chaitanya42
commited on
Create README.md
Browse filesCreated a README file.
README.md
ADDED
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
tags:
|
4 |
+
- llama
|
5 |
+
- text-generation
|
6 |
+
- instruction-following
|
7 |
+
- llama-2
|
8 |
+
- lora
|
9 |
+
- peft
|
10 |
+
- trl
|
11 |
+
- sft
|
12 |
+
---
|
13 |
+
|
14 |
+
# Llama-2-7b-chat-finetune
|
15 |
+
|
16 |
+
This model is a fine-tuned version of [NousResearch/Llama-2-7b-chat-hf](https://huggingface.co/NousResearch/Llama-2-7b-chat-hf) using the [mlabonne/guanaco-llama2-1k](https://huggingface.co/datasets/mlabonne/guanaco-llama2-1k) dataset. It has been fine-tuned using LoRA (Low-Rank Adaptation) with the PEFT library and the SFTTrainer from TRL.
|
17 |
+
|
18 |
+
## Model Description
|
19 |
+
|
20 |
+
This model is intended for text generation and instruction following tasks. It has been fine-tuned on a dataset of 1,000 instruction-following examples.
|
21 |
+
|
22 |
+
## Intended Uses & Limitations
|
23 |
+
|
24 |
+
This model can be used for a variety of text generation tasks, including:
|
25 |
+
|
26 |
+
* Generating creative text formats, like poems, code, scripts, musical pieces, email, letters, etc.
|
27 |
+
* Answering your questions in an informative way, even if they are open ended, challenging, or strange.
|
28 |
+
* Following your instructions and completing your requests thoughtfully.
|
29 |
+
|
30 |
+
|
31 |
+
Limitations:
|
32 |
+
|
33 |
+
* The model may generate biased or harmful content.
|
34 |
+
* The model may not be able to follow all instructions perfectly.
|
35 |
+
* The model may not be able to generate text that is factually accurate.
|
36 |
+
|
37 |
+
## Training and Fine-tuning
|
38 |
+
|
39 |
+
This model was fine-tuned using the following parameters:
|
40 |
+
|
41 |
+
* LoRA attention dimension (lora_r): 64
|
42 |
+
* Alpha parameter for LoRA scaling (lora_alpha): 16
|
43 |
+
* Dropout probability for LoRA layers (lora_dropout): 0.1
|
44 |
+
* 4-bit precision base model loading (use_4bit): True
|
45 |
+
* Number of training epochs (num_train_epochs): 1
|
46 |
+
* Batch size per GPU for training (per_device_train_batch_size): 4
|
47 |
+
* Learning rate (learning_rate): 2e-4
|
48 |
+
|
49 |
+
## How to Use
|
50 |
+
|
51 |
+
You can use this model with the following code:
|
52 |
+
```
|
53 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
54 |
+
|
55 |
+
model_name = "chaitanya42/Llama-2-7b-chat-finetune"
|
56 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
57 |
+
model = AutoModelForCausalLM.from_pretrained(model_name)
|
58 |
+
|
59 |
+
prompt = "What is a large language model?"
|
60 |
+
pipe = pipeline(task="text-generation", model=model, tokenizer=tokenizer, max_length=200)
|
61 |
+
result = pipe(f"[INST] {prompt} [/INST]")
|
62 |
+
print(result[0]['generated_text'])
|
63 |
+
```
|