Article
Train 400x faster Static Embedding Models with Sentence Transformers
•
101
Vikas Kumar
vikas
AI & ML interests
None yet
Recent Activity
upvoted
an
article
3 days ago
Train 400x faster Static Embedding Models with Sentence Transformers
liked
a model
about 2 months ago
alibaba-damo/mgp-str-base
liked
a Space
4 months ago
Tonic/GOT-OCR
Organizations
vikas's activity
upvoted
an
article
3 days ago
liked
a
model
about 2 months ago
liked
a
Space
4 months ago
liked
a
Space
5 months ago
upvoted
an
article
5 months ago
Article
The 5 Most Under-Rated Tools on Hugging Face
•
86
upvoted
a
paper
5 months ago
upvoted
a
paper
6 months ago
upvoted
an
article
6 months ago
Article
Finetuning PaliGemma with AutoTrain
By
•
•
9
upvoted
a
collection
6 months ago
upvoted
2
articles
6 months ago
Article
Fine-tune Llama 3.1 Ultra-Efficiently with Unsloth
By
•
•
264Article
Llama 3.1 - 405B, 70B & 8B with multilinguality and long context
•
226
liked
a
dataset
6 months ago
upvoted
a
collection
6 months ago
upvoted
3
articles
6 months ago
Article
SmolLM - blazingly fast and remarkably powerful
•
298
Article
The Rise of Agentic Data Generation
By
•
•
81Article
ColPali: Efficient Document Retrieval with Vision Language Models 👀
By
•
•
187
reacted to
fdaudens's
post with 👍
6 months ago
Post
3354
🧠 How to create more diverse, realistic synthetic AI training data?
@TencentAIGC-Lab AI Lab created @proj-persona , a vast collection of 1 billion diverse personas, to help create synthetic data with LLMs that encapsulate a wide array of perspectives, knowledge, experiences, interests, and professions.
These personas were created with automatically curated data, representing approximately 13% of the world’s total population.
💡 The authors argue that integrating a persona into data synthesis prompts effectively steers LLMs to adopt specific perspectives, creating unique and relevant synthetic data with minimal effort.
They showcased various practical applications of Persona Hub to demonstrate its effectiveness and versatility in various synthetic data creation scenarios: mathematical and logical reasoning problems, simulating diverse user requests and prompts for LLMs, generating informative and detailed text content across various topics, and more.
🚀 It's one of the trending datasets on Hugging Face. Digging into it is quite fun! I found one that reminds me of several people I know: "A journalist who covers technology and innovation in the print and digital media industries." It helped generate the prompt attached to this post (about which I'd be curious to know your answers 😉).
Synthetic data is a hot topic in AI. It will be interesting to see if this research could help make LLMs more robust, versatile, and capable of handling a wide array of real-world scenarios.
👉Explore the dataset: proj-persona/PersonaHub
👉 Read the paper: https://arxiv.org/pdf/2406.20094
@TencentAIGC-Lab AI Lab created @proj-persona , a vast collection of 1 billion diverse personas, to help create synthetic data with LLMs that encapsulate a wide array of perspectives, knowledge, experiences, interests, and professions.
These personas were created with automatically curated data, representing approximately 13% of the world’s total population.
💡 The authors argue that integrating a persona into data synthesis prompts effectively steers LLMs to adopt specific perspectives, creating unique and relevant synthetic data with minimal effort.
They showcased various practical applications of Persona Hub to demonstrate its effectiveness and versatility in various synthetic data creation scenarios: mathematical and logical reasoning problems, simulating diverse user requests and prompts for LLMs, generating informative and detailed text content across various topics, and more.
🚀 It's one of the trending datasets on Hugging Face. Digging into it is quite fun! I found one that reminds me of several people I know: "A journalist who covers technology and innovation in the print and digital media industries." It helped generate the prompt attached to this post (about which I'd be curious to know your answers 😉).
