Don't miss our VP, Compiler Software, Andrew Ling, speaking at the 8th Annual Toronto Machine Learning Summit (TMLS) about modifying PyTorch models to enable custom data-types and persist precision information through Groq Compiler on July 15 at 11:30 ET. https://hubs.la/Q02BZPB60
Groq’s Post
More Relevant Posts
-
Lemme know what's the output. the GPT is telling it's 40 & the compiler is telling it's 10!! let mul=function(x,y){ console.log(x*y); } let kul=mul.bind(4,5); kul(2);
To view or add a comment, sign in
-
Chapter 5: Intro To Variables🔍Dive into the world of reverse engineering! Learn how variables are stored in memory, initialized, and manipulated in code. Unravel the mysteries of data allocation and compiler signals. 💻 #ReverseEngineering #Coding101 https://lnkd.in/enM_-gnM
To view or add a comment, sign in
-
-
Lately, I've been trying to create a compiler. Not for any use case, mainly just for fun. Currently, I'm working on the Scanner. Instead of using something like Flex to generate Tokens, I thought it would be a good exercise to create something similar to Flex from scratch. Man oh man did I not realize how much theory Flex was covering up! From Parsing a regular expression via Recursive Descent parsing to constructing a non-deterministic finite automata through Thompson's construction all the way to finally generating a deterministic finite automata through the Powerset Construction. Makes me appreciate libraries that abstract this stuff away a TON more 🤓
To view or add a comment, sign in
-
Haven’t had a chance to test much aside from the Hugging Face space (link below), but the open weight model #DBRX (though commercial viability comes with many disclaimers) from Databricks and Mosaic ML team looks to be a great advancement in all the best practices discovered in open source source research. Crazy fast. And trained with an awesome mix of data it seems. Even better, for Mac silicon users via #MLX, you can now quant it and run it with 70gb of ram. https://lnkd.in/guRnnBbG Space to play with the model here: https://lnkd.in/gnA-h9iA
To view or add a comment, sign in
-
Spent so much time and energy in an attempt to implement a fairly simple contrastive setup because kept failing to follow every single pattern per documentation of these high-level off-the-shelf DL packages. Eventually made it in a couple of hours via plain Pytorch 🤧 Never have I truly appreciated the era of low-level DL (RIP #mxnet 🐐) w/o infinite future warnings, unresolvable dependency conflicts, "batteries" blowing up after the slightest tweak, undocumented objects accepting 50 parameters... Maybe it wasn't that bad to casually face papers w/o code back then when you had to quickly set up rather simple, opinionated ad hoc implementations from scratch instead of spending hours depth-first reading the docs and source code to finally run a slightly modified 5-liner. #showerthoughts
To view or add a comment, sign in
-
AI/ML Engineer ll AWS AI&ML Scholar (Among top 500 worldwide) || Xylem Grand Prize Winner 2023 (University Track)
PyTorch Internals! A comprehensive guide to how you can understand the PyTorch codebase & contribute to it. Here's the link: https://lnkd.in/dnDsJTQG
To view or add a comment, sign in
-
PyTorch 2.0 succeeds its predecessor, PyTorch 1.x, with improved performance and efficiency. Here’s a rundown of these new features and capabilities: Backward Compatibility: - Code, model and API compatibility with PyTorch 1.x - Maintained compatibility with libraries and dependencies Fusing Operations: - Fuses multiple operations into a single GPU kernel - Reduces kernel launching overhead Reducing Type Conversion Overhead: - Identifies and minimizes unnecessary type conversion code - Avoids redundancies Reusing Buffers: - Reduced memory allocation - Lower memory consumption during model training Read on to learn about all the improvements PyTorch 2.0 brings and how they work: https://bit.ly/47QsWR2
To view or add a comment, sign in
-
-
The little details add up to making a presentable product; later will code for default behavior to be std::cin (terminal) input; already have used the class destructor of "verbosityfeature" to automatically execute verbosityfeature if it wasn't run earlier... that's philosophy, as to whether this is a good use of a destructor. So: one gets a standard readout. Made "-i" to automatically do automorphisms when only one graph is on the workspace. Made it possible to list any number of input filenames after the "-d" switch. Switched verbosityfeature to take first an output filename, and secondly to take the encoded switches, with the default being a compiler define that takes all possible verbosities together. Very close to having a VERBOSITY_MINIMAL that will make a pleasantly readable output on several dozen input files.
To view or add a comment, sign in
-
https://lnkd.in/dyhd4vjb Recursion is very useful but sometimes challenging. Rock the JVM explains these topics very nicely.
The Power of Recursion - How To Think Code Clearly
https://www.youtube.com/
To view or add a comment, sign in
Lets Connect! Cyber, Risk, IT, Audit, Compliance, Product Development, Entrepreneur, Public Speaker
2moStay tuned into groq. Thier technology looks to be a game changer.