profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/andrew-christianson/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Andrew Christianson andrew-christianson Los Angeles, CA

andrew-christianson/lsp-python-ms 4

emacs lsp-mode client for Microsoft's python language server

andrew-christianson/alerter 0

Send User Alert Notification on Mac OS X 10.8 from the command-line.

andrew-christianson/amazon-sagemaker-workshop 0

Amazon SageMaker workshop lab guides and materials

andrew-christianson/anaconda 0

Anaconda turns your Sublime Text 3 in a full featured Python development IDE including autocompletion, code linting, IDE features, autopep8 formating, McCabe complexity checker and Vagrant for Sublime Text 3 using Jedi, PyFlakes, pep8, PyLint, pep257 and McCabe that will never freeze your Sublime Text 3

andrew-christianson/anaconda-mode 0

Code navigation, documentation lookup and completion for Python.

andrew-christianson/awesome-go 0

A curated list of awesome Go frameworks, libraries and software

andrew-christianson/aws-okta 0

aws-vault like tool for Okta authentication

issue openedfacebook/Ax

documentation for optimize

The optimize example on the main page makes it really easy to get started but I had to hunt around a bit to figure out how to set a parameter to be integer valued ("value_type"). Would be helpful to have a more detailed docstring, say.

created time in 4 hours

issue openedfacebook/Ax

Is there a complete example of use for

I mean: how to prepare the complete experiment with all bells and whistles, as OptimizationConfig, metrics property etc. with the help of ax.storage.sqa_store.encoder.Encoder then save_experiment after that load_experiment back with the help of ax.storage.sqa_store.decoder.Decoder?

I can not understand how to use these two classes in practice.

created time in 5 hours

release conda-forge/miniforge

4.10.1-5

released time in 8 hours

push eventfacebook/Ax

Ax Website Deployment Script

commit sha 41e19c6b6b948a339f6a5ba61de0ccb8d28de9e3

Update latest version of site

view details

push time in 20 hours

pull request commentfacebook/Ax

make reference point optional

This pull request has been merged in facebook/Ax@3d20dc09f297c13902081c24ad14cd99e9fa1c5c.

sdaulton

comment created time in a day

push eventfacebook/Ax

Sam Daulton

commit sha 3d20dc09f297c13902081c24ad14cd99e9fa1c5c

make reference point optional (#601) Summary: Pull Request resolved: https://github.com/facebook/Ax/pull/601 This diff does two things: 1) if a user creates a `MultiObjectiveTorchModelbridge` and calls `gen` without specifying objective thresholds, then we infer the objective thresholds within the `MultiObjectiveBotorchModel`, generate candidates, and return the inferred objective thresholds in `gen_metadata`. 2) it adds a `infer_objective_thresholds` method to the `MultiObjectiveTorchModelbridge`, which can be used to infer objective thresholds without generating candidates. This refactors the Base `Modelbridge.gen` and `ArrayModelbridge._gen` methods and to apply transformations within a utility function. Note that this method returns ObservationData. If the user wants to plot outcomes with objective thresholds, the user would have to create the ObjectiveThresholds and set the objective thresholds on the optimization config. Reviewed By: Balandat Differential Revision: D28163744 fbshipit-source-id: c74908290bf6b7162771c0768e06bf8181fd4185

view details

push time in a day

PR closed facebook/Ax

make reference point optional CLA Signed fb-exported

Summary: This diff does two things: 1) if a user creates a MultiObjectiveTorchModelbridge and calls gen without specifying objective thresholds, then we infer the objective thresholds within the MultiObjectiveBotorchModel, generate candidates, and return the inferred objective thresholds in gen_metadata. 2) it adds a infer_objective_thresholds method to the MultiObjectiveTorchModelbridge, which can be used to infer objective thresholds without generating candidates.

This refactors the Base Modelbridge.gen and ArrayModelbridge._gen methods and to apply transformations within a utility function.

Note that this method returns ObservationData. If the user wants to plot outcomes with objective thresholds, the user would have to create the ObjectiveThresholds and set the objective thresholds on the optimization config.

Reviewed By: Balandat

Differential Revision: D28163744

+748 -64

7 comments

9 changed files

sdaulton

pr closed time in a day

pull request commentfacebook/Ax

make reference point optional

This pull request was exported from Phabricator. Differential Revision: D28163744

sdaulton

comment created time in a day

pull request commentfacebook/Ax

make reference point optional

This pull request was exported from Phabricator. Differential Revision: D28163744

sdaulton

comment created time in a day

pull request commentfacebook/Ax

make reference point optional

This pull request was exported from Phabricator. Differential Revision: D28163744

sdaulton

comment created time in a day

pull request commentfacebook/Ax

make reference point optional

This pull request was exported from Phabricator. Differential Revision: D28163744

sdaulton

comment created time in a day

issue commentfacebook/Ax

Augmenting trial run metadata with metadata from `Runner.stop` call

My understanding is that your flow is different because the metadata about the job/process you use to evaluate the trials is only available when the process is completed –– is that right?

