Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
407 commits
Select commit Hold shift + click to select a range
22e959a
Add export_dtype parameter to `convert_to_hf` function (#2041)
idoh Nov 17, 2025
3819737
[compiler toolkit] Port joint_ac_pass from simplefsdp (#2051)
yiming0416 Nov 18, 2025
f8ab834
Allow more arbitrary dataset selection
janEbert Jan 14, 2025
339d8dd
Allow specifying arbitrary data files
janEbert Feb 26, 2025
076d706
Add gradient accumulation test
janEbert Feb 10, 2025
2652308
fix BitNet Llama's TP plan
janEbert Jan 14, 2025
1fd11c9
Add ByteTokenizer
janEbert Feb 21, 2025
1af0dd0
remove llama-2 spec
janEbert Feb 25, 2025
77b17f6
Add byte-tokenized Llama-3 train spec
janEbert Mar 20, 2025
c873c72
Allow padding vocab size By default, pad to multiple of 128
janEbert Feb 25, 2025
9002002
Allow setting vocab size from tokenizer
janEbert Aug 1, 2025
c812570
Add helper function for forward/param dtypes
janEbert Aug 8, 2025
bbefc8f
Expose additional data loader arguments
janEbert Mar 26, 2025
b7e854a
Add Byte tokenizer spec
janEbert Feb 25, 2025
d2eff6b
add bitnet opts, but disable it on job_config
janEbert Feb 25, 2025
c0959b1
Try to make BitNet work by patching PyTorch
janEbert Feb 25, 2025
4c0d83a
Refactor BitNet model code
janEbert Feb 27, 2025
dd1b6cb
Add QK normalization
janEbert Mar 13, 2025
76dab56
Add model config with QK normalization
janEbert Mar 13, 2025
c52d45d
Allow configuring norm layers
janEbert Aug 1, 2025
79bb06d
Move BitNet model to separate file
janEbert Mar 13, 2025
48d59c5
Add FineWeb data preset
rakkit Mar 14, 2025
413004e
Allow logging additional data
rakkit Mar 14, 2025
29462d9
Add MoE model
rakkit Aug 1, 2025
6833a20
Add Scion optimizer
rakkit Mar 14, 2025
37f20a1
Remove Scion arguments' defaults
ofivite Apr 10, 2025
81fde71
Add multi-token prediction objective
janEbert Mar 19, 2025
afcb80b
[mtp] Convert inputs and outputs to dictionaries
janEbert Mar 26, 2025
56a4cad
Make MTP loss work with other loss functions
janEbert Mar 26, 2025
fb6624b
Fix MoE loss with pipelining
janEbert Mar 26, 2025
28854b4
Refactor out sample packing
janEbert Mar 26, 2025
64715c2
Add data mixing
janEbert Mar 26, 2025
f70c77e
Adapt arguments for dataset mixing
janEbert Mar 28, 2025
20a1492
Allow placing norms after every layer
rakkit May 27, 2025
eae92d8
Support `norm_everywhere=True` model parallelism
rakkit Aug 1, 2025
42a8984
Add activation offloading
rakkit Apr 3, 2025
0c4403f
Allow separate seed for data
janEbert Apr 3, 2025
2b2e197
Add windowed data shuffling
janEbert Apr 3, 2025
7b2ae33
Seed RNG unconditionally
janEbert Apr 3, 2025
2bd927e
Save job config upon Trainer creation
janEbert Apr 3, 2025
2e621f2
Save model args
janEbert Apr 16, 2025
d3f924f
Use more legible JSON format
rakkit May 27, 2025
d8eb36e
Also dump training config as TOML
rakkit May 27, 2025
2bec9a9
Add KV cache
janEbert Apr 7, 2025
73cc61e
Expose more WandB settings
ofivite Apr 7, 2025
5782c09
Allow configuring separate param groups
ofivite Apr 9, 2025
198a31a
Explicitly pass `device_id` to `torch.distributed`
janEbert Apr 9, 2025
5ca2be4
Allow logging parameter and update norms
ofivite Apr 10, 2025
3928627
