issues
search
mosaicml
/
composer
Supercharge Your Model Training
http://docs.mosaicml.com
Apache License 2.0
5.06k
stars
407
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Don't use TP when `tensor_parallel_degree` is 1
#3437
snarayan21
opened
3 hours ago
0
Lower the system metrics logging frequency to reduce MLflow server's load
#3436
chenmoneygithub
closed
11 hours ago
2
Relax hf hub pin
#3435
dakinggg
opened
1 day ago
0
Correctly process `parallelism_config['tp']` when it's a dict
#3434
snarayan21
opened
1 day ago
5
Bump CI testing version
#3433
mvpatel2000
opened
1 day ago
0
Remove MosaicMLLambdaEvalClient
#3432
aspfohl
opened
1 day ago
0
Remove save overwrite
#3431
mvpatel2000
closed
1 day ago
2
Fixes to TP Docs
#3430
snarayan21
closed
15 hours ago
1
Remove CodeQL workflow
#3429
mvpatel2000
closed
2 days ago
0
Report job status to mlflow logger
#3428
chenmoneygithub
opened
3 days ago
0
Add precision change to bf16 when using fp8 eval
#3427
j316chuck
opened
3 days ago
0
Skip HSDP + TP pytests that require torch 2.3 or above
#3426
KuuCi
closed
3 days ago
0
Bumping MLflow version to 2.14.1
#3425
JackZ-db
closed
4 days ago
0
Update psutil requirement from <6,>=5.8.0 to >=5.8.0,<7
#3424
dependabot[bot]
closed
4 days ago
0
Bump deepspeed from 0.8.3 to 0.14.4
#3423
dependabot[bot]
closed
4 days ago
1
Bump coverage[toml] from 7.5.3 to 7.5.4
#3422
dependabot[bot]
closed
4 days ago
0
Training stops after first pass of Evaluation when pretraining MosaicBert
#3421
amishparekh
opened
5 days ago
4
Fix style
#3420
b-chu
closed
1 week ago
0
Patch PyTorch 2.3.1
#3419
mvpatel2000
closed
1 week ago
0
Fixes some typing issues
#3418
dakinggg
closed
1 week ago
0
Restore dev version
#3417
karan6181
closed
1 week ago
0
Add support for variable length dataloaders in DDP
#3416
JAEarly
closed
4 days ago
0
Support DDP with rank-dependent dataloader lengths
#3415
JAEarly
closed
4 days ago
2
Bump version v0.23.3
#3414
karan6181
closed
1 week ago
2
Bump version v0.23.3
#3413
karan6181
closed
1 week ago
1
Move pillow dep as required
#3412
mvpatel2000
closed
1 week ago
0
Computing train metrics at a given frequency
#3411
Ghelfi
opened
1 week ago
1
fixing mlflow logging to Databricks workspace file paths with /Shared/ prefix
#3410
JackZ-db
closed
1 week ago
2
CPU tests image fix
#3409
snarayan21
closed
1 week ago
0
Revert "Optionally use `flash-attn`'s CE loss for metrics (#3394)"
#3408
snarayan21
closed
1 week ago
0
Add setter for epoch in iteration
#3407
b-chu
closed
1 week ago
0
Update numpy requirement from <1.27.0,>=1.21.5 to >=1.21.5,<2.1.0
#3406
dependabot[bot]
closed
1 week ago
1
Bump deepspeed from 0.8.3 to 0.14.3
#3405
dependabot[bot]
closed
1 week ago
1
Add pynvml to mlflow dep group
#3404
dakinggg
closed
1 week ago
0
Test pytorch patch
#3403
j316chuck
closed
3 days ago
0
Add missing import for PyTorch 2.3.1 device mesh slicing
#3402
mvpatel2000
closed
2 weeks ago
0
Add buffering time to mlflow logger
#3401
chenmoneygithub
closed
2 weeks ago
2
Check for 'CUDA error: out of memory' when auto-microbatching
#3400
JAEarly
closed
2 weeks ago
0
Save checkpoint to disk for API with new save layout
#3399
eracah
closed
1 week ago
0
Simplify launcher world size parsing
#3398
mvpatel2000
closed
1 week ago
6
CUDA OOM error not caught with auto microbatching
#3397
JAEarly
closed
2 weeks ago
3
Busy wait utils in dist
#3396
dakinggg
closed
2 weeks ago
0
Remove FSDP restriction from PyTorch 1.13
#3395
mvpatel2000
closed
2 weeks ago
1
Optionally use `flash-attn`'s CE loss for metrics
#3394
snarayan21
closed
1 week ago
4
Skip extra dataset state load
#3393
mvpatel2000
closed
2 weeks ago
0
Update packaging requirement from <24.1,>=21.3.0 to >=21.3.0,<24.2
#3392
dependabot[bot]
closed
2 weeks ago
0
Bump cryptography from 42.0.6 to 42.0.8
#3391
dependabot[bot]
closed
2 weeks ago
0
Bump pytest from 7.4.4 to 8.2.2
#3390
dependabot[bot]
closed
2 weeks ago
1
Only requires `databricks-sdk` when inside the Databricks platform
#3389
antoinebrl
closed
2 weeks ago
0
Restore dev version
#3388
bigning
closed
3 weeks ago
0
Next