notes

notes/ai.scrbl

@@ -70,10 +70,10 @@
 
   Transformer-based architectures like GPT, BERT, and their variants:
 
-  Model:
-  - computational representation of a system or process, that learns from data /
-    dataset to make decisions and predictions. The goal of learning is to
-    minimize the difference between predictions and actual outcomes
+  Model i.e. the "brain":
+  The original data is not don't needed. The model is ready to perform tasks
+  based on what it learned during training. I.e. it can run on your own hardware,
+  For fine-tuning or further training of the brain, you might need additional data.
   - consists of
     * architecture
     * parameters
@@ -92,7 +92,6 @@
   its intraction with the model. In the end all the data is created by the AIs
   The algorithms train another algorithms.
 
-
   Instruct-tune LLaMA on consumer hardware
   7.6K  https://github.com/tloen/alpaca-lora
 

notes/bric_a_brac.scrbl

@@ -1,5 +1,12 @@
 #lang notes
 
+
+@block{@block-name{bc - An arbitrary precision calculator language}
+  printf "1 / 2\nquit\n" > /tmp/math.txt
+  bc --quiet --mathlib /tmp/math.txt 
+  .50000000000000000000
+}
+
 @block{@block-name{Sway / i3 / keys keybingings / Refcard / Reference Card}
   | Alt + Enter         | open new terminal                 |
   | Alt + j             | focus left                        |

notes/cli/linux.scrbl

@@ -299,11 +299,11 @@
   # -C   display progress bar; file descriptor is for GUIs
   # -A   check all filesystems
   # sudo fsck -C -V /dev/disk/by-id/YOUR_FLASH_DRIVE-PARTITION-TO-CHECK
+  # sudo e2fsck -v $diskPart     # -v verbose
   sudo umount        $diskPart
   # or
   sudo umount --lazy $diskPart # detach the filesystem now, clean up things later
 
-  #
   # -w   data-destructive write-mode test. (Mutually exclusive with -n)
   #      Scans for bad blocks by writing some patterns on every block of the
   #      device, reading every block and comparing the contents.