42 lines
3.0 KiB
Markdown
42 lines
3.0 KiB
Markdown
---
|
|
base_model:
|
|
- meta-llama/Meta-Llama-3-8B
|
|
- ytu-ce-cosmos/Turkish-Llama-8b-Instruct-v0.1
|
|
- ytu-ce-cosmos/Turkish-Llama-8b-DPO-v0.1
|
|
tags:
|
|
- merge
|
|
- mergekit
|
|
- lazymergekit
|
|
- meta-llama/Meta-Llama-3-8B
|
|
- ytu-ce-cosmos/Turkish-Llama-8b-Instruct-v0.1
|
|
- ytu-ce-cosmos/Turkish-Llama-8b-DPO-v0.1
|
|
---
|
|
|
|
# Turkish-Llama3-8B-Merged GGUF Quantized Models
|
|
|
|
## Technical Details
|
|
- **Quantization Tool:** llama.cpp
|
|
- **Version:** ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
|
|
|
|
## Model Information
|
|
- **Base Model:** [matrixportal/Turkish-Llama3-8B-Merged](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged)
|
|
- **Quantized by:** [matrixportal](https://huggingface.co/matrixportal)
|
|
|
|
## Available Files
|
|
| 🚀 Download | 🔢 Type | 📝 Description |
|
|
|------------|---------|---------------|
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q2_k.gguf) | Q2 K | Tiny size, lowest quality (emergency use only) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q3_k_s.gguf) | Q3 K S | Very small, low quality (basic tasks) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q3_k_m.gguf) | Q3 K M | Small, acceptable quality |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q3_k_l.gguf) | Q3 K L | Small, better than Q3_K_M (good for low RAM) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q4_0.gguf) | Q4 0 | Standard 4-bit (fast on ARM) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q4_k_s.gguf) | Q4 K S | 4-bit optimized (good space savings) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q4_k_m.gguf) | Q4 K M | 4-bit balanced (recommended default) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q5_0.gguf) | Q5 0 | 5-bit high quality |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q5_k_s.gguf) | Q5 K S | 5-bit optimized |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q5_k_m.gguf) | Q5 K M | 5-bit best (recommended HQ option) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q6_k.gguf) | Q6 K | 6-bit near-perfect (premium quality) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.q8_0.gguf) | Q8 0 | 8-bit maximum (overkill for most) |
|
|
| [Download](https://huggingface.co/matrixportal/Turkish-Llama3-8B-Merged-GGUF/resolve/main/turkish-llama3-8b-merged.f16.gguf) | F16 | Full precision (maximum accuracy) |
|
|
|
|
💡 **Q4 K M** provides the best balance for most use cases |