memhole

10 ms·

121.
by memhole 2y ago
Do you have any suggestions on deploying? I looked over the readme. I’m not understanding how to go from clone to cloud. Meetup sucks and I’m wondering if this might be a good replacement for local groups.
122.
by memhole 2y ago
I’ve been working on writing this year. It’s hard. I have countless posts or ideas for posts sitting there because I don’t like them for one reason or another. For me I don’t collect metrics or really try to think that anyone might read a s
123.
by memhole 2y ago
This is the top comment I hope for on every Christmas. Some of us draw snowmen instead of Santas. We all get a day to enjoy in our own way though. Sounds like a great time.
124.
by memhole 2y ago
Would love to see distill come back
125.
by memhole 2y ago
I’d happily put it with ms paint art. If anything it’ll make it more palatable. I just don’t like the junk food aspect. At first glance it’s neat and a second longer you question what dimension the subjects came from
126.
by memhole 2y ago
Not sure if this counts. I recently went from description of a screenshot of graph to generate pandas code and plot from description. Conceptually it was accurate. I don’t think it reflects any understanding. But to go from screenshot to co
127.
by memhole 2y ago
Nice write up. I don’t think the comments about duckdb spilling to disk are correct. I believe if you create a temp or persistent db duckdb will spill to disk. I might have missed it, but the integration of duckdb and the arrow library make
128.
by memhole 2y ago
This is how I think about it as well. It’s like the equivalent of trying to build an ML model with only positive examples otherwise
129.
by memhole 2y ago
Start with ignoring 90% of the stuff you read about and realize you’re only manipulating strings to send to an API.
130.
by memhole 2y ago
Famously crashy? Any incidents you can share? I’ve only had good experiences is why I ask.
131.
by memhole 2y ago
Wal mode allows for concurrent readers and one writer. I’ve been use a similar architecture for a Flask app using SQLite and Redis. So far I haven’t had any issues sending thousands of events. The write process is a few milliseconds and fro
132.
by memhole 2y ago
Only a data engineer here, but I agree. If I’m understanding right, your competitors are more likely Fivetran/Airbyte, and Snowflake/Databricks. There’s lots of money and marketing you’ll need to overcome. Mage reminds me of the Z
133.
by memhole 2y ago
Thanks for this paper! Still early so not quite production, but I’ve seen positive results on tasks scaling the number of “agents”. I more or less think of agents as a task focused prompt and these agents are slight variations of the task.
134.
by memhole 2y ago
I used this for a project recently: https://fomantic-ui.com/
135.
by memhole 2y ago
Do you avoid things like task sensors? Based off what you described it sounds like an anti pattern if you’re using them. Great description of good orchestration design. Airflow is fairly open ended in how you can construct dags, leading to
136.
by memhole 2y ago
Their Unity Catalog covers lineage and governance. Logging is still pretty bad last time I used it. But that’s more of PySpark issue with only the driver logs being collected.
137.
by memhole 2y ago
I’m so glad to see LLMs spark these conversations lately. It’s been a huge gripe of mine that we don’t question the underlying precision in other areas of AI/ML
138.
by memhole 3y ago
Hierarchical and DBSCAN don’t require upfront knowledge about the number of clusters.