Talking w/ a friend and it's not at all clear what open source means in an era of LLMs (what, exactly, is the source code?) and AI, generally. I suspect we're going to need another big discussion, similar to when cloud changed how we thought about code "distribution".
CODE
-
Defining and Categorizing LLM Fine-tuning Approaches
By
–
How do we define and categorize LLM "finetuning" these days? There's 1) Supervised finetuning: to 1a) train an encoder-LLM for classification or 1b) train a decoder-LLM to generate a specific response text. [1/2]
-
Three types of fine-tuning approaches for language models
By
–
[2/2] 2) Instruction finetuning: usually RLHF to generate a specific response. (But can also refer to 1b depending on the technique) 3) Long-context finetuning: like pretraining but w longer contexts (as per Code Llama paper)?
-
Keras Core Models Compatible with PyTorch Module Objects
By
–
Inversely, Keras Core models behave like torch `Module` objects and can be used anywhere you'd use a `Module`.
-
Keras Core now supports arbitrary PyTorch Modules with TorchModuleWrapper
By
–
Now, it's possible to use arbitrary `Modules` inside a Keras Core model/layer (with the torch backend). The `TorchModuleWrapper` class turns them into Keras layers, keeping tracking of trainable weights / etc.
-
Build Powerful AI Agents in 2 Hours with LLMOps Platform
By
–
Build Powerful AI Agents in 2 hours using our LLMOps Platform. Sep 5, 9:00 am PST Free for everyone to attend. • Custom data transformers
• Vector stores
• Integrate with any LLM
• Chain data, code, and LLMs
• Deploy and monitoring Register here: -
Debugging Strategy: Detective Work vs Trial and Error
By
–
Debugging is half careful detective work & half throwing spaghetti against the wall. Key to productivity is figuring out when to be in either mode.
-
Inference Mode Quantization with BNB NF4 for CodeLlama
By
–
And if you are looking for inference mode quantization, you can use `–quantize "bnb.nf4" with the "generate/base.py" scripts as well.
I am currently using that for running CodeLlama 34B models. -
QLoRA-style Quantization Added with BNB NF4 Support
By
–
Yes! Just added QLoRA-style quantization the other week, too. You can use it via –quantize "bnb.nf4" for example. I have some benchmarks here:
-
Kelsey Hightower critiques Terraform description in technical discussion
By
–
Worth listening to @kelseyhightower in a conversation with @bcantrill (
https://
youtube.com/watch?v=13ctYO
u8TsA#t=39m40s
…). Kelsey, as a contributor to Terraform, clearly isn't impressed with McJannet's description, and rightly so.