Skip to content
Navigation Menu
Sign in
Appearance settings
Platform
AI CODE CREATION
GitHub Copilot
Write better code with AI
GitHub Copilot app
Direct agents from issue to merge
MCP Registry
Integrate external tools
DEVELOPER WORKFLOWS
Actions
Automate any workflow
Codespaces
Instant dev environments
Issues
Plan and track work
Code Review
Manage code changes
Code Quality
Enforce quality at merge
APPLICATION SECURITY
GitHub Advanced Security
Find and fix vulnerabilities
Code security
Secure your code as you build
Secret protection
Stop leaks before they start
EXPLORE
Why GitHub
Documentation
Blog
Changelog
Marketplace
View all features
Solutions
BY COMPANY SIZE
Enterprises
Small and medium teams
Startups
Nonprofits
BY USE CASE
App Modernization
DevSecOps
DevOps
CI/CD
View all use cases
BY INDUSTRY
Healthcare
Financial services
Manufacturing
Government
View all industries
View all solutions
Resources
EXPLORE BY TOPIC
AI
Software Development
DevOps
Security
View all topics
EXPLORE BY TYPE
Customer stories
Events & webinars
Ebooks & reports
Business insights
GitHub Skills
SUPPORT & SERVICES
Documentation
Customer support
Community forum
Trust center
Partners
View all resources
Open Source
COMMUNITY
GitHub Sponsors
Fund open source developers
PROGRAMS
Security Lab
Maintainer Community
Accelerator
GitHub Stars
Archive Program
REPOSITORIES
Topics
Trending
Collections
Enterprise
ENTERPRISE SOLUTIONS
Enterprise platform
AI-powered developer platform
AVAILABLE ADD-ONS
GitHub Advanced Security
Enterprise-grade security features
Copilot for Business
Enterprise-grade AI features
Premium Support
Enterprise-grade 24/7 support
Pricing
Type
/
to search
Sign in
Sign up
Appearance settings
You signed in with another tab or window.
Reload
to refresh your session.
You signed out in another tab or window.
Reload
to refresh your session.
You switched accounts on another tab or window.
Reload
to refresh your session.
Dismiss alert
{{ message }}
Uh oh!
There was an error while loading.
Please reload this page
.
NVIDIA
/
TransformerEngine
Public
Notifications
You must be signed in to change notification settings
Fork
795
Star
3.5k
Code
Issues
157
Pull requests
158
Discussions
Actions
Security and quality
0
Insights
Additional navigation options
Code
Issues
Pull requests
Discussions
Actions
Security and quality
Insights
[PyTorch/Jax] Fix attention mask definition, and sliding window for decoder
- #818
#818
Merged
ptrendx
merged 29 commits into
NVIDIA:main
NVIDIA/TransformerEngine:main
from
cyanguwa:fix_attn_masks
cyanguwa/TransformerEngine:fix_attn_masks
Copy head branch name to clipboard
May 17, 2024
Conversation
Commits
29
(29)
Checks
Files changed
Merged
[PyTorch/Jax] Fix attention mask definition, and sliding window for decoder
#818
ptrendx
merged 29 commits into
NVIDIA:main
NVIDIA/TransformerEngine:main
from
cyanguwa:fix_attn_masks
cyanguwa/TransformerEngine:fix_attn_masks
Copy head branch name to clipboard
Commits
Commits on Apr 26, 2024
fix inconsistency for attn mask; now True means participating in attn
Show description for fae6572
cyanguwa
committed
fae6572
View commit details
Copy full SHA for fae6572
Browse repository at this point
fix sliding window window_size for decoder+padding combination
Show description for dc6869d
cyanguwa
committed
dc6869d
View commit details
Copy full SHA for dc6869d
Browse repository at this point
Merge branch 'NVIDIA:main' into fix_attn_masks
cyanguwa
authored
cca5f49
View commit details
Copy full SHA for cca5f49
Browse repository at this point
Commits on May 1, 2024
revert paddle changes regarding mask
Show description for 81c4605
cyanguwa
committed
81c4605
View commit details
Copy full SHA for 81c4605
Browse repository at this point
Merge branch 'main' into fix_attn_masks
cyanguwa
authored
35de3be
View commit details
Copy full SHA for 35de3be
Browse repository at this point
revert softmax to 1-mask;0-keep
Show description for 4ca194c
cyanguwa
committed
4ca194c
View commit details
Copy full SHA for 4ca194c
Browse repository at this point
Commits on May 2, 2024
enforce 1-mask out; 0-keep rule for jax masks
