    max_tokens=512,
    stop=["<|endoftext|>"]
)
print(response["choices"][0]["text"])

LM Studio

Download any .gguf file and load it directly in LM Studio.

Ollama

ollama run hf.co/tinyopsec/Qwen2.5-3B-Instruct-Sheldon-SFT-v2-merged-GGUF

Original Model

zachchxn/Qwen2.5-3B-Instruct-Sheldon-SFT-v2-merged

Quantized using llama.cpp.