[core] Layerwise Upcasting #10347
update
36b0c37d
update
42046c09
make style
7dc739b1
Merge branch 'main' into layerwise-upcasting-hook
7ed71413
remove dynamo disable
1fa4ee50
add coauthor
da4907ea
update
bc2ada4d
DN6
commented
on 2024-12-27
Merge branch 'main' into layerwise-upcasting-hook
91bfc3d3
update
7c31bb03
update
8975bbfb
update mixin
341fbfca
add some basic tests
5f898a1f
update
558c64e3
update
7858f2c2
Merge branch 'main' into layerwise-upcasting-hook
26630264
non_blocking
3d84b9e2
improvements
93726473
Merge branch 'main' into layerwise-upcasting-hook
a0f1de71
update
e586ef38
norm.* -> norm
cfe63182
Merge branch 'main' into layerwise-upcasting-hook
9235f778
apply suggestions from review
76274151
add example
b9e12171
update hook implementation to the latest changes from pyramid attentiā¦
bde103c1
deinitialize should raise an error
64e6c9c9
update doc page
7037133c
Merge branch 'main' into layerwise-upcasting-hook
f1b46d69
stevhliu
approved these changes
on 2025-01-16
DN6
commented
on 2025-01-08
Apply suggestions from code review
390742b0
update docs
19901e7a
update
3ae32b47
refactor
bf797e77
Merge branch 'main' into layerwise-upcasting-hook
d22465a1
fix _always_upcast_modules for asym ae and vq_model
5956a9e3
DN6
approved these changes
on 2025-01-21
fix lumina embedding forward to not depend on weight dtype
93bd8eee
refactor tests
77a32a74
add simple lora inference tests
1335d7e1
_always_upcast_modules -> _precision_sensitive_module_patterns
a263e1a0
Merge branch 'main' into layerwise-upcasting-hook
93e36bab
remove todo comments about review; revert changes to self.dtype in unā¦
245137fe
check layer dtypes in lora test
b713511d
Merge branch 'main' into layerwise-upcasting-hook
4450b1c4
fix UNet1DModelTests::test_layerwise_upcasting_inference
ed14d260
_precision_sensitive_module_patterns -> _skip_layerwise_casting_patteā¦
2c9c33f3
skip test in NCSNppModelTests
08211f75
skip tests for AutoencoderTinyTests
59e04c3d
skip tests for AutoencoderOobleckTests
0a16826d
skip tests for UNet1DModelTests - unsupported pytorch operations
1d306b8e
layerwise_upcasting -> layerwise_casting
a9364bd3
skip tests for UNetRLModelTests; needs next pytorch release for curreā¦
c4d5a2b2
add layerwise fp8 pipeline test
d175d93d
use xfail
bf116912
DN6
approved these changes
on 2025-01-22
Apply suggestions from code review
1c523b29
Merge branch 'main' into layerwise-upcasting-hook
78033647
add assertion with fp32 comparison; add tolerance to fp8-fp32 vs fp32ā¦
376adf90
add note about memory consumption on tesla CI runner for failing test
719e8d39
a-r-r-o-w
merged
beacaa55
into main 1 year ago
a-r-r-o-w
deleted the layerwise-upcasting-hook branch 1 year ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub