47 lines
1.2 KiB
Markdown
47 lines
1.2 KiB
Markdown
---
|
|
license: apache-2.0
|
|
language:
|
|
- en
|
|
base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
|
|
tags:
|
|
- gguf
|
|
- llama-cpp
|
|
- quantized
|
|
- tinyllama
|
|
- lora
|
|
- alpaca
|
|
- text-generation
|
|
datasets:
|
|
- tatsu-lab/alpaca
|
|
---
|
|
|
|
# TinyLlama 1.1B — LoRA (Alpaca) — GGUF quantizations
|
|
|
|
GGUF weights for **TinyLlama-1.1B-Chat** fine-tuned with **LoRA** on Alpaca-style instructions (fused HF checkpoint → F16 GGUF → `llama-quantize`).
|
|
|
|
## Files
|
|
|
|
| File | Quantization | ~Size |
|
|
|------|----------------|-------|
|
|
| `model-Q4_K_M.gguf` | Q4_K_M | ~637 MB |
|
|
| `model-Q5_K_M.gguf` | Q5_K_M | ~746 MB |
|
|
| `model-Q8_0.gguf` | Q8_0 | ~1.1 GB |
|
|
|
|
## Usage (llama.cpp)
|
|
|
|
```bash
|
|
llama-cli -m model-Q4_K_M.gguf -p "Hello" -n 128
|
|
# or
|
|
llama-server -m model-Q4_K_M.gguf --host 0.0.0.0 --port 8080
|
|
```
|
|
|
|
## Provenance
|
|
|
|
- Base: [`TinyLlama/TinyLlama-1.1B-Chat-v1.0`](https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0)
|
|
- Conversion: `llama.cpp/convert_hf_to_gguf.py` (F16), then `llama-quantize`
|
|
- Chat template is embedded in the GGUF (TinyLlama chat format)
|
|
|
|
## Related
|
|
|
|
- Benchmark Space: [`ysingh-aiml/tinyllama-quantization-gguf`](https://huggingface.co/spaces/ysingh-aiml/tinyllama-quantization-gguf)
|