.

content/experience.md

@@ -8,6 +8,8 @@ author = "Carl Pearson"
 
 ## Community
 
+- IPDPS'24 AsHES workshop program chair
+- SC'23 ExaMPI workshop committee member
 - ICPP'23 program committee member
 - ISC'23 workshop commitee member
 - IPDPS'23 AsHES Workshop program committee member

content/post/20211201-links/index.md

@@ -1,7 +1,7 @@
 +++
 title = "Interesting Links"
 date = 2022-11-26T00:00:00
-lastmod = 2022-11-27T00:00:00
+lastmod = 2023-06-04T00:00:00
 draft = false
 
 # Authors. Comma separated list, e.g. `["Bob Smith", "David Jones"]`.
@@ -58,6 +58,22 @@ Presented without comment or endorsement.
 * [A Chemical Hunger – Interlude I: The Fattest Cities in the Land](https://slimemoldtimemold.com/2021/11/08/a-chemical-hunger-interlude-i-the-fattest-cities-in-the-land/) 
 * [A Chemical Hunger – Part X: What to Do About It](https://slimemoldtimemold.com/2021/11/23/a-chemical-hunger-part-x-what-to-do-about-it/)-->
 
+## June 2023
+
+* [DreamBerd Programming Language](https://github.com/TodePond/DreamBerd)
+
+## May 2023
+
+* [Sandspiel](https://sandspiel.club/#)
+* [DALLE mini](https://craiyon.com)
+* [The Scaling Hyopthesis](https://gwern.net/scaling-hypothesis)
+* [Learning the Elite Class](https://aella.substack.com/p/learning-the-elite-class)
+* [I'm trying to put a command in a variable, but the complex cases always fail!](http://mywiki.wooledge.org/Bash/FAQ/050)
+* [(manifold) Will and independent RCT study evaluating Silexan  for anxiety find a standardized effect size greater than 0.5?](https://manifold.markets/AlvarodeMenard/will-an-independent-rct-study-evalu)
+* [Things your manager might not know](jvns.ca/blog/things-your-manager-might-not-know/)
+* [doodle : ai](https://doodleai.darraghoriordan.com/)
+
+
 ## November 2022
 
 * [RFC 1855: Netiquette Guidelines](https://www.rfc-editor.org/rfc/rfc1855) ([archive](https://web.archive.org/web/20221118154752/https://www.rfc-editor.org/rfc/rfc1855))

content/post/20230506_vicuna/index.md

@@ -1,5 +1,5 @@
 +++
-title = "Running Vicuna-13B in the Cloud"
+title = "Running Vicuna-13B in Google Cloud"
 date = 2023-05-06T00:00:00
 lastmod = 2023-05-06T00:00:00
 draft = false
@@ -9,7 +9,7 @@ authors = ["Carl Pearson"]
 
 tags = []
 
-summary = ""
+summary = "How to experiment with hosting Vicuna-13B on a cloud VM"
 
 # Projects (optional).
 #   Associate this post with one or more of your projects.
@@ -39,7 +39,11 @@ categories = []
 
 +++
 
-Vicuna-13B is an LLM based off of the LLaMa model.
+[Vicuna-13B](https://lmsys.org/blog/2023-03-30-vicuna/) is an LLM chatbot based off of the LLaMa model.
+It's authors claim it achieves 90% of the quality of ChatGPT in a "fun and non-scientific" evaluation.
+
+You can rent some cloud hardware and experiment with Vicuna 13B yourself!
+Using CPU-only is slow (couple tokens per second), but fast enough for yout to get an idea of what to expect.
 
 ## Set up your Cloud Instance
 
@@ -47,7 +51,9 @@ Create a cloud VM with
 * 150 GB of disk space
 * 64 GB of CPU memory
 
-When everything was done, my VM had 132GB of disk space used, so you'll probably want at least 150GB
+I used a Google Compute Engine `e2-standard-16`, which costs around $0.70/hour, so it may not be something you want to leave running. You can stop the instance when you're not using it.
+
+When everything was done, my VM had 132GB of disk space used.
 
 Ordinarily I wouldn't recommend setting up python like this, but since we're just experimenting:
 
@@ -123,7 +129,7 @@ I got an error about regenerating protobuf functions if I used protobuf > 3.20.
 
 ## Apply the vicuna deltas
 
-FastChat has done the work of getting a little chat interface set up.
+[FastChat](https://github.com/lm-sys/FastChat) has done the work of getting a little chat interface set up.
 We'll use their package to download the deltas and apply them as well.
 
 ```
@@ -138,4 +144,8 @@ I had `fschat 0.2.5`.
 
 ## Start Chatting
 
-python3 -m fastchat.serve.cli --device cpu --model-path vicuna-13b/
+This will open up a little Chat-GPT-style interface in your terminal.
+
+```
+python3 -m fastchat.serve.cli --device cpu --model-path vicuna-13b/
+```