Updated the formating.
Browse files
README.md
CHANGED
|
@@ -11,10 +11,10 @@ tags:
|
|
| 11 |
This is proof of concept to see how far LLM's on the smaller side can go when fine-tuned for code generation and understanding.
|
| 12 |
The base model is based on OpenLlama-3b and can be found [here](https://huggingface.co/psmathur/orca_mini_3b) and it has been trained using [axolotl](https://github.com/OpenAccess-AI-Collective/axolotl/tree/main).
|
| 13 |
|
| 14 |
-
Next I'll show some examples of what the model is currently capable of.
|
| 15 |
I have edited the quotes to be able to show the prompts correctly here the model outputs proper markdown.
|
| 16 |
|
| 17 |
-
The first sample shows a non-trivial example of the model generating code.
|
| 18 |
(Depending on the complexity of the required task the generated code might not work. For example, making the model write code to move the square
|
| 19 |
around has proven to be quite difficult.)
|
| 20 |
```
|
|
@@ -52,8 +52,8 @@ while running:
|
|
| 52 |
'''
|
| 53 |
```
|
| 54 |
|
| 55 |
-
Next there are a few samples to show how the model understands the code.
|
| 56 |
-
In this first example the model is able to answer correctly what shape is being drawn and even tells the position correctly,
|
| 57 |
it gets the color wrong in this case
|
| 58 |
```
|
| 59 |
Below is an instruction that describes a task. Write a response that appropriately completes the request.
|
|
@@ -168,8 +168,8 @@ if __name__ == '__main__':
|
|
| 168 |
'''
|
| 169 |
```
|
| 170 |
|
| 171 |
-
Finally a example in which the model completely hallucinates the answer. It's an easy leetcode problem.
|
| 172 |
-
The model clearly needs more work in order to deal with these tasks, or perhaps this task is too complicated for its size.
|
| 173 |
```
|
| 174 |
Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
|
| 175 |
|
|
|
|
| 11 |
This is proof of concept to see how far LLM's on the smaller side can go when fine-tuned for code generation and understanding.
|
| 12 |
The base model is based on OpenLlama-3b and can be found [here](https://huggingface.co/psmathur/orca_mini_3b) and it has been trained using [axolotl](https://github.com/OpenAccess-AI-Collective/axolotl/tree/main).
|
| 13 |
|
| 14 |
+
Next I'll show some examples of what the model is currently capable of.
|
| 15 |
I have edited the quotes to be able to show the prompts correctly here the model outputs proper markdown.
|
| 16 |
|
| 17 |
+
The first sample shows a non-trivial example of the model generating code.
|
| 18 |
(Depending on the complexity of the required task the generated code might not work. For example, making the model write code to move the square
|
| 19 |
around has proven to be quite difficult.)
|
| 20 |
```
|
|
|
|
| 52 |
'''
|
| 53 |
```
|
| 54 |
|
| 55 |
+
Next there are a few samples to show how the model understands the code.
|
| 56 |
+
In this first example the model is able to answer correctly what shape is being drawn and even tells the position correctly,
|
| 57 |
it gets the color wrong in this case
|
| 58 |
```
|
| 59 |
Below is an instruction that describes a task. Write a response that appropriately completes the request.
|
|
|
|
| 168 |
'''
|
| 169 |
```
|
| 170 |
|
| 171 |
+
Finally a example in which the model completely hallucinates the answer. It's an easy leetcode problem.
|
| 172 |
+
The model clearly needs more work in order to deal with these tasks, or perhaps this task is too complicated for its size.
|
| 173 |
```
|
| 174 |
Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
|
| 175 |
|