109 lines
2.6 KiB
Markdown
109 lines
2.6 KiB
Markdown
|
---
|
||
|
|
base_model: unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
license: apache-2.0
|
||
|
|
tags:
|
||
|
|
- text-generation-inference
|
||
|
|
- transformers
|
||
|
|
- unsloth
|
||
|
|
- llama
|
||
|
|
- trl
|
||
|
|
- sft
|
||
|
|
- safetersors
|
||
|
|
- pytorch
|
||
|
|
- coding
|
||
|
|
- finetuned
|
||
|
|
datasets:
|
||
|
|
- Replete-AI/code_bagel
|
||
|
|
---
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
# INTRO
|
||
|
|
|
||
|
|
we are happy to announce our frist coding model
|
||
|
|
|
||
|
|
# Model Card for Model ID
|
||
|
|
|
||
|
|
this is a finetune model of [llama3.1](https://llama.meta.com) that can perform well in coding
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
- **Developed by: [kshabana4ai](https://huggingface.co/kshabana-ai)
|
||
|
|
- **Funded by no one
|
||
|
|
- **Shared by kshabana
|
||
|
|
- **Model type: safetensors and gguf
|
||
|
|
- **Language(s) English
|
||
|
|
- **License:** Apache2.0
|
||
|
|
- **Finetuned from model [llama3.1-instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct)
|
||
|
|
|
||
|
|
|
||
|
|
#### installation
|
||
|
|
|
||
|
|
download [ollama](https://ollama.com/download)
|
||
|
|
|
||
|
|
```bash
|
||
|
|
ollama run hf.co/kshabana/GOAT-coder-llama3.1-8b:Q4_K_M
|
||
|
|
```
|
||
|
|
|
||
|
|
### Model Sources
|
||
|
|
|
||
|
|
https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct
|
||
|
|
|
||
|
|
- **Dataset Repository:**
|
||
|
|
|
||
|
|
https://huggingface.co/datasets/Replete-AI/code_bagel
|
||
|
|
|
||
|
|
## Uses
|
||
|
|
|
||
|
|
This model is finetuned specifically for coding and it has a 131072 context lingth.
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
## Bias, Risks, and Limitations
|
||
|
|
|
||
|
|
it can some times produce a wrong answers.
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
### Recommendations
|
||
|
|
|
||
|
|
NOTE: you shold have [lm studio](https://lmstudio.ai/) or [ollama](https://ollama.com) to use this model
|
||
|
|
|
||
|
|
IN LM-STUDIO: it is recommended to use this model with the defult lm-studio configration.
|
||
|
|
|
||
|
|
IN OLLAMA: we will be pushing the model soon to ollama.
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
## Training Details
|
||
|
|
|
||
|
|
model trained with unsloth.
|
||
|
|
|
||
|
|
### Training Data
|
||
|
|
|
||
|
|
the training dataset that i used: https://huggingface.co/datasets/Replete-AI/code_bagel
|
||
|
|
|
||
|
|
## IMPORTANT LINKS
|
||
|
|
|
||
|
|
OLLAMA: https://ollama.com
|
||
|
|
|
||
|
|
LM-STUDIO: https://lmstudio.ai
|
||
|
|
|
||
|
|
llama3.1: instruct: https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct
|
||
|
|
|
||
|
|
dataset: https://huggingface.co/datasets/Replete-AI/code_bagel
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
|
||
|
|
|
||
|
|
[<img src="" width="200"/>](https://github.com/unslothai/unsloth)
|
||
|
|
|
||
|
|
kshabana4ai
|
||
|
|
|
||
|
|
 width="200"/>](https://huggingface.co/kshabana-ai)
|
||
|
|
|