Minueza-32M-Chat / README.md

Adding Evaluation Results (#2)

cca87ef verified 4 months ago

No virus

12 kB

	---
	language:
	- en
	license: apache-2.0
	datasets:
	- databricks/databricks-dolly-15k
	- Felladrin/ChatML-databricks-dolly-15k
	- euclaise/reddit-instruct-curated
	- Felladrin/ChatML-reddit-instruct-curated
	- THUDM/webglm-qa
	- Felladrin/ChatML-WebGLM-QA
	- starfishmedical/webGPT_x_dolly
	- Felladrin/ChatML-webGPT_x_dolly
	- LDJnr/Capybara
	- Felladrin/ChatML-Capybara
	- Open-Orca/SlimOrca-Dedup
	- Felladrin/ChatML-SlimOrca-Dedup
	- HuggingFaceH4/ultrachat_200k
	- Felladrin/ChatML-ultrachat_200k
	- nvidia/HelpSteer
	- Felladrin/ChatML-HelpSteer
	- sablo/oasst2_curated
	- Felladrin/ChatML-oasst2_curated
	- CohereForAI/aya_dataset
	- Felladrin/ChatML-aya_dataset
	- argilla/distilabel-capybara-dpo-7k-binarized
	- Felladrin/ChatML-distilabel-capybara-dpo-7k-binarized
	- argilla/distilabel-intel-orca-dpo-pairs
	- Felladrin/ChatML-distilabel-intel-orca-dpo-pairs
	- argilla/ultrafeedback-binarized-preferences
	- Felladrin/ChatML-ultrafeedback-binarized-preferences
	- sablo/oasst2_dpo_pairs_en
	- Felladrin/ChatML-oasst2_dpo_pairs_en
	- NeuralNovel/Neural-DPO
	- Felladrin/ChatML-Neural-DPO
	base_model: Felladrin/Minueza-32M-Base
	pipeline_tag: text-generation
	widget:
	- messages:
	- role: system
	content: You are a career counselor. The user will provide you with an individual
	looking for guidance in their professional life, and your task is to assist
	them in determining what careers they are most suited for based on their skills,
	interests, and experience. You should also conduct research into the various
	options available, explain the job market trends in different industries, and
	advice on which qualifications would be beneficial for pursuing particular fields.
	- role: user
	content: Heya!
	- role: assistant
	content: Hi! How may I help you?
	- role: user
	content: I am interested in developing a career in software engineering. What
	would you recommend me to do?
	- messages:
	- role: system
	content: You are a highly knowledgeable assistant. Help the user as much as you
	can.
	- role: user
	content: How can I become a healthier person?
	- messages:
	- role: system
	content: You are a helpful assistant who gives creative responses.
	- role: user
	content: Write the specs of a game about mages in a fantasy world.
	- messages:
	- role: system
	content: You are a helpful assistant who answers user's questions with details.
	- role: user
	content: Tell me about the pros and cons of social media.
	- messages:
	- role: system
	content: You are a helpful assistant who answers user's questions with details
	and curiosity.
	- role: user
	content: What are some potential applications for quantum computing?
	inference:
	parameters:
	max_new_tokens: 250
	do_sample: true
	temperature: 0.65
	top_p: 0.55
	top_k: 35
	repetition_penalty: 1.176
	model-index:
	- name: Minueza-32M-Chat
	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: 20.39
	name: normalized accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Minueza-32M-Chat
	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: 26.54
	name: normalized accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Minueza-32M-Chat
	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: 25.75
	name: accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Minueza-32M-Chat
	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: 47.27
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Minueza-32M-Chat
	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: 50.99
	name: accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Minueza-32M-Chat
	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: 0.0
	name: accuracy
	source:
	url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Minueza-32M-Chat
	name: Open LLM Leaderboard
	---

