Skip to content

Issues: NVIDIA/TransformerEngine

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

173 Open 257 Closed
173 Open 257 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

[BUG] Weight gradients with TransformerEngine v2.1 don't match those with TransformerEngine v1.12 bug Something isn't working tp_overlap
#1616 opened Mar 26, 2025 by okoge-kaz
1
[BUG] Wrong attention gradient in Transformer Engine bug Something isn't working
#1615 opened Mar 26, 2025 by i-love-megatron
fp8 can not set pp>1
#1604 opened Mar 22, 2025 by mdy666
2
Can we only replace part of nn.Linear with te.Linear and others keep unchanged?
#1595 opened Mar 20, 2025 by zigzagcai
3
Does Transformer Engine support FP8 Grouped GEMM?
#1594 opened Mar 20, 2025 by zigzagcai
1
How to debug tex.fused_attn_bwd getting cuDNN Error: [cudnn_frontend] Error: No execution plans support the graph bug Something isn't working
#1591 opened Mar 19, 2025 by Ir1d
1
did not get improvement from tp/sp overlap
#1587 opened Mar 18, 2025 by artetaout
cuBLAS Error
#1585 opened Mar 18, 2025 by wangli68
5
curious about atomic_gemm_overlap_ag
#1583 opened Mar 17, 2025 by ehion
Does TransformerEngine support FP8 communication such like all-gather or all-to-all?
#1579 opened Mar 14, 2025 by zigzagcai
2
Is it necessary to perform layer replacement on te.xx? If not, is it effective to use te.fp8.autocast directly
#1556 opened Mar 11, 2025 by wangli68
2
Install earlier version 1.13
#1553 opened Mar 9, 2025 by velocirraptor23
@ksivaman
7
Context parallelism with MLA
#1552 opened Mar 8, 2025 by SuperCB
@cyanguwa @xrennvidia
4
When I import the package ’transformer_engine.pytorch‘, the error message is as follows
#1541 opened Mar 6, 2025 by wangli68
@ksivaman
1
How can we use te.Linear with weight parallel?
#1532 opened Mar 4, 2025 by zigzagcai
3
Causal mask ignored in DotProductAttention good first issue Good for newcomers
#1524 opened Feb 28, 2025 by anthony-Neo
@KshitijLakhani @phu0ngng
1
How can we integrate the DeepGEMM Fp8 GEMM implementation in TE's block-wise scaling？
#1509 opened Feb 26, 2025 by BolongLin
4
Failed to build transformer-engine
#1506 opened Feb 25, 2025 by sfwu2003
6
Question about the performace of GroupedLinear performance Performance issues
#1499 opened Feb 20, 2025 by XLzed
5
Float8Quantizer::create_tensor calculates scale_inv instead of creating an empty buffer performance Performance issues
#1491 opened Feb 18, 2025 by yaox12
3
qwen1.5-0.5B failed to save model with huggingface transformers bug Something isn't working
#1482 opened Feb 13, 2025 by xinpengzz
2
When flash-attn >2.6.1, use context parallel will cause error
#1467 opened Feb 9, 2025 by south-ocean
@cyanguwa
1
How to set NVTE_FWD/BWD_LAYERNORM_SM_MARGIN?
#1459 opened Feb 5, 2025 by cailun01
1
If no Windows support is planned for 5090 as it wasn't for 4090, pass along to corporate to mention this in advertisement
#1436 opened Jan 29, 2025 by NeedsMoar
4
HF Accelerate FP8 use more gpu memory then FP16 in training LLM
#1429 opened Jan 28, 2025 by Liufeiran123
1
ProTip! Updated in the last three days: updated:>2025-03-23.