ogno-monarch-jaskier-merge-7b / README.md

Adding Evaluation Results (#3)

57e9c59 verified 4 months ago

No virus

5.61 kB

	---
	language:
	- en
	license: cc-by-nc-4.0
	library_name: diffusers
	tags:
	- merge
	- mergekit
	- lazymergekit
	- mlabonne/Monarch-7B
	- paulml/OGNO-7B
	- bardsai/jaskier-7b-dpo-v5.6
	base_model:
	- mlabonne/Monarch-7B
	- paulml/OGNO-7B
	- bardsai/jaskier-7b-dpo-v5.6
	model-index:
	- name: ogno-monarch-jaskier-merge-7b
	results:
	- task:
	type: text-generation
	name: Text Generation
	dataset:
	name: AI2 Reasoning Challenge (25-Shot)
	type: ai2_arc
	config: ARC-Challenge
	split: test
	args:
	num_few_shot: 25
	metrics:
	- type: acc_norm
	value: 73.04
	name: normalized accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/ogno-monarch-jaskier-merge-7b
	name: Open LLM Leaderboard
	- task:
	type: text-generation
	name: Text Generation
	dataset:
	name: HellaSwag (10-Shot)
	type: hellaswag
	split: validation
	args:
	num_few_shot: 10
	metrics:
	- type: acc_norm
	value: 89.09
	name: normalized accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/ogno-monarch-jaskier-merge-7b
	name: Open LLM Leaderboard
	- task:
	type: text-generation
	name: Text Generation
	dataset:
	name: MMLU (5-Shot)
	type: cais/mmlu
	config: all
	split: test
	args:
	num_few_shot: 5
	metrics:
	- type: acc
	value: 64.78
	name: accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/ogno-monarch-jaskier-merge-7b
	name: Open LLM Leaderboard
	- task:
	type: text-generation
	name: Text Generation
	dataset:
	name: TruthfulQA (0-shot)
	type: truthful_qa
	config: multiple_choice
	split: validation
	args:
	num_few_shot: 0
	metrics:
	- type: mc2
	value: 77.44
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/ogno-monarch-jaskier-merge-7b
	name: Open LLM Leaderboard
	- task:
	type: text-generation
	name: Text Generation
	dataset:
	name: Winogrande (5-shot)
	type: winogrande
	config: winogrande_xl
	split: validation
	args:
	num_few_shot: 5
	metrics:
	- type: acc
	value: 84.77
	name: accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/ogno-monarch-jaskier-merge-7b
	name: Open LLM Leaderboard
	- task:
	type: text-generation
	name: Text Generation
	dataset:
	name: GSM8k (5-shot)
	type: gsm8k
	config: main
	split: test
	args:
	num_few_shot: 5
	metrics:
	- type: acc
	value: 69.45
	name: accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/ogno-monarch-jaskier-merge-7b
	name: Open LLM Leaderboard
	---

	# DPO Fine-tuned version

	I DPO finetuned this model later to obtain a slightly better model (open llm leaderboard benchmark performance) https://huggingface.co/eren23/ogno-monarch-jaskier-merge-7b-OH-PREF-DPO

	# ogno-monarch-jaskier-merge-7b

	ogno-monarch-jaskier-merge-7b is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
	* [mlabonne/Monarch-7B](https://huggingface.co/mlabonne/Monarch-7B)
	* [paulml/OGNO-7B](https://huggingface.co/paulml/OGNO-7B)
	* [bardsai/jaskier-7b-dpo-v5.6](https://huggingface.co/bardsai/jaskier-7b-dpo-v5.6)

	## 🧩 Configuration

	```yaml
	models:
	- model: eren23/dpo-binarized-NeutrixOmnibe-7B
	# No parameters necessary for base model
	- model: mlabonne/Monarch-7B
	#Emphasize the beginning of Vicuna format models
	parameters:
	weight: 0.6
	density: 0.59
	- model: paulml/OGNO-7B
	parameters:
	weight: 0.1
	density: 0.55
	# Vicuna format
	- model: bardsai/jaskier-7b-dpo-v5.6
	parameters:
	weight: 0.3
	density: 0.55

	merge_method: dare_ties
	base_model: eren23/dpo-binarized-NeutrixOmnibe-7B
	parameters:
	int8_mask: true
	dtype: bfloat16
	random_seed: 0
	```

	## 💻 Usage

	```python
	!pip install -qU transformers accelerate

	from transformers import AutoTokenizer
	import transformers
	import torch

	model = "eren23/ogno-monarch-jaskier-merge-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"])
	```

	GGUF Version: https://huggingface.co/eren23/ogno-monarch-jaskier-merge-7b-GGUF
	# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
	Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_eren23__ogno-monarch-jaskier-merge-7b)

	\| Metric \|Value\|
	\|---------------------------------\|----:\|
	\|Avg. \|76.43\|
	\|AI2 Reasoning Challenge (25-Shot)\|73.04\|
	\|HellaSwag (10-Shot) \|89.09\|
	\|MMLU (5-Shot) \|64.78\|
	\|TruthfulQA (0-shot) \|77.44\|
	\|Winogrande (5-shot) \|84.77\|
	\|GSM8k (5-shot) \|69.45\|