FrankenVillain-7B-v1
FrankenVillain-7B-v1 is a Franken merge of the following models using [mergekit](https://github.com/cg123/mergekit
🧩 Configuration
slices:
- sources:
- model: jeonsworld/CarbonVillain-en-10.7B-v1
layer_range: [0, 24]
- sources:
- model: jeonsworld/CarbonVillain-en-10.7B-v1
layer_range: [8, 32]
merge_method: passthrough
dtype: bfloat16
💻 Usage
!pip install -qU transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "luqmanxyz/FrankenVillain-7B-v1"
messages = [{"role": "user", "content": "What are the 3 planets closest to the sun"}]
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"])
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
Metric | Value |
---|---|
Avg. | 45.34 |
AI2 Reasoning Challenge (25-Shot) | 42.75 |
HellaSwag (10-Shot) | 51.52 |
MMLU (5-Shot) | 48.60 |
TruthfulQA (0-shot) | 56.19 |
Winogrande (5-shot) | 73.01 |
GSM8k (5-shot) | 0.00 |
- Downloads last month
- 61
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for luqmanxyz/FrankenVillain-7B-v1
Base model
jeonsworld/CarbonVillain-en-10.7B-v1
Finetuned
this model
Evaluation results
- normalized accuracy on AI2 Reasoning Challenge (25-Shot)test set Open LLM Leaderboard42.750
- normalized accuracy on HellaSwag (10-Shot)validation set Open LLM Leaderboard51.520
- accuracy on MMLU (5-Shot)test set Open LLM Leaderboard48.600
- mc2 on TruthfulQA (0-shot)validation set Open LLM Leaderboard56.190
- accuracy on Winogrande (5-shot)validation set Open LLM Leaderboard73.010
- accuracy on GSM8k (5-shot)test set Open LLM Leaderboard0.000