Skip to content

Issues: AI-Hypercomputer/torchprime

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

45 Open 24 Closed
45 Open 24 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

[launcher] Specify an existing docker image URL that contains torchprime, for reproducibility. good first issue Good for newcomers
#141 opened Mar 3, 2025 by tengyifei
[Llama 3.1 405B] Optimize device ID assignment to avoid data shuffling after DCN all-gather
#140 opened Mar 3, 2025 by tengyifei
@tengyifei
Adopt apply_xla_patch_to_nn_linear to replace all nn.Linear ops with einsum
#139 opened Mar 3, 2025 by tengyifei
@tengyifei
Mixtral 8x7B perf data
#137 opened Mar 1, 2025 by cloudchrischan
Llama 3.1-70B perf data
#136 opened Mar 1, 2025 by cloudchrischan
Llama 3.1-8B perf data
#135 opened Mar 1, 2025 by cloudchrischan
Clarify how to use non xpk clusters
#134 opened Feb 28, 2025 by cloudchrischan
2
Performance of Llama 3.1 8B matches Huggingface fork
#133 opened Feb 28, 2025 by tengyifei Good quality Llama 3.1 8B and 70B in torch_xla_models
@zpcore
Add example datasets
#132 opened Feb 28, 2025 by cloudchrischan
Add a guide for how to pick the best sharding strategy
#131 opened Feb 28, 2025 by cloudchrischan
Add best practices or guides for troubleshooting the top 3 OOM errors
#130 opened Feb 28, 2025 by cloudchrischan
Address correctness guarantees
#129 opened Feb 28, 2025 by cloudchrischan
Print of co["Flops"] in train.py errors out - TypeError: string indices must be integers
#128 opened Feb 28, 2025 by brianchunkang
Document the trick of reshaping 64x4 device mesh into ring
#124 opened Feb 26, 2025 by tengyifei
@tengyifei
Able to persist logs into GCS good first issue Good for newcomers
#116 opened Feb 20, 2025 by tengyifei
Able to load a Huggingface dataset from a GCS cache
#115 opened Feb 20, 2025 by tengyifei
[torch_xla] Changing the sharding of model.embed_tokens.weight produces NaN gradients in Llama 3.1 405B
#114 opened Feb 18, 2025 by tengyifei
@tengyifei
Introduce a torch and torchax roller
#103 opened Feb 10, 2025 by tengyifei
Create nightly and stable release docker images
#99 opened Feb 8, 2025 by tengyifei
[Performance Report] What metrics to report?
#96 opened Feb 7, 2025 by zpcore
2
pytest --fork makes test easily hang
#94 opened Feb 7, 2025 by tengyifei
[torch_xla] MVP correctness check for Llama 3.0 8B
#90 opened Feb 6, 2025 by tengyifei Good quality Llama 3.1 8B and 70B in torch_xla_models
@zpcore
5
Bring up Stable Diffusion 2.1
#88 opened Feb 5, 2025 by tengyifei
Bring up Stable Diffusion 2.0
#87 opened Feb 5, 2025 by tengyifei
Hide ShardedModule from state dict
#82 opened Feb 5, 2025 by tengyifei
@tengyifei
ProTip! Mix and match filters to narrow down what you’re looking for.