Skip to content
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.

Sign up for GitHub

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

Jump to bottom

Update baseline #1

Closed
wants to merge 190 commits into from
Closed

Update baseline #1

wants to merge 190 commits into from

Conversation

rickardp
Copy link
Owner

@rickardp rickardp commented Jun 7, 2023 •
edited
Loading

Sync with latest main branch from upstream

TimDettmers and others added 30 commits February 4, 2023 14:52
@TimDettmers
Added fp4 quant/dequant and dequant optimizations.
3ac5840
@TimDettmers
Forward matmul_fp4 tests pass.
160a835
@TimDettmers
Backward matmul_fp4 passes.
13c0a4d
@TimDettmers
Added matmul_fp4 to the benchmark.
cfe4705
@TimDettmers
Fixed matmul_fp4 transpose.
c361f84
@TimDettmers
Added bias test for LinearFP4 and basic test.
c0c352b
@TimDettmers
Added backprop test for Linear8bitLt and LinearFP4.
7f0773a
@TimDettmers
Fixed FP4 import and data type conversion in backward.
c93a90d
@TimDettmers
Added cast to fp4 layer for speed.
9851a10
@artidoro
t5 model fix
6c31a5f
@TimDettmers
Some small changes.
6981052
@TimDettmers
Added normal quant.
8645d1f
@TimDettmers
Added bf16 Adam.
c4cfe4f
@TimDettmers
Added 8-bit compression to quantization statistics.
51a21df
@TimDettmers
Generalized FP4 data type.
2dd5d69
@TimDettmers
Added normal with extra value.
0d332a6
@TimDettmers
Added quantization tree generation.
4ad999d
@TimDettmers
First draft of NF4.
64cc059
@TimDettmers
Refactor FP4 into 4Bit and integrate NF4 data type.
4ea489d
@TimDettmers
Fixed ParamsIn4 init; fixed PyTorch 2.0 test failure.
1ccb7bd
@TimDettmers
Some fixed for loading PEFT modules with Params4bit.
e9fa03b
@TimDettmers
Fixed bias conversion in Linear4bit
b8ea2b4
@TimDettmers
Added 32-bit optimizer for bfloat16 gradients.
7dc198f
@TimDettmers
Added nested quantization for blockwise quantization.
0f9d302
@rapsealk
fix: Replace libcudart with pytorch api
97b2567
@rapsealk
fix: Get CUDA compiled version through pytorch
eb54c55
@rapsealk
fix: Use raw int
9836b0b
@TimDettmers
Initial template.
6bfd7a4
@TimDettmers
Added cutlass example.
6e2544d
@TimDettmers
CUTLASS compiles.
84964db
poedator and others added 29 commits November 9, 2023 00:09
@poedator
partially reverted 76b40a5
781fcd5
@poedator
cleanup commented out deletions
070f45d
@poedator
remove moving offset if not nested
74c00eb
@poedator
type hints in Params4bit constructors
5486053
@TimDettmers
Merge pull request bitsandbytes-foundation#864 from poedator/save4_fixes
726f147 
fixes to recent PR bitsandbytes-foundation#753
@jph00
Avoid double-quantizing when calling cuda()
a403c0e
@poedator
fixes for init and tests
ffd46ce
@poedator
unshared nested_quant_map
de44be1
@poedator
test comment removed
4586426
@poedator
extra check for qs_dict['shape'] is None
079d7af
@akrentsel
Update env_vars.py to include GOOGLE_VM_CONFIG_LOCK_FILE
a49708c 
See this comment: bitsandbytes-foundation#620 (comment)

adding this variable to be ignored allows the bitsandbytes import and setup to proceed on a GCP VM in a jupyter notebook.
@theamdara
Update cuda_install.sh
e048dc9 
Add option to download and install Cuda 12.3
@Titus-von-Koeller
Merge pull request bitsandbytes-foundation#895 from theamdara/main
a409738 
Update cuda_install.sh
@Titus-von-Koeller
Merge pull request bitsandbytes-foundation#883 from akrentsel/patch-1
744d36f 
Update env_vars.py to include GOOGLE_VM_CONFIG_LOCK_FILE
@Titus-von-Koeller
Merge pull request bitsandbytes-foundation#867 from jph00/patch-2
2ee289f 
Avoid double-quantizing when calling `cuda()`
@Titus-von-Koeller
chore: update dev setup
2c605d0
@Titus-von-Koeller
Merge pull request bitsandbytes-foundation#896 from Titus-von-Koeller…
bbbed83 
…/improved_dev_setup

chore: update dev setup
@TimDettmers
Merge pull request bitsandbytes-foundation#868 from poedator/fix_1108
c8f564d 
Fix for 4bit without compress_statistics.
@TimDettmers
Added latest changes; bumped pip version.
31c84f9
@Titus-von-Koeller
Revert "Merge PR#867 from jph00/patch-2" due to failing PEFT integration
b601979 
tests.

This reverts commit 2ee289f, reversing
changes made to 744d36f.
@Titus-von-Koeller
Merge pull request bitsandbytes-foundation#904 from Titus-von-Koeller…
f2e4b6d 
…/fix-failing-peft-tests

Fix issue bitsandbytes-foundation#901: Failing PEFT integration tests by reversing PR#867 from jph00/patch-2
@Titus-von-Koeller
postfix release 0.41.3.post1
c75b6de
@Titus-von-Koeller
cuda installation tweaks
e3b318b
@Titus-von-Koeller
add pkg build + upload dependencies
8a45bfa
@younesbelkada
add stale bot (bitsandbytes-foundation#925)
2603cad
@younesbelkada
add bug report and feature request to bnb (bitsandbytes-foundation#924)
d88cd05
@younesbelkada
move to workflows (bitsandbytes-foundation#932)
f63abb5
@rickardp
Merge remote-tracking branch 'upstream/main' into apple-silicon-merge2
2801948
@rickardp
Post merge fix errors
84a6b78
@rickardp rickardp closed this Jan 2, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.