|
|
--- |
|
|
license: apache-2.0 |
|
|
language: |
|
|
- en |
|
|
library_name: transformers |
|
|
pipeline_tag: text-generation |
|
|
--- |
|
|
|
|
|
# MulitLoRA-Mistral-Merging |
|
|
 |
|
|
|
|
|
MultiLoRA-Mistral-Merge is a MultiLoRA Ties Merge made with the following adapters using [🧜 AutoLoRAMerging](https://colab.research.google.com/drive/1cEj5p42NZ6Vg2HVYEGL2IM6n0G0gwvQU?usp=sharing): |
|
|
* [Yhyu13/dolphin-2.6-mistral-7b-dpo-laser-function-calling-lora](https://huggingface.co/Yhyu13/dolphin-2.6-mistral-7b-dpo-laser-function-calling-lora) |
|
|
* [predibase/legal](https://huggingface.co/predibase/legal) |
|
|
* [predibase/wikisql](https://huggingface.co/predibase/wikisql) |
|
|
|
|
|
The merged adapter can generate SQL statements, give legal advices, and perform function calling. |
|
|
|
|
|
## 🧩 Configuration |
|
|
```yaml |
|
|
density: 0.2 |
|
|
merging_type: "ties" |
|
|
weights: [2.0, 0.3, 0.7] |
|
|
``` |
|
|
|
|
|
## 💻 Usage |
|
|
|
|
|
```python |
|
|
!pip install -qU transformers bitsandbytes accelerate peft |
|
|
|
|
|
from peft import PeftConfig, PeftModel |
|
|
from peft import PeftModel, PeftConfig |
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
import torch |
|
|
|
|
|
peft_model = "abideen/MulitLoRA-Mistral-Merging" |
|
|
config = PeftConfig.from_pretrained(peft_model) |
|
|
model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, device_map="auto") |
|
|
tokenizer = AutoTokenizer.from_pretrained(peft_model) |
|
|
model.resize_token_embeddings(len(tokenizer)) |
|
|
model = PeftModel.from_pretrained(model, peft_model) |
|
|
|
|
|
prompt = "Table: Sports; Columns: ['Team', 'Head Coach', 'President', 'Home Ground', 'Location'] Natural Query: Who is the Head Coach of the team whose President is Mario Volarevic? SQL Query:" # @param {type:"string"} |
|
|
messages = [ |
|
|
{"role": "user", "content": prompt}, |
|
|
] |
|
|
text = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False) |
|
|
inputs = tokenizer(text, return_tensors="pt") # , add_special_tokens=False) |
|
|
inputs = {k: v for k, v in inputs.items()} |
|
|
outputs = model.generate( |
|
|
**inputs, |
|
|
max_new_tokens=256, |
|
|
do_sample=True, |
|
|
top_p=0.95, |
|
|
temperature=0.2, |
|
|
repetition_penalty=1.2, |
|
|
eos_token_id=tokenizer.eos_token_id, |
|
|
) |
|
|
print(tokenizer.decode(outputs[0])) |
|
|
``` |
|
|
|
