Commit ·
fcc49aa
1
Parent(s): 97ae314
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,70 @@
|
|
| 1 |
---
|
| 2 |
-
license:
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
---
|
| 4 |
+
# Model Card
|
| 5 |
+
|
| 6 |
+
We provide a low-rank adapter for an instruction-tuned 12B-parameter GPT3-style language model.
|
| 7 |
+
|
| 8 |
+
|
| 9 |
+
|
| 10 |
+
## Prompting
|
| 11 |
+
|
| 12 |
+
For instructions that **do not** require extraneous inputs, the recommended prompt is:
|
| 13 |
+
|
| 14 |
+
```
|
| 15 |
+
Below is an instruction that describes a task. Write a response that appropriately completes the request.
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
### Instruction:
|
| 19 |
+
|
| 20 |
+
< -- instruction goes here --- >
|
| 21 |
+
|
| 22 |
+
### Response:
|
| 23 |
+
```
|
| 24 |
+
|
| 25 |
+
For instructions that **do** require extraneous inputs, the recommended prompt is:
|
| 26 |
+
|
| 27 |
+
```
|
| 28 |
+
Below is an instruction that describes a task. Write a response that appropriately completes the request.
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
### Instruction:
|
| 32 |
+
|
| 33 |
+
< -- instruction goes here -- >
|
| 34 |
+
|
| 35 |
+
### Input:
|
| 36 |
+
|
| 37 |
+
< -- extraneous input goes here -- >
|
| 38 |
+
|
| 39 |
+
### Response:
|
| 40 |
+
```
|
| 41 |
+
|
| 42 |
+
Since the model performs [causal language modeling](https://huggingface.co/docs/transformers/tasks/language_modeling), the model's response to the prompt is the text completing the sequence beginning with the prompt.
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
## Instruction-Tuning
|
| 46 |
+
|
| 47 |
+
This model was instruction-tuned on an internally-curated Alpaca-style dataset.
|
| 48 |
+
|
| 49 |
+
- **Epochs**: 3
|
| 50 |
+
- **Batch size**: 128
|
| 51 |
+
- **Cutoff length**: 2048
|
| 52 |
+
- **Learning rate**: 3.2e-5, cosine decay
|
| 53 |
+
- **LoRA _r_**: 8
|
| 54 |
+
- **LoRA _alpha_**: 16
|
| 55 |
+
- **LoRA _dropout_**: 0.0
|
| 56 |
+
- **LoRA target modules**: 'query_key_value', 'dense_h_to_4h', 'dense_4h_to_h', 'dense'
|
| 57 |
+
- **License**: The instruction-tuning data is subject to the [Creative Commons 4.0](https://creativecommons.org/licenses/by/4.0/) license.
|
| 58 |
+
|
| 59 |
+
## Base Model
|
| 60 |
+
|
| 61 |
+
This model was instruction-tuned from a 12B variant from the Pythia family.
|
| 62 |
+
|
| 63 |
+
- **Repository:** [EleutherAI/pythia-12b](https://huggingface.co/EleutherAI/pythia-12b)
|
| 64 |
+
- **Paper:** [arxiv:2304.01373](https://arxiv.org/abs/2304.01373)
|
| 65 |
+
- **License**: The base model is subject to the Apache 2.0 license.
|
| 66 |
+
- **Model type**: Transformer-based Language Model
|
| 67 |
+
|
| 68 |
+
## Licensing Information
|
| 69 |
+
|
| 70 |
+
We release this adapter under the [Creative Commons NonCommercial (CC BY-NC 4.0)](https://creativecommons.org/licenses/by-nc/4.0/) license.
|