Tracking Examples: linalg and ma

[bmwoodruff]

• Claim a group by making a comment below.
• ~While Nebari is generating examples~ See the log files for the output from Llama3. Spend some time with the functions to see what they do.
• Some of these functions are NOT something we need to add examples to. You'll have to decide that.
• When you've finished the group of 10, check it off.

np.linalg

np.linalgGroups:

• [ ] Group 1
• [ ] Group 2
• [ ] Group 3

Group 1	Group 2	Group 3
cholesky	lstsq	slogdet
cond	matmul	solve
cross	matrix_norm	svd
det	matrix_power	svdvals
diagonal	matrix_rank	tensordot
eig	matrix_transpose	tensorinv
eigh	multi_dot	tensorsolve
eigvals	norm	trace
eigvalsh	outer	vecdot
inv	pinv	vector_norm

np.ma

np.maGroups:

• [ ] Group 1
• [ ] Group 2
• [ ] Group 3
• [ ] Group 4
• [ ] Group 5
• [ ] Group 6
• [ ] Group 7
• [ ] Group 8
• [ ] Group 9
• [ ] Group 10
• [ ] Group 11

Group 1	Group 2	Group 3	Group 4	Group 5
allclose	asanyarray	compress_rows	dot	getmask
allequal	asarray	compressed	ediff1d	getmaskarray
alltrue	average	concatenate	filled	in1d
amax	choose	convolve	fix_invalid	inner
amin	clump_masked	correlate	flatnotmasked_contiguous	innerproduct
append	clump_unmasked	count_masked	flatnotmasked_edges	intersect1d
apply_along_axis	common_fill_value	cov	flatten_mask	isMA
apply_over_axes	compress_cols	default_fill_value	flatten_structured_array	isMaskedArray
argsort	compress_nd	diag	fromflex	is_mask
array	compress_rowcols	diff	getdata	is_masked

Group 6	Group 7	Group 8	Group 9	Group 10
isarray	masked_all	masked_object	ndim	reshape
isin	masked_all_like	masked_outside	notmasked_contiguous	resize
left_shift	masked_equal	masked_values	notmasked_edges	right_shift
make_mask	masked_greater	masked_where	outer	round
make_mask_descr	masked_greater_equal	max	outerproduct	round_
make_mask_none	masked_inside	maximum_fill_value	polyfit	set_fill_value
mask_cols	masked_invalid	median	power	setdiff1d
mask_or	masked_less	min	ptp	setxor1d
mask_rowcols	masked_less_equal	minimum_fill_value	put	shape
mask_rows	masked_not_equal	ndenumerate	putmask	size

Group 11
sometrue
sort
take
transpose
union1d
unique
vander
where

[otieno-juma]

I will work on ma group 1

[bmwoodruff]

@otieno-juma No need to tag me on these, but you will want to specify if you're working on linalg group 1, or ma group 1.

[jud-sdev]

I'll work on np.linalg group 2

[bmwoodruff]

Note that the output of Llama3-70B is now stored in these log files .

[jud-sdev]

Here's what I got when I tried generating examples for lstsq function in group 2.

[bmwoodruff]

It looks like you haven't cloned this repo yet to Nebari. You'll want to start with that.

[jud-sdev]

I've cloned the repo. I think the issue is related to the upgrade of Numpy. Have you tried generating an example today?

[bmwoodruff]

All the examples have been generated already by AI. The log files are in the possee-org/gen-ai repo.

The error message you see above has always shown when using Llama 3. You can ignore it, and things still work.

[jud-sdev]

Okay, I get it. Thank you so much, I truly appreciate it!

[otieno-juma]

I'm going to work on ma group 9

[otieno-juma]

I'm going to work on ma group 9

[bmwoodruff]

@otieno-juma For now, focus only on functions that have no examples in the code base. You can manually select these with the example_post_processing.py file, and then generate just those.

The Triage team does not want to review 1000 examples that are injected systematically throughout the entire codebase. So while we have a tool that can do that, it isn't wanted in that form.

[bmwoodruff]

After discussion at the triage meeting, adding examples to every function in NumPy is not wanted. Focus on just the functions that don't have examples. After we review the prompt generator, and guarantee that the newly generated examples do indeed add something of more than just marginal value, we can revisit this topic. I'm closing this issue for now.