Update README.md #220
Update README.md #220
Conversation
Some copy edits based on the existing text but I have a bunch of questions...
jorgeorpinel
added
the
A: docs
Codecov Report
@@ Coverage Diff @@
## main iterative/mlem#220 +/- ##
==========================================
+ Coverage 85.48% 90.29% +4.81%
==========================================
Files 77 78 +1
Lines 5675 5882 +207
==========================================
+ Hits 4851 5311 +460
+ Misses 824 571 -253
Continue to review full report at Codecov.
|
Copy edits, typo fix, separate paragraph sentences with new lines, etc.
README.md
Outdated
| - **Simple text file to save model metadata:** | ||
| Automatically include Python requirements and input data needs into a deployment-ready format. | ||
| Use the same format on any ML framework. |
There was a problem hiding this comment.
Is the format YAML? Where can we see the schema?
There was a problem hiding this comment.
Yes, YAML.
Do you think it worth in showing schema somewhere? You aren't supposed to write those Yaml files, only read, if you need it for some reason.
There was a problem hiding this comment.
- In that case no need to include the schema in the README but probably somewhere for reference in the docs. Also consider creating something like https://github.com/iterative/dvcyaml-schema
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
There was a problem hiding this comment.
.mlem yaml schema is dependent on the type of the object. For model, data, deployment, etc, it will be a different schema. IIRC, this is the reason why we cannot easily create the schema as DVC can. We can take a look later, but this is not the priority rn.
There was a problem hiding this comment.
Created the issue to address this later https://github.com/iterative/mlem/issues/306
There was a problem hiding this comment.
OK np. Putting a pseudo schema spec somewhere in mlem.ai.doc would be nice though ⏳
README.md
Outdated
| ## Why is MLEM special? | ||
|
|
||
| The main reason to use MLEM instead of other tools is to adopt a **GitOps approach**, helping you manage model lifecycles in Git: | ||
|
|
||
| - **Git as a single source of truth:** MLEM writes model metadata to a plain text file that can be versioned in a Git repo along with your code. |
There was a problem hiding this comment.
- Consider: This whole section could be merged in with the intro, as well as the bullet lists combined (not as-is).
There was a problem hiding this comment.
UPDATE: See @dberenbaum's recommendation:
the first section bullets should explain the basic workflow [numbered list] rather than the benefits, and the rest should be moved to “Why MLEM?” [bullet list].
|
A few questions above but really I think this we'd need a short usage guide covering all the commands and/or API endpoints (whatever we recommend or expect to get more usage) to get a better sense of what the tool does and how it looks and feels. Is that coming? UPDATE: https://mlem.ai/doc/get-started |
Some copy edits based on the existing text
but I have a bunch of questions...Closes iterative/mlem.ai#88