	# Minueza-32M-Chat: A chat model with 32 million parameters

	- Base model: [Felladrin/Minueza-32M-Base](https://huggingface.co/Felladrin/Minueza-32M-Base)
	- Datasets used during SFT:
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-databricks-dolly-15k)] [databricks/databricks-dolly-15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-reddit-instruct-curated)] [euclaise/reddit-instruct-curated](https://huggingface.co/datasets/euclaise/reddit-instruct-curated)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-WebGLM-QA)] [THUDM/webglm-qa](https://huggingface.co/datasets/THUDM/webglm-qa)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-webGPT_x_dolly)] [starfishmedical/webGPT_x_dolly](https://huggingface.co/datasets/starfishmedical/webGPT_x_dolly)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-Capybara)] [LDJnr/Capybara](https://huggingface.co/datasets/LDJnr/Capybara)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-SlimOrca-Dedup)] [Open-Orca/SlimOrca-Dedup](https://huggingface.co/datasets/Open-Orca/SlimOrca-Dedup)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-ultrachat_200k)] [HuggingFaceH4/ultrachat_200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-HelpSteer)] [nvidia/HelpSteer](https://huggingface.co/datasets/nvidia/HelpSteer)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-oasst2_curated)] [sablo/oasst2_curated](https://huggingface.co/datasets/sablo/oasst2_curated)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-aya_dataset)] [CohereForAI/aya_dataset](https://huggingface.co/datasets/CohereForAI/aya_dataset)
	- Datasets used during DPO:
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-distilabel-capybara-dpo-7k-binarized)] [argilla/distilabel-capybara-dpo-7k-binarized](https://huggingface.co/datasets/argilla/distilabel-capybara-dpo-7k-binarized)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-distilabel-intel-orca-dpo-pairs)] [argilla/distilabel-intel-orca-dpo-pairs](https://huggingface.co/datasets/argilla/distilabel-intel-orca-dpo-pairs)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-ultrafeedback-binarized-preferences)] [argilla/ultrafeedback-binarized-preferences](https://huggingface.co/datasets/argilla/ultrafeedback-binarized-preferences)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-oasst2_dpo_pairs_en)] [sablo/oasst2_dpo_pairs_en](https://huggingface.co/datasets/sablo/oasst2_dpo_pairs_en)
	- [[ChatML](https://huggingface.co/datasets/Felladrin/ChatML-Neural-DPO)] [NeuralNovel/Neural-DPO](https://huggingface.co/datasets/NeuralNovel/Neural-DPO)
	- License: [Apache License 2.0](https://huggingface.co/Felladrin/Minueza-32M-Chat/resolve/main/license.txt)
	- Availability in other ML formats:
	- GGUF: [Felladrin/gguf-Minueza-32M-Chat](https://huggingface.co/Felladrin/gguf-Minueza-32M-Chat)
	- ONNX: [Felladrin/onnx-Minueza-32M-Chat](https://huggingface.co/Felladrin/onnx-Minueza-32M-Chat)

	## Recommended Prompt Format

	```
	<\|im_start\|>system
	{system_message}<\|im_end\|>
	<\|im_start\|>user
	{user_message}<\|im_end\|>
	<\|im_start\|>assistant
	```

	## Recommended Inference Parameters

	```yml
	do_sample: true
	temperature: 0.65
	top_p: 0.55
	top_k: 35
	repetition_penalty: 1.176
	```

	## Usage Example

	```python
	from transformers import pipeline

	generate = pipeline("text-generation", "Felladrin/Minueza-32M-Chat")

	messages = [
	{
	"role": "system",
	"content": "You are a helpful assistant who answers the user's questions with details and curiosity.",
	},
	{
	"role": "user",
	"content": "What are some potential applications for quantum computing?",
	},
	]

	prompt = generate.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

	output = generate(
	prompt,
	max_new_tokens=256,
	do_sample=True,
	temperature=0.65,
	top_k=35,
	top_p=0.55,
	repetition_penalty=1.176,
	)

	print(output[0]["generated_text"])
	```

	## How it was trained

	This model was trained with [SFT Trainer](https://huggingface.co/docs/trl/main/en/sft_trainer) and [DPO Trainer](https://huggingface.co/docs/trl/main/en/dpo_trainer), in several sessions, using the following settings:

	For Supervised Fine-Tuning:

	\| Hyperparameter \| Value \|
	\| :-------------------------- \| :-------------------------------------------- \|
	\| learning_rate \| 2e-5 \|
	\| total_train_batch_size \| 24 \|
	\| max_seq_length \| 2048 \|
	\| weight_decay \| 0 \|
	\| warmup_ratio \| 0.02 \|

	For Direct Preference Optimization:

	\| Hyperparameter \| Value \|
	\| :-------------------------- \| :-------------------------------------------- \|
	\| learning_rate \| 7.5e-7 \|
	\| total_train_batch_size \| 6 \|
	\| max_length \| 2048 \|
	\| max_prompt_length \| 1536 \|
	\| max_steps \| 200 \|
	\| weight_decay \| 0 \|
	\| warmup_ratio \| 0.02 \|
	\| beta \| 0.1 \|

	## [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_Felladrin__Minueza-32M-Chat)

	\| Metric \|Value\|
	\|---------------------------------\|----:\|
	\|Avg. \|28.49\|
	\|AI2 Reasoning Challenge (25-Shot)\|20.39\|
	\|HellaSwag (10-Shot) \|26.54\|
	\|MMLU (5-Shot) \|25.75\|
	\|TruthfulQA (0-shot) \|47.27\|
	\|Winogrande (5-shot) \|50.99\|
	\|GSM8k (5-shot) \| 0.00\|