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

27 Open 8 Closed
27 Open 8 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

DeepSeek prefill test fails under torch 2.6.0
#75 opened Feb 3, 2025 by tengyifei
Remove generated xplane_pb2.py
#74 opened Feb 3, 2025 by tengyifei
[torch_xla] Performance of Llama 3.1 405B matches Huggingface fork on 1 pod
#73 opened Feb 1, 2025 by tengyifei
@tengyifei
1
tp log to follow the logs of a task
#68 opened Jan 31, 2025 by tengyifei
Calculate step duration and MFU after a training run is done
#67 opened Jan 31, 2025 by tengyifei
@tengyifei
Test torchprime on multiple TPU pods
#66 opened Jan 30, 2025 by bhavya01
Extract common model layers in a separate directory and test each one
#65 opened Jan 30, 2025 by bhavya01
@bhavya01
[torchax] Initialize the model weights with the right distribution for each weight
#58 opened Jan 28, 2025 by tengyifei
End-to-end tests
#52 opened Jan 25, 2025 by tengyifei
1
Mixtral 8x22B is implemented and optimized
#45 opened Jan 21, 2025 by tengyifei
@bhavya01
Performance of Mixtral 8x7B matches the Huggingface fork
#44 opened Jan 21, 2025 by tengyifei
@bhavya01
[torch_xla_models] Integrate scan
#41 opened Jan 15, 2025 by tengyifei
[torch_xla_models] Llama 3.1 8B convergence testing
#40 opened Jan 15, 2025 by tengyifei
[torch_xla] Training loss is NaN under the AdamW optimizer
#36 opened Jan 14, 2025 by tengyifei
XPK experiment affordances
#35 opened Jan 14, 2025 by tengyifei
Address some painpoints in the xpk launcher
#33 opened Jan 14, 2025 by tengyifei
4 of 7 tasks
@tengyifei
Add TPU CI
#32 opened Jan 14, 2025 by tengyifei
@tengyifei
Investigate Torchax and native PyTorch output error on Llama model
#30 opened Jan 11, 2025 by zpcore
Study and impl MaxText's per-host dataloading
#23 opened Jan 9, 2025 by tengyifei
@bhavya01
Support both FSDP and 2D sharding with the same trainer
#22 opened Jan 9, 2025 by tengyifei
@tengyifei
Support changing the sharding from a model config
#21 opened Jan 9, 2025 by tengyifei
@tengyifei
1
[torch_xla_models] Trained model is evaluated using a common eval dataset
#20 opened Jan 9, 2025 by tengyifei Good quality Llama 3.1 8B and 70B in torch_xla_models
[torch_xla_models] Train different Llama variants from our own config files
#19 opened Jan 9, 2025 by tengyifei Good quality Llama 3.1 8B and 70B in torch_xla_models
@bhavya01
Enable minibatch dataloading in torch_xla_models
#18 opened Jan 9, 2025 by tengyifei Good quality Llama 3.1 8B and 70B in torch_xla_models
@bhavya01
Performance of Llama 3.1 70B matches Huggingface fork
#17 opened Jan 9, 2025 by tengyifei Good quality Llama 3.1 8B and 70B in torch_xla_models
@zpcore
ProTip! no:milestone will show everything without a milestone.