|
--- |
|
tags: |
|
- merge |
|
- mergekit |
|
- lazymergekit |
|
--- |
|
|
|
# GZA-14B-sce |
|
|
|
GZA-14B-sce is a merge of the following models: |
|
RZA & GZA Models provided me data to which model ranked to influence the merge. |
|
RZA blended somewhere between two models with different variations to answers questions in different formats. |
|
While GZA continued to use one type of formatting. |
|
https://huggingface.co/Quazim0t0/GZA-14B-sce |
|
|
|
## 🧩 Configuration |
|
|
|
```yaml |
|
models: |
|
# Pivot model |
|
- model: Quazim0t0/Phi4.Turn.R1Distill.16bit |
|
# Target models |
|
- model: Quazim0t0/ThinkPhi1.1-Tensors |
|
- model: Pinkstack/SuperThoughts-CoT-14B-16k-o1-QwQ |
|
- model: prithivMLmods/Phi-4-QwQ |
|
merge_method: sce |
|
base_model: Quazim0t0/Phi4.Turn.R1Distill.16bit |
|
parameters: |
|
select_topk: 1.0 |
|
dtype: bfloat16 |
|
``` |
|
|
|
## 💻 Usage |
|
|
|
```python |
|
!pip install -qU transformers accelerate |
|
|
|
from transformers import AutoTokenizer |
|
import transformers |
|
import torch |
|
|
|
model = "Quazim0t0/GZA-14B-sce" |
|
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"]) |
|
``` |
