issues
search
google
/
paxml
Pax is a Jax-based machine learning framework for training large scale models. Pax allows for advanced and fully configurable experimentation and parallelization, and has demonstrated industry leading model flop utilization rates.
Apache License 2.0
458
stars
69
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
How to do gradient accumulation?
#104
gramesh-amd
opened
2 months ago
0
Change tensorflow and protobuf version
#103
Santiago-Castellano
closed
2 months ago
1
Question about Integration Timeline for PR #99
#102
Santiago-Castellano
opened
3 months ago
8
[NVIDIA] Enable GQA with cuDNN flash attention
#101
kaixih
closed
2 months ago
0
docs: update README.md
#100
eltociear
opened
3 months ago
0
Bump tensorflow from 2.9.3 to 2.12.1 in /paxml/pip_package in the pip group across 1 directory
#99
dependabot[bot]
opened
4 months ago
0
Add cuda profiler hook
#98
shawnwang18
opened
4 months ago
0
Bump the pip group in /paxml/pip_package with 2 updates
#97
dependabot[bot]
closed
4 months ago
1
LoRA support in transformers
#96
tanmayshishodia
opened
5 months ago
1
Error running tutorial notebook
#95
ajikmr
opened
5 months ago
0
Bump the pip group across 1 directory with 3 updates
#94
dependabot[bot]
closed
5 months ago
1
Add flag to use dpa in grok
#93
hx89
opened
5 months ago
0
Bump the pip group across 1 directory with 2 updates
#92
dependabot[bot]
closed
5 months ago
1
Support fractional PERCORE_BATCH_SIZE in synthetic dataset
#91
hx89
closed
6 months ago
0
Update README.md
#90
eltociear
opened
6 months ago
0
lingvo issue while installing paxml in vscode
#89
juneedpk
opened
6 months ago
0
Bump the pip group across 1 directory with 2 updates
#88
dependabot[bot]
closed
6 months ago
1
Add pipeline parallelism for Grok
#87
hx89
closed
6 months ago
1
update grok model param
#86
hx89
closed
6 months ago
1
Bump tensorflow from 2.9.3 to 2.11.1 in /paxml/pip_package in the pip group across 1 directory
#85
dependabot[bot]
closed
6 months ago
1
Bump the pip group across 1 directory with 3 updates
#84
dependabot[bot]
closed
6 months ago
1
[NVIDIA] Add LLaMA SFT and LoRA PEFT support
#83
ashors1
closed
6 months ago
1
[NVIDIA] Add MoE Configs
#82
ashors1
closed
6 months ago
0
Revert "[NVIDIA ] Remove references to deprecated XLA flags."
#81
laurentes
closed
7 months ago
0
[NVIDIA ] Remove references to deprecated XLA flags.
#80
Tixxx
closed
7 months ago
0
Added new config to nvidia.py
#79
abhinavgoel95
closed
7 months ago
0
adding new configs to nvidia.py
#78
abhinavgoel95
closed
7 months ago
2
DEADLINE_EXCEEDED on 1024 GPUs.
#77
mhugues
opened
7 months ago
0
Bump idna from 3.6 to 3.7 in /paxml/pip_package
#76
dependabot[bot]
closed
7 months ago
1
Adding support for expert parallelism
#75
abhinavgoel95
closed
7 months ago
0
Bump pillow from 10.2.0 to 10.3.0 in /paxml/pip_package
#74
dependabot[bot]
closed
7 months ago
1
[NVIDIA] Add config option to use cudnn flash attention
#73
kaixih
closed
4 months ago
2
Set offload checkpoint policy
#72
jaro-sevcik
opened
8 months ago
0
Jax + tpu and AQT int8 train model loss is abnormal
#71
Lisennlp
opened
9 months ago
0
[NVIDIA] Configure steps/sec interval to compute steps/sec only when writing a summary
#70
ashors1
closed
9 months ago
0
Make `compute_steps_per_sec_interval_steps` configurable
#69
ashors1
closed
9 months ago
0
[NVIDIA] Add LLaMA configs and scripts
#68
ashors1
closed
9 months ago
4
Set eval_loop_num_batches=-1 for lambada eval
#67
ashors1
closed
10 months ago
0
Use bfloat16 for eval
#66
tbaker2
opened
10 months ago
1
[Question] Very low MFU(30%~35%) when train bf16 Llama2 and GPT model with single SXM4 A100 machine.
#65
MoFHeka
opened
10 months ago
0
[Feature Request] Need ZeRo-1/2 to cooperate with PP+TP+DP. Which may more faster than FSDP sometimes.
#64
MoFHeka
opened
10 months ago
0
Bump notebook from 7.0.6 to 7.0.7 in /paxml/pip_package
#63
dependabot[bot]
closed
10 months ago
1
Bump jupyterlab from 4.0.10 to 4.0.11 in /paxml/pip_package
#62
dependabot[bot]
closed
10 months ago
1
[NVIDIA] Add synthetic configs for benchmarking
#61
ashors1
closed
10 months ago
0
Bump jupyter-lsp from 2.2.1 to 2.2.2 in /paxml/pip_package
#60
dependabot[bot]
closed
10 months ago
1
Bump jinja2 from 3.1.2 to 3.1.3 in /paxml/pip_package
#59
dependabot[bot]
closed
10 months ago
1
Remaining tutorials
#58
rahulbatra85
opened
11 months ago
0
Bump jupyter-server from 2.11.1 to 2.11.2 in /paxml/pip_package
#57
dependabot[bot]
closed
11 months ago
1
[NVIDIA] Add Llama2 configs
#56
ashors1
closed
1 year ago
0
[NVIDIA] Simplify the unit test for overwrite_with_gradient
#55
kaixih
closed
9 months ago
0
Next