-
Notifications
You must be signed in to change notification settings - Fork 282
Pull requests: apple/axlearn
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Support remat and its HF weights conversion for Bert-based text embedding encoder
#99
by JinhaoLei
was merged Oct 5, 2023
Loading…
Adds support for torch param conversion for LayerNormStateless.
#89
by ruomingp
was merged Oct 1, 2023
Loading…
enable argnum donation on all architectures, call _step_log() for early steps, and add JAX scope annotations
#88
by apghml
was merged Oct 3, 2023
Loading…
have inference runner use the same custom create_device_mesh as trainer
#87
by JinhaoLei
was merged Sep 29, 2023
Loading…
Makes Initializer a subclass of Configurable for consistency.
#85
by ruomingp
was merged Sep 28, 2023
Loading…
Use BertLMHead as vocab_mapping child of SpladePooling layer
#82
by JinhaoLei
was merged Sep 27, 2023
Loading…
Improves tpu runner status check and test coverage.
#81
by markblee
was merged Sep 26, 2023
Loading…
Modify trainer's _create_device_mesh method for better behavior on GPU.
#80
by tgunter
was merged Sep 24, 2023
Loading…
Increase sharding-spec flexibility in trainer.py and attention.py
#75
by tgunter
was merged Sep 18, 2023
Loading…
ProTip!
Adding no:label will show everything without a label.