Fix MoE norm logging
rakkit Apr 11, 2025
519b406
Allow logging learning rates per parameter group
ofivite Apr 10, 2025
0cf1558
Add config for norm experiments
ofivite Apr 10, 2025
44041d7
Allow model initialization configuration
janEbert Apr 10, 2025
65925e5
Refactor updating model config from job config
janEbert Apr 10, 2025
40b905d
Expose initialization configurations
janEbert Apr 10, 2025
4b0e244
Use muP-like initialization in norm experiments
janEbert Apr 10, 2025
d9a7acd
Implement muP hyperparameter width multipliers
ofivite Apr 10, 2025
409cdcc
Disable weight decay for norm experiments
janEbert Apr 10, 2025
78751e7
Support Scion initialization scheme
ofivite Apr 10, 2025
6adaf01
Add debug model configs for scaling tests
ofivite Apr 10, 2025
85bc8ed
Change norm experiment momentum to Scion's default
ofivite Apr 10, 2025
8b658ca
Change model scaling debug configs
ofivite Apr 10, 2025
be7e14d
Prefer non-light Scion for norm experiments
ofivite Apr 10, 2025
edfb59d
Allow disabling gradient clipping
ofivite Apr 10, 2025
2837df5
Disable gradient clipping for norm experiments
ofivite Apr 10, 2025
732d330
Fix saving/loading for light Scion
janEbert Apr 11, 2025
5f3b30c
Fix expert parallelism
rakkit Apr 14, 2025
ddba03b
Expose `start_pos` argument
janEbert Apr 16, 2025
3f595e9
Add distributed Scion
rakkit Apr 16, 2025
11e6c16
Fix generator argument handling
janEbert Apr 19, 2025
a3dff1a
Pass `std` argument to orthogonal init
janEbert Apr 19, 2025
2fb7bc1
Fix scaling debug model configs
ofivite May 7, 2025
2d46c20
Allow logging condition numbers
ofivite May 7, 2025
e75d2d6
Do not include epsilon in sqrt for normalization
ofivite May 7, 2025
c7d2842
Fix light Scion and refactor
ofivite May 7, 2025
ce26df2
Add scaled orthogonal initialization
ofivite May 7, 2025
d55db43
Fix Scion initialization
ofivite May 7, 2025
fefbecd
Replace spectral with RMS->RMS norm logging
ofivite May 7, 2025
b545956
Improve condition number numerical stability
ofivite May 7, 2025
2e9c6ea
Make scaling-relevant configs part of model flavor
ofivite May 7, 2025
8f97e72
Allow disabling depth-wise residual init scaling
ofivite May 7, 2025
690fa16
Do not include epsilon in sqrt for normalization
janEbert May 7, 2025
7229cdf
Fix distributed Scion
rakkit May 7, 2025
d7ea4fe
Align APIs
rakkit May 7, 2025
8536216
Add padding token support to byte tokenizer
janEbert May 7, 2025
c7290b8
Add padding token support to model
janEbert May 7, 2025
499069b
Take `eos_id` from tokenizer if available
janEbert May 28, 2025
969ee72
Improve optimizer logging
rakkit May 26, 2025
7ceee8c
Allow configuring Scion backend
rakkit May 26, 2025
d23279b
Expose Scion backend config
rakkit May 26, 2025
587ee12
Force SVD dtype to float32
rakkit May 26, 2025
b4a4617
Adapt MoE debug config to Llama-3 and Scion
rakkit May 26, 2025
7df9af0
Support MoE router customization
rakkit May 26, 2025
b0eda73
Add image domain orthogonal initialization
rakkit May 27, 2025
5ea5e57
Add image domain norm for Scion
rakkit May 27, 2025
fa35fd9
Fix style
rakkit May 27, 2025
7930f92
Add debug model with QK norm
rakkit May 27, 2025
769d165
Do not modify routing parameter group for Scion
