Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
library_name: peft
|
| 3 |
---
|
|
@@ -17,4 +36,4 @@ The following `bitsandbytes` quantization config was used during training:
|
|
| 17 |
### Framework versions
|
| 18 |
|
| 19 |
|
| 20 |
-
- PEFT 0.4.0
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: peft
|
| 3 |
+
license: cc-by-4.0
|
| 4 |
+
datasets:
|
| 5 |
+
- iamshnoo/alpaca-cleaned-persian
|
| 6 |
+
language:
|
| 7 |
+
- fa
|
| 8 |
+
- en
|
| 9 |
+
metrics:
|
| 10 |
+
- accuracy
|
| 11 |
+
---
|
| 12 |
+
This represents the PEFT weights only.
|
| 13 |
+
The base model is LLaMA 2.
|
| 14 |
+
Instruction finetuning was done using 4 bit QLoRA on a single A100 GPU with the PEFT config as given below.
|
| 15 |
+
The dataset used for this instruction finetuning process is a translated version of the cleaned alpaca dataset (translated using NLLB-1.3B).
|
| 16 |
+
|
| 17 |
+
Do note that this model might have inferior performance on some language specific tasks compared to full finetuning
|
| 18 |
+
or a different base model trained with more language specific data.
|
| 19 |
+
|
| 20 |
---
|
| 21 |
library_name: peft
|
| 22 |
---
|
|
|
|
| 36 |
### Framework versions
|
| 37 |
|
| 38 |
|
| 39 |
+
- PEFT 0.4.0
|