Alex
@quantext.bsky.social
Quant Trader, Local LLM Enthusiast. Transformers. ASR. TTS. Time-series forecasting
YT channel: https://www.youtube.com/@quantext
Insta(fitness): https://www.instagram.com/being_fit4ever/
X( gonna be ex): https://x.com/BeingHorizontal
YT channel: https://www.youtube.com/@quantext
Insta(fitness): https://www.instagram.com/being_fit4ever/
X( gonna be ex): https://x.com/BeingHorizontal
China slapped an extra 34% tariffs on US imports, and Canada is hot on their heels.
April 4, 2025 at 5:58 PM
China slapped an extra 34% tariffs on US imports, and Canada is hot on their heels.
If you use a SATA SSD and are tempted by NVMe M.2 Gen4 SSD, reconsider. Performance improvement is mainly for frequent large file transfers. App and OS load times will have minimal changes. For LLMs, only load times decrease; inference speed stays the same.
March 1, 2025 at 7:34 PM
If you use a SATA SSD and are tempted by NVMe M.2 Gen4 SSD, reconsider. Performance improvement is mainly for frequent large file transfers. App and OS load times will have minimal changes. For LLMs, only load times decrease; inference speed stays the same.
Getting in the rabbit hole of Local LLMs and transformers, this is the new world and a big paradigm shift.
Dare I say Timeseries forecasting. In my short test, Transformer architecture absolutely outperformed traditional sequence-to-sequence models like LsTM. I will soon put my thoughts together.
Dare I say Timeseries forecasting. In my short test, Transformer architecture absolutely outperformed traditional sequence-to-sequence models like LsTM. I will soon put my thoughts together.
March 1, 2025 at 7:00 PM
Getting in the rabbit hole of Local LLMs and transformers, this is the new world and a big paradigm shift.
Dare I say Timeseries forecasting. In my short test, Transformer architecture absolutely outperformed traditional sequence-to-sequence models like LsTM. I will soon put my thoughts together.
Dare I say Timeseries forecasting. In my short test, Transformer architecture absolutely outperformed traditional sequence-to-sequence models like LsTM. I will soon put my thoughts together.