-
Notifications
You must be signed in to change notification settings - Fork 12
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Float16 support + Improved Context hyperparameter support + GNNs layer-wise inference #51
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
… gradient updates.
Added MapInputsToEdges to create map with edge indices.
… it is created -- if variable is created after the checkpoint. * Added `ogbnmag.ExcludeOgbnMagVariablesFromSave`.
Renamed: shapes.UInt{X} -> shapes.Uint{X}; Added an alias to the previous UInt{X} form.
Added test for layer-wise inference.
…per embeddings preserving old value.
… gather to use the correct sortedness.
…?) bug for non-sorted scatter on those models for float16.
janpfeifer
changed the title
Improved Context hyperparameter support + GNNs layer-wise inference
Float16 support + Improved Context hyperparameter support + GNNs layer-wise inference
Jun 10, 2024
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Outdated suggestions cannot be applied.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.
types.shapes
package:Float16
training -- tested with GNNs.Float16
.Adam
optimizer -- it requires at leastfloat32
, even if the model runs onfloat16
.epsilon
values forSoftmax
andAdam
-- current values would lead toNaN
withfloat16
.DType.IsFloat16
to check forFloat16
orBFloat16
(not yet well-supported).Int8
,Int16
,Uint8
andUint16
.UInt{X}
toUint{X}
and added a deprecated alias to the old form (so it still compiles).Variable.SetValueGraph
when the shape changes. Improved some documentation.optimizers
:FromContext
to allow further configuration of the optimizer by setting other hyperparameters into context.clip_step_by_value
, a clip by value applied to gradient updates.Adam
optimizer:"clip_step_by_value", "adam_epsilon", "adam_dtype"
hyperparameters support.MustOptimizerByName
now takes also the context for the optimizer hyperparameters. -- this breaks the API.checkpoints
:slices.CloseToEpsilon
to easily customize tests.Variable.SetValuePreservingOld
when shapes change.Scatter
doesn't assume indices are sorted or unique.WithCustomMetricFn
for custom metrics andScheduleEveryNSteps
.Shift
,ShiftLeft
,ShiftRight
,ShiftWithScalar
,ShiftWithValue
.