9d46161353
UPDATE: bug fixes when missing config variables
2023-02-25 04:00:09 +00:00
cb809d7852
fixup: Format Python code with Black
2023-02-25 02:03:50 +00:00
ab4fe43d51
UPDATE: adds support for using lookup values for variables outside the model (exogeneous variables like OAT and WBT), and includes support for loading initial state values from the training dataset. Updates `ddm_trainer.py`, and `ddm_predictor.py` to utilize said updates, and updates the `hvac_b1` example to demonstrate
2023-01-26 23:21:01 +00:00
2e2881e516
UPDATE: exogeneous variables lookup
2023-01-11 15:56:22 +00:00
2a95cfc24a
UPDATE: allow for iteration to be used as a feature. Remove tune import from torch models until sweep.
2022-12-14 00:34:49 +00:00
df5e2ed8d4
UPDATE: support for output types when doing prediction; does not enforce any constraints during training, however!
2022-12-06 22:54:57 +00:00
c95985405a
UPDATE: preprocess pipeline support for training and prediction, with example of calculating temperature delta setpoints
2022-12-05 18:19:47 +00:00
8421608edb
UPDATE: ensures ddm_predictor updates only the variables specified in concatenate_var_length and not concatenate length, and that takes precedence during epsiode_step. Also fixes tests, save for skipping over the long sweep tests
2022-11-15 05:22:08 +00:00
eb1b4c353d
UPDATE: timeseries class implementation with example yaml and some docs
2022-09-29 15:06:22 +00:00
64555700d0
UDPATE: cartpole config with variable length lags
2022-08-05 14:46:16 +00:00
09d3089498
UPDATE: adds field for variable length lag for both ddm_trainer and ddm_predictor
2022-08-05 05:14:44 +00:00
5c228e2ca5
fixup: Format Python code with Black
2022-08-04 00:57:31 +00:00
b0e1272b23
MERGE
2022-08-01 22:05:42 -07:00
f180facd4c
premerge
2022-08-01 22:02:15 -07:00
44435e3282
UPDATE: include changes for dataclass to work with ddm_predictor
2022-06-02 13:55:12 +00:00
6f7db4a596
Merge branch 'journeyman/scheduler' of github.com:microsoft/datadrivenmodel into journeyman/scheduler
2021-09-24 15:35:49 -07:00
11259afbaf
bugfix on samples that don't use signal builder
2021-09-24 15:35:36 -07:00
dcde3da721
fixup: Format Python code with Black
2021-09-24 22:21:51 +00:00
2742f5e104
Now overriding signal params if provided from Inkling
2021-09-24 15:20:58 -07:00
f1c9e2e9cb
fixup: Format Python code with Black
2021-09-23 17:30:14 +00:00
2ac3eb9d19
Made Tout usable as input to DDM from signal builder. Added piecewise function
2021-09-23 00:43:40 -07:00
fe8c855c08
fixup: Format Python code with Black
2021-08-25 19:19:51 +00:00
8b95584459
Working signal builder for Tset and Tout using SignalBuilder class.
2021-08-24 22:51:45 -07:00
960133f130
Initial commit of signal builder trying to instantiate using target, but not working
2021-08-24 10:33:36 -07:00
c95cbdbbb5
fixup: Format Python code with Black
2021-08-19 21:28:09 +00:00
6a41c91f61
UPDATE: cleaner yaml examples
2021-08-17 18:32:47 +00:00
e02af1dc1c
fixup: Format Python code with Black
2021-08-13 20:47:02 +00:00
268e1827cd
generalized test_policy. Now has exported brain and random. changed min/max in sample
2021-08-13 13:46:12 -07:00
acedf1e98d
fixup: Format Python code with Black
2021-08-12 23:39:02 +00:00
f3c0eefb7f
UPDATE: push completed yaml example
2021-08-12 23:37:54 +00:00
8646f775cb
UPDATE: initial states mapper; docs todo
2021-08-12 23:31:39 +00:00
1424e36312
bugfix: test loaders now tests correctly and ddm_predictor.get_state() returns a Dict
2021-07-21 18:34:30 +00:00
132420a10d
fixup: Format Python code with Black
2021-07-15 17:00:45 +00:00
9608ccb86f
UPDATE: initializations from episode_init and from initial_states passthru
2021-07-14 13:12:55 -07:00
491180749a
UPDATE: pass episode_inits and configs through test_random_policy
2021-07-14 11:56:14 -07:00
fcdfc03246
slightly better docstring
2021-07-13 17:17:49 -07:00
4041ff825e
UPDATE: initial states in episode_start and in cartpole-updated-simparam.yaml
...
initial states should be defined before episode_step. Previously these were randomly initialized as standard gaussians in episode_start using the keys defined by simulator.states. Now a new value is provdied in simulator.yaml called initial_states which takes a dictionary of initial_values. Moreover, episode_start looks in the `config` Dict (for instance when using scenario parameters in Inkling from lessons) and updates initial_states if provided.
TODO: update documentation and all simulator.yaml files
2021-07-13 17:07:21 -07:00
85074dd902
UPDATE: ddm_predictor torch -> pytorch
2021-07-13 09:31:50 -07:00
48b940da6f
UPDATE: iteration-level logging in simulator.policy=random
2021-06-22 10:20:32 -07:00
de3e226dc0
🐛 fix: track history of data in ddm_predictor and define schema episode_step properly
2021-06-21 19:13:53 -07:00
ae3bbe92a8
:BUG fix + updated cartpole dataset: make sure definition of features matches the states during ddm_predictor; plus new dataset matching in-platform example
2021-06-10 15:43:33 -07:00
846bb30656
UPDATE: add a episode initialization parameter in simulator yaml and ddm_predictor; TODO: add option to calculate summary statistics of configs during ddm_trainer and save to simulator.yaml
2021-06-10 11:09:30 -07:00
aef442424d
UPDATE: env_setup param for ddm_predictor
2021-06-09 14:51:56 -07:00
b316c18aac
UPDATE: make torch installation optional for simulator packages, upgrade mba to 0.1.3, and update dockerfile
2021-06-09 12:38:33 -07:00
556f92a4bb
BUGFIXES: don't rebuild model after load
2021-06-02 16:28:05 -07:00
5de46a1a1b
removing excess azure logs
2021-05-11 12:07:44 -07:00
4e860fff33
fixup: Format Python code with Black
2021-04-23 00:46:17 +00:00
894d15f088
UPDATE: remove azure-cli dependency since resolving its msal dependency is a pain
2021-04-06 13:59:50 -07:00
ee0fd59a73
ADD: github workflows for linting and pytest coverage report
2021-04-06 13:16:55 -07:00
5b2ba393df
UPDATE: add diffstate to trainer + some docs; todos
2021-04-01 12:06:06 -07:00
Ali Zaidi
Journey McDowell
sandip.kulkarni