Synthetic data is a hot topic in AI. It will be interesting to see if this research could help make LLMs more robust, versatile, and capable of handling a wide array of real-world scenarios.
👉Explore the dataset: proj-persona/PersonaHub
👉 Read the paper: https://arxiv.org/pdf/2406.20094
reacted to
mrm8488's
post with ❤️
7 months ago
Post
4911
🚨Exciting news for the Multilingual Synthetic Data Community!🚨
I’ve taken inspiration from the MAGPIE paper on Llama-3-8B-instruct and extended its capabilities. Here’s what’s new!
🗞 The MAGPIE paper showcased that if you use the instruction-tuned version (
🤔 While reading a script by Sebastian Raschka, PhD, I wondered: Could these advancements be replicated in other languages? Specifically, could they benefit non-English datasets?
🎉 And the answer is YES! At least for Spanish. I've successfully adapted the techniques for Spanish, proving the model's flexibility and multilingual capabilities.
👩💻 To make this accessible, I created a basic script (heavily inspired by the Sebastian Raschka one) that allows you to generate similar datasets using
[Script](https://gist.github.com/mrm8488/4650a5e3cc45523798a527a3446eb312)
🔍 Explore the datasets 📚 generated using our new script!
- [Llama-3-8B](https://huggingface.co/datasets/mrm8488/dataset_llama3_5000_samples_es_4231_filtered)
- [Phi-3-medium](https://huggingface.co/datasets/mrm8488/dataset_phi3-medium_5000_samples_es_3906_filtered)
- [Phi-3-mini](https://huggingface.co/datasets/mrm8488/dataset_phi3_5000_samples_es_3282_filtered)
Note: These datasets have basic filtering. Apply additional quality filters before using them to fine-tune large language models.
Inspiration and base script:
https://github.com/rasbt/LLMs-from-scratch/blob/main/ch07/05_dataset-generation/llama3-ollama.ipynb
https://www.linkedin.com/feed/update/urn:li:activity:7210982019751661568/
I’ve taken inspiration from the MAGPIE paper on Llama-3-8B-instruct and extended its capabilities. Here’s what’s new!
🗞 The MAGPIE paper showcased that if you use the instruction-tuned version (
Llama-3-8B-instruct) to generate synthetic instructions and then fine-tune the base version (Llama-3-8B) on this dataset, you can improve even the it-tuned version🤔 While reading a script by Sebastian Raschka, PhD, I wondered: Could these advancements be replicated in other languages? Specifically, could they benefit non-English datasets?
🎉 And the answer is YES! At least for Spanish. I've successfully adapted the techniques for Spanish, proving the model's flexibility and multilingual capabilities.
👩💻 To make this accessible, I created a basic script (heavily inspired by the Sebastian Raschka one) that allows you to generate similar datasets using
ollama models (initially phi and llama3) automatically and upload it to the Hugging Face Hub
🔍 Explore the datasets 📚 generated using our new script!
- [Llama-3-8B](https://huggingface.co/datasets/mrm8488/dataset_llama3_5000_samples_es_4231_filtered)
- [Phi-3-medium](https://huggingface.co/datasets/mrm8488/dataset_phi3-medium_5000_samples_es_3906_filtered)
- [Phi-3-mini](https://huggingface.co/datasets/mrm8488/dataset_phi3_5000_samples_es_3282_filtered)
Note: These datasets have basic filtering. Apply additional quality filters before using them to fine-tune large language models.
Inspiration and base script:
https://github.com/rasbt/LLMs-from-scratch/blob/main/ch07/05_dataset-generation/llama3-ollama.ipynb
https://www.linkedin.com/feed/update/urn:li:activity:7210982019751661568/
- 7 replies
upvoted
a
collection
7 months ago
upvoted
an
article
7 months ago
Article
Fine-tuning Florence-2 - Microsoft's Cutting-edge Vision Language Models
•
183