rakkit May 27, 2025
266a966
Disable experimental activation checkpointing
rakkit May 27, 2025
c1ce271
Update Scion implementation capabilities
rakkit May 27, 2025
f6a4f8b
Fix distributed Scion query for light version
rakkit May 27, 2025
0bc6216
Allow configuring LR when resuming from checkpoint
rakkit May 27, 2025
bb54204
Allow configuring whether LRs are loaded
janEbert May 27, 2025
197fb0a
Fix distributed Scion param group configuration
rakkit May 27, 2025
3fb2a59
Fix line lengths
janEbert May 27, 2025
6104063
Add baseline configs
rakkit May 29, 2025
c860620
Fix target shape indexing
rakkit May 29, 2025
8827d36
Refactor Scion backend selection
rakkit Jun 3, 2025
ba4fc11
Do not add empty parameter groups
rakkit Jun 3, 2025
d29c99b
Fix RNG state loading
rakkit Jun 4, 2025
d6eae15
Hackily force process exits upon completion
rakkit Jul 7, 2025
6f8d194
Hackily fix circular import error
rakkit Jul 7, 2025
7270b9b
Refactor optional Attention norms
rakkit Jul 7, 2025
dfef8f5
Change Scion parametrization settings
rakkit Jul 7, 2025
ad12138
Fix parameter norm calculation for MoE
rakkit Jul 7, 2025
4bfaf66
Fix Scion with MoE
rakkit Jul 7, 2025
bd804fc
Fix bug
rakkit Jul 7, 2025
8900c4b
Skip empty local tensors
rakkit Jul 7, 2025
758b0ee
Fix missing dependency for activation offloading
rakkit Jul 30, 2025
495338d
Fix typing and adjust argument names
janEbert Jul 30, 2025
bf527eb
Move `_wrap_orthogonal` into global scope
janEbert Jul 30, 2025
7e4fb7d
Refactor function selection in `build_init_fn`
rakkit Jul 30, 2025
7f97d10
Add no-communication version of `orthogonal_`
rakkit Jul 30, 2025
7a44aef
Handle `None` gradients
rakkit Jul 30, 2025
3688cc6
Refactor module name clean-up function
janEbert Jul 30, 2025
4ee293d
Refactor norm functions
janEbert Jul 30, 2025
af282ec
Refactor Scion
rakkit Jul 30, 2025
eb05c33
Refactor parameter norm calculations
rakkit Jul 30, 2025
9c69e53
Refactor orthogonal init
rakkit Jul 30, 2025
852b7f3
Add Polar Express backend
rakkit Jul 30, 2025
054feaf
Add additional norms
rakkit Aug 5, 2025
8ae25e8
Add SSNorm
rakkit Aug 5, 2025
811c63e
Fix SSNorm treatment with Scion
rakkit Aug 7, 2025
2dc586d
Allow setting `depth_init` from job config
rakkit Aug 5, 2025
d6ed876
Warn about regexp mismatch more prominently
rakkit Aug 5, 2025
ddd57e7
Add Scion gradient debugging mode
rakkit Aug 5, 2025
0358ed0
Allow configuring model residual scaling
rakkit Aug 6, 2025
bce1df8
Allow selection of logged norms
rakkit Aug 6, 2025
177ce3e
Skip model weight initialization if possible
janEbert Aug 7, 2025
c33316d
Fix and refactor distributed Scion
rakkit Aug 7, 2025
ceda66a
Log expert metrics if available
rakkit Aug 7, 2025
44df7ef
Add more model flavors
rakkit Aug 7, 2025
b2b3317
Add debug model flavors
rakkit Aug 8, 2025
c0a366f
Refactor MoE optimizer modifications
rakkit Aug 19, 2025
0101c62
Improve readability
janEbert Aug 19, 2025
62aa68b
Fix MoE model args
rakkit Aug 19, 2025
5e190f2
Refactor MoE pipelining
rakkit Aug 19, 2025
abb3468
Refactor setting layer ID
rakkit Aug 19, 2025
54e9d17
Move class to instance variables
janEbert Aug 19, 2025
8af93ef
Fix MoE weight initialization
rakkit Aug 19, 2025
18d5b50
