Deepak Mishra

@nicopreme I was literally thinking about making something like this last night. Makes it easy to give feedback to the agent message generated in plan mode.

Added a fixed-position editor to Pi that stays put while the chat can stream and scroll above it. Also added configurable shortcuts for jumping between user/LLM messages, to the bottom of chat, and to the start/end of a multiline editor. pi install npm:pi-powerline-footer github.com/nicobailon/pi-…

@runwayml There seems to be a loophole. The AI ones generally load first!

We showed 1,000 people a series of videos. Some real and some generated with Gen-4.5. Over 90% of people couldn't tell the difference. Can you? Take the test at the link below.

@Helios_Movement @grok what should be the food intake for vegetarians

Here's the list of nutrients most people don't get enough of and experience: -Fatigue -High blood pressure -Skin issues -Low libido -Brain fog -High blood sugar -Hair loss -A compromised immune system and more, as a result. Thread🧵

@abhi1thakur Hello from Gurugram!

Catalyst is now open source!🎉 It’s surreal that something the entire @1mgOfficial Frontend team has contributed to over the last 4 years was rolled out to the public in its first beta release @react_india Docs: catalyst.1mg.com/public_docs/ CCA: npmjs.com/package/create…

Doing good science is 90% finding a science buddy to constantly talk to about the project.

@OfficialLoganK I've grown to like the editing experience in @obsdmd but if I were to say something specific, working with lists is terrible. Assuming the markdown mode, the editor should know that I am writing list items and entering a new line should automatically start a new list item.

@skilp4d Anything specific you want in the editor experience?

Hey @OfficialLoganK, any chance the team is working on improving the editor experience of aistudio? I love the fact that the prompts auto save now. As soon as the editor improves, I think aistudio might become my goto tool given its crazy long context length 🙌🏻

Hey @perplexity_ai @AravSrinivas, are there any plans to allow data export?

@yacineMTB Shots fired

if web devs were smart they wouldn't be web devs. they would be software engineers

Ball so hard...

New feature alert in the @huggingface ecosystem! Flash Attention 2 natively supported in huggingface transformers, supports training PEFT, and quantization (GPTQ, QLoRA, LLM.int8) First pip install flash attention and pass use_flash_attention_2=True when loading the model!

🔥 Excited to open-source the code of our Evol-Instruct！ 🏆It's one the most important techniques for the success of WizardLM, WizardCoder, and WizardMath. ❤️Thanks authors: @CanXu20 , @victorsungo_ai , @tangmensan. 🏇Let's build the Evol Universe Together! 💥Code Link: github.com/nlpxucan/Wizar…

#TIL that dropping that last batch from your dataloader is a wise idea. They are as bad as the Trojan horse. Took me hours to realise this mistake. But why is this the default value in #PyTorch :(

@rasbt @jeremyphoward @radekosmulski @HamelHusain @amin__asadi Agree with @rasbt on this. Could it be that the conversion was done keeping in mind a better support for float16 format ?

@jeremyphoward @radekosmulski @HamelHusain @amin__asadi Wouldn’t that be only a problem the other way around, going from float16 -> bfloat16? Since float16 supports more digits after the decimal points, going bfloat16 -> float16 there shouldn’t be no rounding?

What are you using for fine tuning OSS foundation models?

This is huge: Llama-v2 is open source, with a license that authorizes commercial use! This is going to change the landscape of the LLM market. Llama-v2 is available on Microsoft Azure and will be available on AWS, Hugging Face and other providers Pretrained and fine-tuned models are available with 7B, 13B and 70B parameters. Llama-2 website: ai.meta.com/llama/ Llama-2 paper: ai.meta.com/research/publi… A number of personalities from industry and academia have endorsed our open source approach: about.fb.com/news/2023/07/l…

@Sachin_kalsi Absolutely agree, it just makes an army of lazy programmers.

The invention of the try-catch block has unintentionally made programmers rely more on it & become less attentive and resourceful. If this feature hadn't been introduced, it's possible that we would have seen the emergence of even more skilled and capable #programmers

Continuing on my previous post, I have tried to explain the challenges of Ingesting data from IoT devices at Loconav. I have tried to keep things simple but might have lapsed a bit, let me know if you any feedback or questions around it. Happy Reading (: @vaibhav-vg/art-of-ingesting-data-at-loconav-94bdacc27d6f" target="_blank" rel="nofollow noopener">medium.com/@vaibhav-vg/ar…

Starting a series on what I've been doing for past 4 years. Feedback is always appreciated. link.medium.com/naFnuCeIaAb