70B is general, this one is specifically trained for code tasks. With a smaller model, you can do better on a smaller task scope if you have a ton of great data.
MACHINE LEARNING
-
Mistral AI dominates HF trending with 7B Instruct and Codestral
By
–
.
@MistralAI holds 3 spots in the top 10 trending models on HF including #1 with 7B Instruct. Congrats for the launch of Codestral, I'm sure it'll get #1 trending very soon! https://
huggingface.co/models -
Fine-tuning LLMs for 2x Faster Inference with Speculative Decoding
By
–
Fine-tuning #LLMs isn't just for customization You can also #finetune LLMs to increase throughput using speculative decoding Check out the replay of our recent tech talk to learn how we fine-tuned an LLM using to increase inference by 2x https://
pbase.ai/4bV6rNc. -
Scale launches SEAL Leaderboards for frontier LLM evaluation
By
–
New from Scale: SEAL Leaderboards — a new benchmark arena for frontier LLMs – Private, novel assessments that models can’t train on
– ELO-scale rankings (via Bradley-Terry)
– Domain leaderboards (today: coding, math, instruct, Spanish — more soon!) (Links in reply) -
Contributing sampling functions to Keras random module
By
–
Do you want to open a PR? You can definitely add more sampling functions to `keras.random`
-
LLM Evaluation Improvements and the Challenge of Creating Good Benchmarks
By
–
Nice, a serious contender to @lmsysorg in evaluating LLMs has entered the chat. LLM evals are improving, but not so long ago their state was very bleak, with qualitative experience very often disagreeing with quantitative rankings. This is because good evals are very difficult
-
Programmatic Labeling Accelerates AI Development From Months to Days
By
–
Data labeling taking too long? SMEs and data scientists need to transition from manual to programmatic labeling. See how these teams have taken AI development time from months to days! Don't miss out on Snorkel’s upcoming webinar; register now
-
Optimizing batch size and throughput for GPT-3 training
By
–
Okay that's good to know 🙂
I was just following the GPT-3 paper numbers in the table, but like I mentioned it's possible the settings are way too conservative.
Few more things to try: we want to increase the batch size as much as possible to get higher tok/s. Are you using -r 1 -
Leading AI Models Evaluated: Coding, Math, Multilingual Performance
By
–
3/ We eval'd many of the leading models: – GPT-4o
– GPT-4 Turbo
– Claude 3 Opus
– Gemini 1.5 Pro
– Gemini 1.5 Flash
– Llama3
– Mistral Large On Coding, Math, Instruction Following, and Multilinguality (Spanish). See leaderboard results below. -
Scale Launches SEAL Leaderboards for Frontier Model Evaluation
By
–
1/ We are launching SEAL Leaderboards—private, expert evaluations of leading frontier models. Our design principles:
Private + Unexploitable. No overfitting on evals!
Domain Expert Evals
Continuously Updated w/new Data and Models Read more in http://
scale.com/leaderboard