Is Gpt4.5 really more emotionally sensitive?Scientists at OpenAI launched gpt4.5 which is supposed to be more empathetic. I think, not just their effort to make an LLM empathetic have…3d ago3d ago
Huggingface vs llama.cpp tokenizersLlama cpp python repository mention that there is a discrepency between llama.cpp server vs huggingface tokenizer, so I had to test what…Mar 28Mar 28
Fixing GPG key conflicts in apt (Ubuntu)Sometimes when working with Nvidia apt registry or other packages, we can encounter a conflict between signing keys.Jan 26Jan 26
Automating gcloud auth login using tampermonkey scriptWhen working with google cloud, we often get logged out after 10 hours and we have to run gcloud auth login and/or gcloud auth…Jan 14Jan 14
An effective eval_strategy for huggingface trainerThe only options you get in eval_strategy in huggingface trainer are:Nov 28, 2024Nov 28, 2024
Difference between model_max_length and max_lengthTokenizer.model_max_length truncates inputNov 19, 2024Nov 19, 2024
Making deepspeed work on Lambda Labs GPU On-Demand InstanceLambda labs on-demand instance come with Nvidia drivers and CUDA toolkit already installed on their linux machines. The setup however is…Nov 2, 2024Nov 2, 2024
GCP Cloud GPU Quota WorkaroundThe search on GCP UI when creating new on-demand VM with GPU attached is not great and it’s very often a hit and trial on if your machine…Oct 28, 2024Oct 28, 2024
My yabai config to exclude apps easilyConfig to easily exclude apps from yabai configOct 23, 2024Oct 23, 2024