Expert Support case study: Bolstering a RAG app with LLM-as-a-Judge
•
25
Aymeric Roucher
m-ric
AI & ML interests
Leading Agents at Hugging Face 🤗
Recent Activity
updated
a dataset
2 days ago
huggingface/documentation-images
updated
a dataset
2 days ago
m-ric/agents_medium_benchmark
liked
a Space
6 days ago
data-agents/jupyter-agent
Articles
Organizations
Posts
80
Post
1627
After 6 years, BERT, the workhorse of encoder models, finally gets a replacement: 𝗪𝗲𝗹𝗰𝗼𝗺𝗲 𝗠𝗼𝗱𝗲𝗿𝗻𝗕𝗘𝗥𝗧! 🤗
We talk a lot about ✨Generative AI✨, meaning "Decoder version of the Transformers architecture", but this is only one of the ways to build LLMs: encoder models, that turn a sentence in a vector, are maybe even more widely used in industry than generative models.
The workhorse for this category has been BERT since its release in 2018 (that's prehistory for LLMs).
It's not a fancy 100B parameters supermodel (just a few hundred millions), but it's an excellent workhorse, kind of a Honda Civic for LLMs.
Many applications use BERT-family models - the top models in this category cumulate millions of downloads on the Hub.
➡️ Now a collaboration between Answer.AI and LightOn just introduced BERT's replacement: ModernBERT.
𝗧𝗟;𝗗𝗥:
🏛️ Architecture changes:
⇒ First, standard modernizations:
- Rotary positional embeddings (RoPE)
- Replace GeLU with GeGLU,
- Use Flash Attention 2
✨ The team also introduced innovative techniques like alternating attention instead of full attention, and sequence packing to get rid of padding overhead.
🥇 As a result, the model tops the game of encoder models:
It beats previous standard DeBERTaV3 for 1/5th the memory footprint, and runs 4x faster!
Read the blog post 👉 https://huggingface.co./blog/modernbert
We talk a lot about ✨Generative AI✨, meaning "Decoder version of the Transformers architecture", but this is only one of the ways to build LLMs: encoder models, that turn a sentence in a vector, are maybe even more widely used in industry than generative models.
The workhorse for this category has been BERT since its release in 2018 (that's prehistory for LLMs).
It's not a fancy 100B parameters supermodel (just a few hundred millions), but it's an excellent workhorse, kind of a Honda Civic for LLMs.
Many applications use BERT-family models - the top models in this category cumulate millions of downloads on the Hub.
➡️ Now a collaboration between Answer.AI and LightOn just introduced BERT's replacement: ModernBERT.
𝗧𝗟;𝗗𝗥:
🏛️ Architecture changes:
⇒ First, standard modernizations:
- Rotary positional embeddings (RoPE)
- Replace GeLU with GeGLU,
- Use Flash Attention 2
✨ The team also introduced innovative techniques like alternating attention instead of full attention, and sequence packing to get rid of padding overhead.
🥇 As a result, the model tops the game of encoder models:
It beats previous standard DeBERTaV3 for 1/5th the memory footprint, and runs 4x faster!
Read the blog post 👉 https://huggingface.co./blog/modernbert
Post
2024
𝐇𝐮𝐠𝐠𝐢𝐧𝐠 𝐅𝐚𝐜𝐞 𝐫𝐞𝐥𝐞𝐚𝐬𝐞𝐬 𝐏𝐢𝐜𝐨𝐭𝐫𝐨𝐧, 𝐚 𝐦𝐢𝐜𝐫𝐨𝐬𝐜𝐨𝐩𝐢𝐜 𝐥𝐢𝐛 𝐭𝐡𝐚𝐭 𝐬𝐨𝐥𝐯𝐞𝐬 𝐋𝐋𝐌 𝐭𝐫𝐚𝐢𝐧𝐢𝐧𝐠 𝟒𝐃 𝐩𝐚𝐫𝐚𝐥𝐥𝐞𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 🥳
🕰️ Llama-3.1-405B took 39 million GPU-hours to train, i.e. about 4.5 thousand years.
👴🏻 If they had needed all this time, we would have GPU stories from the time of Pharaoh 𓂀: "Alas, Lord of Two Lands, the shipment of counting-stones arriving from Cathay was lost to pirates, this shall delay the building of your computing temple by many moons "
🛠️ But instead, they just parallelized the training on 24k H100s, which made it take just a few months.
This required parallelizing across 4 dimensions: data, tensor, context, pipeline.
And it is infamously hard to do, making for bloated code repos that hold together only by magic.