Yes, and your temporary suggestion equally works until future updates.

This part may not be as I predicted but, unlike synthetic metrics that implement an offline function f(), I couldn't think of another way to handle process data.

I'm not 100% sure I understand what you mean by "handling process data" here...

I meant "calculating the metric cost based on the trial metadata" by "handling" as you correctly continued in your answer. Thank you.

That sounds like our docs could use some improvement : ) If you have thoughts on what we should add or notes on what parts you found confusing/lacking, please let us know!

The main problem was I read everything on the website site months ago, and I didn't realize that you were updating the site, so I didn't bother going through the same content. Instead, I felt that I had to dig the code. Although that worked pretty well in the end, a single trial add-run-fetch cycle made me go through several distracting BoTorch details while my major concern was first to understand the experimentation workflow. Recently, I realized a new content "multi-objective opt.". Now, I saw that you have completely changed the order of the content. So I will reread everything and let you know if have any problem with the documentation.

An "updates" section with a date of the update on the main page would probably be very beneficial to track the changes to the website. But periodically checking this folder also works for me.

ugurmengilli

comment created time in a day

pull request commentfacebook/Ax

make reference point optional

Codecov Report

Merging #601 (0860936) into master (d969efc) will increase coverage by 0.04%. The diff coverage is 98.12%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master     #601      +/-   ##
==========================================
+ Coverage   93.77%   93.81%   +0.04%     
==========================================
  Files         345      345              
  Lines       26900    27150     +250     
==========================================
+ Hits        25225    25471     +246     
- Misses       1675     1679       +4     
Impacted Files Coverage Δ
ax/models/torch/botorch_moo.py 95.72% <90.24%> (-2.96%) :arrow_down:
ax/modelbridge/multi_objective_torch.py 94.11% <96.55%> (+1.13%) :arrow_up:
ax/modelbridge/array.py 99.19% <100.00%> (+0.11%) :arrow_up:
ax/modelbridge/base.py 99.36% <100.00%> (+0.02%) :arrow_up:
...ge/tests/test_multi_objective_torch_modelbridge.py 96.24% <100.00%> (+2.30%) :arrow_up:
ax/models/tests/test_botorch_moo_model.py 98.80% <100.00%> (+0.42%) :arrow_up:
ax/models/tests/test_torch.py 100.00% <100.00%> (ø)
ax/models/torch_base.py 96.15% <100.00%> (+0.50%) :arrow_up:
ax/utils/testing/core_stubs.py 93.27% <0.00%> (+0.23%) :arrow_up:

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update d969efc...0860936. Read the comment docs.

sdaulton

comment created time in a day

pull request commentfacebook/Ax

make reference point optional

This pull request was exported from Phabricator. Differential Revision: D28163744

sdaulton

comment created time in a day

issue commentfacebook/Ax

Trying to save a experiment_type:str to SQA store

Good point!

alxfed

comment created time in a day

pull request commentfacebook/Ax

refactor subset_model utility

This pull request has been merged in facebook/Ax@d969efcb85287f065910362bd88ddd36e5013d40.

sdaulton

comment created time in a day

push eventfacebook/Ax

Sam Daulton

commit sha d969efcb85287f065910362bd88ddd36e5013d40

refactor subset_model utility (#599) Summary: Pull Request resolved: https://github.com/facebook/Ax/pull/599 Return subset indices in subset_model and use new SubsetModelData dataclass Reviewed By: Balandat Differential Revision: D29214188 fbshipit-source-id: ecaad8a4bf5c30b6cdf2aea3470ef14040d16e40

view details

push time in a day

PR closed facebook/Ax

refactor subset_model utility CLA Signed fb-exported

Summary: Return subset indices in subset_model and use new SubsetModelData dataclass

Differential Revision: D29214188

+104 -50

3 comments

8 changed files

sdaulton

pr closed time in a day

issue closedfacebook/Ax

Trying to save a experiment_type:str to SQA store

I'm trying to save an experiment_type string that has been assigned to the experiment_type property of my Experiment to SQA Store. It is declared as Optional[str] property in the signature of the Experiment class, and it assigns and reads beautifully. However, if you store the Experiment with this assigned (and present) experiment_type to the SQA store there is a 'null' in the column experiment_type of the experiment_v2 table. All the other fields are ORMapped to the table perfectly.

