solar-merge-v1.0 / README.md
jieunhan's picture
Upload folder using huggingface_hub
fa88d3d verified
metadata
license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - upstage/SOLAR-10.7B-Instruct-v1.0
  - heavytail/kullm-solar
base_model:
  - upstage/SOLAR-10.7B-Instruct-v1.0
  - heavytail/kullm-solar

solar-merge-v1.0

solar-merge-v1.0 is a Mixture of Experts (MoE) made with the following models using LazyMergekit:

๐Ÿงฉ Configuration

base_model: upstage/SOLAR-10.7B-v1.0
dtype: float16
experts:
  - source_model: upstage/SOLAR-10.7B-Instruct-v1.0
    positive_prompts: ["๋‹น์‹ ์€ ์นœ์ ˆํ•œ ๋ณดํŽธ์ ์ธ ์–ด์‹œ์Šคํ„ดํŠธ์ด๋‹ค."]
  - source_model: heavytail/kullm-solar
    positive_prompts: ["๋‹น์‹ ์€ ์นœ์ ˆํ•œ ์–ด์‹œ์Šคํ„ดํŠธ์ด๋‹ค."]
gate_mode: cheap_embed
tokenizer_source: base

๐Ÿ’ป Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "jieunhan/solar-merge-v1.0"

tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)

messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
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"])