SpotML
AI
ml training infrastructure
SHOW HN2021
157points
68comments
#309of 2,533 in 2021
Top 13%of 2021 by points
How HN took it
MixedNo clear lean
7 positive, 11 neutral or asking, 5 mixed, 5 critical, out of 28 comments about the project. Thread score +0.05.
Idea: 5 for, 5 againstDocs: 1 for, 0 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Useful for teams and hobbyists paying for spot GPU time
From 3 comments
This looks useful, I like the pricing of free/$9.99 a month. No one asked, but it's HN so I'll say it anyway. I think it's a questionable "vc business" but a great business for…
talolard on HNHa! That's interesting. We use AWS instances for model development and costs are definitely an issue. Sending this to my team. Good luck!
cpalumbo on HNReally excited to try it out. I've had a heck of a time setting up spot instance training on Sagemaker in the past so simplification efforts are much appreciated.
zachthewf on HN
Objections
Nimbo and Spotty are very similar and were not clearly credited
From 2 comments
I'd like to point out that this seems extremely similar to Nimbo (https://nimbo.sh), to the extent that even some of the terminal messages are exactly the same, and even parts of…
mjaques on HNYou should really mention / give attribution / emphasize more that this is a fork of https://spotty.cloud and you took a lot from https://github.com/nimbo-sh/nimbo as well.
rgj on HN
Native retry on spot instances or managed tools like Metaflow, AWS Batch or GKE may already cover this
From 3 comments
Read through the homepage, but not entirely sure -- Why not just train on Spot Instances with a retry implemented? I see that SpotML has a configurable fall back to On-Demand…
Gatesyp on HNNeat. Congratulations on the launch, Vishnu! Apart from the fact that it could deploy to both GCP and AWS, what does it do differently than AWS Batch [0]? When we had a similar…
ignoramous on HNI'm not sure what this provides over what GCP already offers. 4 years ago I switched my co's ML training to use GKE (Google Kubernetes Engine) using a cluster of preemptible…
twistedpair on HN
Cloud providers could copy this functionality
From 1 comment
Kudos; this is likely a very useful tool! So useful (and potentially lucrative) that AWS/GCP would likely take over this functionality eventually -- either by building on top of…
ssivark on HN
Requests
Support distributed training and multiple instances
From 2 comments
Does it apply to jobs running on multiple instances, e.g using dask?
oakfr on HNdo you have an example dockerfile for a ML job ? say xgboost with a CSV data file ... or tensorflow with image data ? unable to figure out how to construct the job itself and how…
sandGorgon on HN
Add examples and documentation
From 2 comments
do you have an example dockerfile for a ML job ? say xgboost with a CSV data file ... or tensorflow with image data ? unable to figure out how to construct the job itself and how…
sandGorgon on HNThis looks cool. How much SpotML-specific code is needed to target SpotML? I'm assuming it doesn't just magically detect the training loop in an existing pytorch codebase and…
ShamelessC on HN
Charge more or branch into inference
From 1 comment
Disclosure: I used to work for GCP and launched Preemptible VMs. Congrats! Can I suggest charging more? IIUC, your business plan is $10/month for the company, regardless of…
boulos on HN
Questions
How it differs from spot retry and Metaflow or AWS Batch
From 2 comments
Read through the homepage, but not entirely sure -- Why not just train on Spot Instances with a retry implemented? I see that SpotML has a configurable fall back to On-Demand…
Gatesyp on HNNeat. Congratulations on the launch, Vishnu! Apart from the fact that it could deploy to both GCP and AWS, what does it do differently than AWS Batch [0]? When we had a similar…
ignoramous on HN
How state is saved within the shutdown time limit
From 1 comment
I'm interested in how they 'hibernate'/save the state of the instances within the shutdown time limit. I was also looking into this for myself, there are ways of using docker to…
JZL003 on HN
Which AWS permissions are needed and how billing works
From 1 comment
How does this work? From the documentation it's clear that AWS credentials are required. Which permissions are necessary? This leads me to the assumption that the SpotML cli…
sknzl on HN
Alternatives named
Nimbo
From 1 comment
I'd like to point out that this seems extremely similar to Nimbo (https://nimbo.sh), to the extent that even some of the terminal messages are exactly the same, and even parts of…
mjaques on HN
Spotty
From 1 comment
Yes we liked the elegance of both the tool and the docs. So it's very much inspired from it. I must also give credit to another great tool https://spotty.cloud/ from which this…
vishnukool on HN
AWS Batch
From 1 comment
Neat. Congratulations on the launch, Vishnu! Apart from the fact that it could deploy to both GCP and AWS, what does it do differently than AWS Batch [0]? When we had a similar…
ignoramous on HN
Metaflow
From 1 comment
Read through the homepage, but not entirely sure -- Why not just train on Spot Instances with a retry implemented? I see that SpotML has a configurable fall back to On-Demand…
Gatesyp on HN
Cortex
From 1 comment
This is cool project, will try it out. how does it compare to cortex.dev? they use AWS spot instances too.
praveenhm on HN
Grid.ai
From 1 comment
How does it compare with grid.ai?
barefeg on HN