If you look at the properties of the class ax.storage.sqa_store.sqa_classes.SQAExperiment(*args: Any, **kwargs: Any) experiment_type is Optional[int] = Column(None, Integer() there so it must be Enumed along the way somewhere

Should the experiment_types be declared or registered in some way before being used in Experiment classes so that they would store in the SQA store properly?

closed time in a day

alxfed

issue openedjulia-vscode/LanguageServer.jl

wrong import for include automatically added

Sometimes when I type include, I get this autocomplete suggestion:

Schermafbeelding 2021-06-07 om 17 42 52

and selecting it will add this first line to my file:

Schermafbeelding 2021-06-07 om 17 43 12

which is not what we want: include and eval are defined inside every module, and I want to use the local one.

created time in a day

issue commentjulia-vscode/LanguageServer.jl

StackOverflowError()

I'm also seeing these stack overflows in crash reporting. Should probably be a priority for us to try to fix this.

jtschneider

comment created time in a day

issue commentfacebook/Ax

Trying to save a experiment_type:str to SQA store

Hi, again @lena-kashtelyan :) Thank you. That's exactly what I expected, but didn't manage to find this (really valuable for me) link. Thank you.

alxfed

comment created time in a day

issue commentfacebook/Ax

Trying to save a experiment_type:str to SQA store

Hi, @alxfed! Great question. There are some docs on specifying experiment type here: https://ax.dev/docs/storage.html#customizing-1. Let me know if that helps!

alxfed

comment created time in a day

issue openedfacebook/Ax

Trying to save a experiment_type:str to SQA store

I'm trying to save an experiment_type string that has been assigned to the experiment_type property of my Experiment to SQA Store. It is declared as Optional[str] property in the signature of the Experiment class, and it assigns and reads beautifully. However, if you store the Experiment with this assigned (and present) experiment_type to the SQA store there is a 'null' in the column experiment_type of the experiment_v2 table. All the other fields are ORMapped to the table perfectly.

If you look at the properties of the class ax.storage.sqa_store.sqa_classes.SQAExperiment(*args: Any, **kwargs: Any) experiment_type is Optional[int] = Column(None, Integer() there so it must be Enumed along the way somewhere

Should the experiment_types be declared in some way before being used in Experiment classes so that they would store in the SQA store properly?

created time in a day

push eventfacebook/Ax

Lena Kashtelyan

commit sha 8f39ac3ef9c7c4aa1d4198c98a0038e016ac5ca6

Allow to specify custom arms to `ax_sweep_impl` Summary: Taking care of one of our TODOs for IGML Reviewed By: bernardbeckerman Differential Revision: D29205954 fbshipit-source-id: ba370dbf4bc7c821628731f6732037c79bcaf39b

view details

push time in a day

issue openedEleutherAI/gpt-neo

Enable GPT-Neo-125M for downstream training Effectively

Issue while finetuning GPT-Neo-125M on Downstream task

We have a GPT-Neo model with comparatively fewer parameters that can be used for the downstream tasks. It can understand longer sequences.

Currently, it can't be trained in Mini batches because of no padding token in the tokenizer used during pretraining because of that I was facing an issue

Solution:

https://github.com/huggingface/transformers/issues/12022#issuecomment-854711470

created time in a day

pull request commentfacebook/Ax

make reference point optional

This pull request was exported from Phabricator. Differential Revision: D28163744

sdaulton

comment created time in a day

PR opened facebook/Ax

make reference point optional

Summary: This diff does two things: 1) if a user creates a MultiObjectiveTorchModelbridge and calls gen without specifying objective thresholds, then we infer the objective thresholds within the MultiObjectiveBotorchModel, generate candidates, and return the inferred objective thresholds in gen_metadata. 2) it adds a infer_objective_thresholds method to the MultiObjectiveTorchModelbridge, which can be used to infer objective thresholds without generating candidates.

This refactors the Base Modelbridge.gen and ArrayModelbridge._gen methods and to apply transformations within a utility function.

Note that this method returns ObservationData. If the user wants to plot outcomes with objective thresholds, the user would have to create the ObjectiveThresholds and set the objective thresholds on the optimization config.

Reviewed By: Balandat

Differential Revision: D28163744

+741 -112

0 comment

15 changed files

pr created time in a day

pull request commentfacebook/Ax

refactor subset_model utility

This pull request was exported from Phabricator. Differential Revision: D29214188

sdaulton

comment created time in a day