README.md

---
base_model:
- microsoft/Phi-3.5-mini-instruct
- microsoft/Phi-3.5-mini-instruct
- microsoft/Phi-3.5-mini-instruct
tags:
- merge
- mergekit
- lazymergekit
- microsoft/Phi-3.5-mini-instruct
license: mit
---
Should ideally be used to further fine-tune. 
# phi-3.5-6b

phi-3.5-6b is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
* [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)
* [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)
* [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct)

## Config

```yaml
base_model: microsoft/Phi-3.5-mini-instruct
merge_method: passthrough
slices:
  - sources:
      - model: microsoft/Phi-3.5-mini-instruct
        layer_range: [0, 32]  # Base model layers
  - sources:
      - model: microsoft/Phi-3.5-mini-instruct
        layer_range: [16, 32]  # Add 16 layers
  - sources:
      - model: microsoft/Phi-3.5-mini-instruct
        layer_range: [24, 32]  # Add 8 more layers
tokenizer_source: microsoft/Phi-3.5-mini-instruct
dtype: float16
```

## Usage

```python
!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Pinkstack/phi-3.5-6b"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
```