-
Notifications
You must be signed in to change notification settings - Fork 321
Pull requests: AI-Hypercomputer/maxtext
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[config] Add a llama2-7b-int8 test config for decode
#1284
opened Feb 19, 2025 by
xy12181
Loading…
4 tasks done
Add flag quantize_only to skip decode steps during quantization
#1283
opened Feb 19, 2025 by
vivianrwu
Loading…
4 tasks done
Add llama3.3-70b-instruct model configs and HF golden logits
#1282
opened Feb 18, 2025 by
hengtaoguo
Loading…
4 tasks done
create separate emergency ckpt directories for each process
#1281
opened Feb 18, 2025 by
keshavb96
Loading…
Fix bugs in Pathways Path for Maxtext Benchmark Runner
#1280
opened Feb 18, 2025 by
SujeethJinesh
Loading…
4 tasks done
Prefix Caching with HBM and latency test
#1278
opened Feb 17, 2025 by
yuyanpeng-google
Loading…
4 tasks done
Link to maxtext_xpk_runner in readme for getting started with XPK
pull ready
#1275
opened Feb 14, 2025 by
Obliviour
Loading…
4 tasks done
fix: Pin orbax-checkpoint to v0.10.3 to resolve dependency error (#1273)
#1274
opened Feb 14, 2025 by
shota-inoue-lts
Loading…
Added support of training with NANOO fp8 GEMM on AMD MI300/MI325 GPUs.
#1262
opened Feb 11, 2025 by
wenchenvincent
•
Draft
3 of 4 tasks
Use causal mask type for training or prefill with fixed seqlen when
#1248
opened Feb 7, 2025 by
wenchenvincent
•
Draft
3 of 4 tasks
Enable mlperf to use mixtral-v1 tokenized dataset to avoid dropping in total_weights when splitting data across multiple host
#1246
opened Feb 7, 2025 by
aireenmei
Loading…
4 tasks done
[wip-prototype]chunked prefill on 1k prompt len
#1237
opened Feb 4, 2025 by
mailvijayasingh
Loading…
4 tasks
Add Pathways Recipe Support for Scale Testing (and fix bugs)
#1220
opened Jan 31, 2025 by
SujeethJinesh
•
Draft
4 tasks
[DRAFT] Initial commit Maxtext unit tests with Pathways backend.
#1211
opened Jan 28, 2025 by
RoshaniN
Loading…
4 tasks
Previous Next
ProTip!
What’s not been updated in a month: updated:<2025-01-19.