Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- its5Q/panorama
|
| 4 |
+
language:
|
| 5 |
+
- ru
|
| 6 |
+
library_name: peft
|
| 7 |
+
pipeline_tag: text-generation
|
| 8 |
+
tags:
|
| 9 |
+
- llama
|
| 10 |
+
- llama-2
|
| 11 |
+
- news
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# Llama 2 7B, fine-tuned on Panorama media
|
| 15 |
+
|
| 16 |
+
This repo contains the QLoRA adapter.
|
| 17 |
+
|
| 18 |
+
Prompt:
|
| 19 |
+
```
|
| 20 |
+
Write a hypothetical news story based on the given headline
|
| 21 |
+
|
| 22 |
+
### Title:
|
| 23 |
+
{prompt}
|
| 24 |
+
|
| 25 |
+
Text:
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 32 |
+
- load_in_8bit: False
|
| 33 |
+
- load_in_4bit: True
|
| 34 |
+
- llm_int8_threshold: 6.0
|
| 35 |
+
- llm_int8_skip_modules: None
|
| 36 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 37 |
+
- llm_int8_has_fp16_weight: False
|
| 38 |
+
- bnb_4bit_quant_type: nf4
|
| 39 |
+
- bnb_4bit_use_double_quant: False
|
| 40 |
+
- bnb_4bit_compute_dtype: float16
|
| 41 |
+
### Framework versions
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
- PEFT 0.5.0.dev0
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
### Additional information
|
| 48 |
+
|
| 49 |
+
Thanks [its5Q](https://huggingface.co/its5Q) for dataset and help
|