π» Merges
Collection
Playing around with different LLM Merges (Slerp, Linear, Ties, Dare, Task Arithmetic).
β’
4 items
β’
Updated
DareVox-7B is a merge of the following models:
models:
- model: mistralai/Mistral-7B-v0.1
# No parameters necessary for base model
- model: teknium/OpenHermes-2.5-Mistral-7B
parameters:
density: 0.53
weight: 0.4
- model: abacusai/Slerp-CM-mist-dpo
parameters:
density: 0.53
weight: 0.3
- model: berkeley-nest/Starling-LM-7B-alpha
parameters:
density: 0.5
weight: 0.4
merge_method: dare_ties
base_model: mistralai/Mistral-7B-v0.1
parameters:
int8_mask: true
dtype: bfloat16
!pip install -qU transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "abideen/DareVox-7B"
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"])