chapter: mlops-engines #28
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
casperdcl
suggested changes
Aug 31, 2023
There was a problem hiding this comment.
- intro too long
- intro mostly off-topic (what does it have to do with MLOps Engines? a lot of this looks more relevant to other chapters)
- missing clear definition of "MLOps Engine"
- missing list/table of MLOps Engines & feature comparison (see e.g. this)
- missing justifications (every claim needs to be backed up, you can't just state personal opinions)
- use clean book-not-blog language, e.g.
## Some Thoughts About The Future->## FutureWith large language models, the story is no different->LLMs are similar- stick to third-person, and definitely don't keep changing between first, second, and third
- avoid repetition
- did you see the original section outline? e.g. Python Bindings, Apache TVM, links to read before writing anything, etc.
mlops-engines.md
Outdated
|
|
||
| ## The MLOps Lifecycle | ||
|
|
||
|  |
There was a problem hiding this comment.
I'm not sure I understand this diagram - what is it trying to show?
There was a problem hiding this comment.
This diagram shows the LLMOps lifecycle. Its just suppose to serve as a banner image
There was a problem hiding this comment.
I'm particularly struggling to understand the meaning of the arrows and the colours.
mlops-engines.md
Outdated
|
|
||
| ## Challenges With Open-Source MLOps | ||
|
|
||
| MLOps has always been available in two flavors. One is the managed version, where all the components are provided out of the box for a steep price. The other is a DIY setup where you stitch together various open-source components. |
mlops-engines.md
Outdated
Comment on lines
163
to
165
| Due to the challenge of running LLMs, enterprises will opt to use an inference server instead of containerizing the model in-house. Most companies don't have the expertise to optimize these models, but they still want the performance benefits. Inference servers, whether they are open-source or not, will be the path forward. | ||
|
|
||
| Another pattern that's emerging is that models will move to the data instead of the data moving to the model. Right now if you call the ChatGPT API, you would be sending your data to the model. Enterprises have worked very hard over the past decade to set up robust data infrastructure in the cloud. It makes a lot more sense to bring the model into the same cloud environment where the data is. This is where open-source models being cloud agnostic have a huge advantage. |
There was a problem hiding this comment.
citations needed. If it's your personal opinion, sate why. Otherwise assume the reader disagrees with every claim you make.
casperdcl
reviewed
Aug 31, 2023
mlops-engines.md
Outdated
| * Takes a while for new LLMs to be supported | ||
|
|
||
|
|
||
| Many other open-source projects like [BentoML](https://www.bentoml.com/), [FastAPI](https://fastapi.tiangolo.com/), and [Flask](https://flask.palletsprojects.com/en/2.3.x/) have been used for serving models in the past. The reason I have not included them on this list is that these open-source tools do not provide the optimizations you need to run LLMs in production. |
There was a problem hiding this comment.
that's not a reason. Also, what are "the optimizations you need to run LLMs in production"?
casperdcl
force-pushed
the
mlops-engines
branch
2 times, most recently
from
671da89
to
3c8bae8
Compare
casperdcl
force-pushed
the
mlops-engines
branch
from
11eb2f9
to
53b7dca
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Review checklist
# h1-Title## Summary/Introductiontitle or equivalent## Futuredevelopments{{ comments }}new-chapter.md), add_toc.ymlentry &index.mdtable row_config.yml:sphinx.config.linkcheck*fixes #4