🤏 𝗕𝘂𝘁 𝗻𝗼𝘄 𝘄𝗲 𝗱𝗼𝗻'𝘁 𝗻𝗲𝗲𝗱 𝗵𝘂𝗴𝗲 𝗿𝗲𝗽𝗼𝘀 𝗮𝗻𝘆𝗺𝗼𝗿𝗲! Instead of building mega-training codes, Hugging Face colleagues cooked in the other direction, towards tiny 4D parallelism libs. A team has built Nanotron, already widely used in industry.
And now a team releases Picotron, a radical approach to code 4D Parallelism in just a few hundred lines of code, a real engineering prowess, making it much easier to understand what's actually happening!
⚡ 𝗜𝘁'𝘀 𝘁𝗶𝗻𝘆, 𝘆𝗲𝘁 𝗽𝗼𝘄𝗲𝗿𝗳𝘂𝗹:
Counting in MFU (Model FLOPs Utilization, how much the model actually uses all the compute potential), this lib reaches ~50% on SmolLM-1.7B model with 8 H100 GPUs, which is really close to what huge libs would reach. (Caution: the team is leading further benchmarks to verify this)
Go take a look 👉 https://github.com/huggingface/picotron/tree/main/picotron
🕰️ Llama-3.1-405B took 39 million GPU-hours to train, i.e. about 4.5 thousand years.
👴🏻 If they had needed all this time, we would have GPU stories from the time of Pharaoh 𓂀: "Alas, Lord of Two Lands, the shipment of counting-stones arriving from Cathay was lost to pirates, this shall delay the building of your computing temple by many moons "
🛠️ But instead, they just parallelized the training on 24k H100s, which made it take just a few months.
This required parallelizing across 4 dimensions: data, tensor, context, pipeline.
And it is infamously hard to do, making for bloated code repos that hold together only by magic.
🤏 𝗕𝘂𝘁 𝗻𝗼𝘄 𝘄𝗲 𝗱𝗼𝗻'𝘁 𝗻𝗲𝗲𝗱 𝗵𝘂𝗴𝗲 𝗿𝗲𝗽𝗼𝘀 𝗮𝗻𝘆𝗺𝗼𝗿𝗲! Instead of building mega-training codes, Hugging Face colleagues cooked in the other direction, towards tiny 4D parallelism libs. A team has built Nanotron, already widely used in industry.
And now a team releases Picotron, a radical approach to code 4D Parallelism in just a few hundred lines of code, a real engineering prowess, making it much easier to understand what's actually happening!
⚡ 𝗜𝘁'𝘀 𝘁𝗶𝗻𝘆, 𝘆𝗲𝘁 𝗽𝗼𝘄𝗲𝗿𝗳𝘂𝗹:
Counting in MFU (Model FLOPs Utilization, how much the model actually uses all the compute potential), this lib reaches ~50% on SmolLM-1.7B model with 8 H100 GPUs, which is really close to what huge libs would reach. (Caution: the team is leading further benchmarks to verify this)
Go take a look 👉 https://github.com/huggingface/picotron/tree/main/picotron
Collections
14
-
Scaling Autoregressive Models for Content-Rich Text-to-Image Generation
Paper • 2206.10789 • Published • 4 -
Beyond Chinchilla-Optimal: Accounting for Inference in Language Model Scaling Laws
Paper • 2401.00448 • Published • 28 -
Training Compute-Optimal Large Language Models
Paper • 2203.15556 • Published • 10 -
Scaling Laws for Neural Language Models
Paper • 2001.08361 • Published • 6
spaces
26
pinned
Running
67
🗺️🏕️
AI Travel Planner
Plan your next vacation with the help of an AI!
pinned
Running
191
🍫👀
Chunk Visualizer
Pick a text splitter => visualize chunks. Great for RAG.
pinned
Running
on
Zero
121
✍👀
Beam Search Visualizer
View how beam search decoding works, in detail!
pinned
Running
102
🤔📊
Agent Data Analyst
Need to analyze data? Let a Llama-3.1 agent do it for you!
Build error
📚
Test
Build error
🏢
Hf Model Downloads
models
3
datasets
15
m-ric/agents_medium_benchmark
Viewer
•
Updated
•
172
•
72
•
3
m-ric/newsletter
Updated
•
36
m-ric/songs_july_24
Viewer
•
Updated
•
85.6k
•
39
m-ric/huggingface_doc_qa_eval
Viewer
•
Updated
•
65
•
342
•
8
m-ric/songs
Preview
•
Updated
•
13
m-ric/rag_scores_cookbook
Viewer
•
Updated
•
5
•
71
m-ric/images
Viewer
•
Updated
•
1
•
9
m-ric/agents_small_benchmark
Viewer
•
Updated
•
100
•
48
•
8
m-ric/huggingface_doc
Viewer
•
Updated
•
2.65k
•
2.25k
•
10
m-ric/transformers_documentation_en
Viewer
•
Updated
•
359
•
55