Show description for 3a4b377
cyanguwa
committed
3a4b377
View commit details
Copy full SHA for 3a4b377
Browse repository at this point
fix jax lint
Show description for dfa0ece
cyanguwa
committed
dfa0ece
View commit details
Copy full SHA for dfa0ece
Browse repository at this point
revert pytorch mask changes; some kept in tests
Show description for aa6eaca
cyanguwa
committed
aa6eaca
View commit details
Copy full SHA for aa6eaca
Browse repository at this point
revert to jax fused attn on main
Show description for 44062a9
cyanguwa
committed
44062a9
View commit details
Copy full SHA for 44062a9
Browse repository at this point
inverse mask logic for get_cu_seqlens/_and_indices in PyTorch implementation and mask generation in unit tests
Show description for 1c8a073
cyanguwa
committed
1c8a073
View commit details
Copy full SHA for 1c8a073
Browse repository at this point
temporarily disable update_weight_scale_inv
Show description for 19f271c
cyanguwa
committed
19f271c
View commit details
Copy full SHA for 19f271c
Browse repository at this point
enforce window_size for decoder
Show description for 2a489bf
cyanguwa
committed
2a489bf
View commit details
Copy full SHA for 2a489bf
Browse repository at this point
add docstring for mask definition 1-mask out;0-keep
Show description for 87d02b6
cyanguwa
committed
87d02b6
View commit details
Copy full SHA for 87d02b6
Browse repository at this point
Merge branch 'main' into fix_attn_masks
cyanguwa
authored
ea06868
View commit details
Copy full SHA for ea06868
Browse repository at this point
Commits on May 14, 2024
Merge branch 'main' into fix_attn_masks
Show description for 0489d74
cyanguwa
authored
0489d74
View commit details
Copy full SHA for 0489d74
Browse repository at this point
add aux_ctx_tensors to save_for_backward
Show description for d759aee
cyanguwa
committed
d759aee
View commit details
Copy full SHA for d759aee
Browse repository at this point
Merge branch 'main' into fix_attn_masks
cyanguwa
authored
6b42b93
View commit details
Copy full SHA for 6b42b93
Browse repository at this point
Merge branch 'main' into fix_attn_masks
cyanguwa
authored
d1b9ecb
View commit details
Copy full SHA for d1b9ecb
Browse repository at this point
tweak make_decoder_mask and make_mask in jax tests
Show description for 2e84099
cyanguwa
committed
2e84099
View commit details
Copy full SHA for 2e84099
Browse repository at this point
Commits on May 15, 2024
skip dBias for shapes other than 1HSS; otherwise dq/dk/dv NaNs
Show description for 49a38f0
cyanguwa
committed
49a38f0
View commit details
Copy full SHA for 49a38f0
Browse repository at this point
expand attn_biases from list to variables in save_for_backward
Show description for 5bd7a1a
cyanguwa
committed
5bd7a1a
View commit details
Copy full SHA for 5bd7a1a
Browse repository at this point
fix use of variable before assignment in jax dact_lu
Show description for bf9a851
cyanguwa
committed
bf9a851
View commit details
Copy full SHA for bf9a851
Browse repository at this point
remove window size definition for decoder
Show description for 84aa282
cyanguwa
committed
84aa282
View commit details
Copy full SHA for 84aa282
Browse repository at this point
Merge branch 'main' into fix_attn_masks
cyanguwa
authored
3911898
View commit details
Copy full SHA for 3911898
Browse repository at this point
add change notes in README for padding mask in PyTorch
Show description for 37f25c3
cyanguwa
committed
37f25c3
View commit details
Copy full SHA for 37f25c3
Browse repository at this point
tweak padding mask notes in README
Show description for 87cea82
cyanguwa
committed
87cea82
View commit details
Copy full SHA for 87cea82
Browse repository at this point
Commits on May 16, 2024
Merge branch 'main' into fix_attn_masks
cyanguwa
authored
5e65f15
View commit details
Copy full SHA for 5e65f15
Browse repository at this point
expand list to tensors for save_for_backwards
Show description for f07b0bf
cyanguwa
committed
f07b0bf
View commit details
Copy full SHA for f07b0bf
Browse repository at this point
You can’t perform that action at this time.
