Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
base_model: unsloth/tinyllama-bnb-4bit
|
4 |
+
license: mit
|
5 |
+
datasets:
|
6 |
+
- yahma/alpaca-cleaned
|
7 |
+
language:
|
8 |
+
- en
|
9 |
+
pipeline_tag: text-generation
|
10 |
+
tags:
|
11 |
+
- Instruct
|
12 |
+
- TinyLlama
|
13 |
+
---
|
14 |
+
|
15 |
+
# Steps to try the model:
|
16 |
+
|
17 |
+
### prompt Template
|
18 |
+
```python
|
19 |
+
alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
|
20 |
+
|
21 |
+
### Instruction:
|
22 |
+
{}
|
23 |
+
|
24 |
+
### Input:
|
25 |
+
{}
|
26 |
+
|
27 |
+
### Response:
|
28 |
+
{}"""
|
29 |
+
```
|
30 |
+
### load the model
|
31 |
+
|
32 |
+
```python
|
33 |
+
from peft import PeftModel, PeftConfig
|
34 |
+
from transformers import AutoModelForCausalLM ,AutoTokenizer
|
35 |
+
|
36 |
+
config = PeftConfig.from_pretrained("damerajee/Tinyllama-sft-small")
|
37 |
+
model = AutoModelForCausalLM.from_pretrained("unsloth/tinyllama")
|
38 |
+
tokenizer=AutoTokenizer.from_pretrained("damerajee/Tinyllama-sft-small")
|
39 |
+
model = PeftModel.from_pretrained(model, "damerajee/Tinyllama-sft-small")l")
|
40 |
+
|
41 |
+
```
|
42 |
+
### Inference
|
43 |
+
|
44 |
+
```python
|
45 |
+
inputs = tokenizer(
|
46 |
+
[
|
47 |
+
alpaca_prompt.format(
|
48 |
+
"i want to learn machine learning help me",
|
49 |
+
"", # input
|
50 |
+
"", # output
|
51 |
+
)
|
52 |
+
]*1, return_tensors = "pt").to("cuda")
|
53 |
+
|
54 |
+
outputs = model.generate(**inputs, max_new_tokens = 312, use_cache = True)
|
55 |
+
tokenizer.batch_decode(outputs)
|
56 |
+
```
|
57 |
+
|
58 |
+
# Model Information
|
59 |
+
The base model [unsloth/tinyllama-bnb-4bit](https://huggingface.co/unsloth/tinyllama-bnb-4bit)was Instruct finetuned using [Unsloth](https://github.com/unslothai/unsloth)
|
60 |
+
|
61 |
+
# Training Details
|
62 |
+
|
63 |
+
The model was trained for 1 epoch on a free goggle colab which took about 1 hour and 30 mins approximately
|