Refactor expert parallelism
rakkit Aug 19, 2025
8973665
Remove temporary patches
rakkit Aug 19, 2025
f693ffe
Adjust MoE parallelization to upstream Llama-3
rakkit Aug 19, 2025
8a722b5
Refactor/fix MoE
rakkit Aug 19, 2025
17818f3
Log additional througput-related metrics
rakkit Aug 19, 2025
7f8a244
Remove legacy code
rakkit Aug 25, 2025
422869c
Delete unused code
rakkit Aug 25, 2025
7149767
Allow configuring activation function
rakkit Aug 25, 2025
e7a45bb
Add more MoE job configurations
rakkit Aug 25, 2025
48e5e68
Allow configuring MoE activation function
rakkit Aug 25, 2025
89a9491
Remove redundant pre-defined models
rakkit Aug 25, 2025
8ec93de
Remove initializing all MoE experts the same
rakkit Aug 25, 2025
6cbc5a5
Fix passing MoE arguments to sublayer
rakkit Aug 25, 2025
1526633
Remove option to use bias for routing
rakkit Aug 25, 2025
2429f61
Make sequence-wise auxiliary loss staticmethod
rakkit Aug 25, 2025
b6c291f
Fix docstring placement
janEbert Aug 25, 2025
c35d72d
Add zero-mean MoE router bias update options
rakkit Aug 25, 2025
6f5b02e
Fix `norm_everywhere=True` with tensor parallelism
rakkit Aug 25, 2025
08f96a9
Apply PP compatibility hacks
rakkit Aug 25, 2025
c936185
Optimize MoE metrics logging
rakkit Aug 25, 2025
64c797c
Make Scion work with biases and weight decay
rakkit Aug 26, 2025
593ba0d
Add optimized zeropower backend Triton kernels
rakkit Aug 26, 2025
80dd38f
Disable MoE aux loss by default
rakkit Aug 27, 2025
2e2887d
Fix Polar Express
rakkit Aug 27, 2025
ec1dd2d
Fix MoE bias update scale factor
rakkit Aug 27, 2025
7ae36b9
Log MoE router data only on DP+CP rank 0
janEbert Aug 27, 2025
db89486
Allow more selective logging
janEbert Aug 27, 2025
425d92f
Allow enabling logging for all first PP stages
rakkit Aug 27, 2025
5a2415c
Log Scion norms only on DP+CP rank 0
rakkit Aug 28, 2025
4f47c03
Do not pad vocab size by default
janEbert Aug 28, 2025
eb59b4d
Allow setting vocab size via CLI
janEbert Aug 28, 2025
20b9e72
Fix MoE logging
rakkit Sep 1, 2025
9cef2c6
Transpose grouped expert weights
rakkit Sep 1, 2025
8d4f82b
Prefer floor division for integer metrics
rakkit Sep 1, 2025
27dfd5f
Add forced uniform MoE load balancer prototype
rakkit Sep 1, 2025
6c16352
Rename "feed_forward" to "moe" when using MoE
rakkit Sep 8, 2025
2d4490e
Sync MoE parallelization upstream changes
rakkit Sep 8, 2025
7dc850f
Handle unsupported MoE+Scion setting
rakkit Sep 8, 2025
3ba66a2
Fix `depth_init` string comparisons
janEbert Sep 15, 2025
84f4099
sync some upstream changes. and rename .ffn of moe to .moe align up …
rakkit Sep 2, 2025
0d91a00
add conv code for scion, correct name to rms->inf
rakkit Sep 2, 2025
eecf257
Allow more arbitrary dataset selection
janEbert Jan 14, 2025
687ce31
Allow specifying arbitrary data files
janEbert Feb 26, 2025
7091669
Add multi-token prediction objective
janEbert Mar 19, 2025
ed6bd9a
Refactor out sample packing
janEbert Mar 26, 2025
ebb988d
refactor scion init, disable scion-light until we check it
rakkit Sep 2, 2025
06d839c
dont register optim hook for none moe model_parts, hack for PP+MoE
rakkit Sep 21, 2025
4cd7f54
fix some bugs
rakkit Sep 21, 2025
f46f412
improve readability of depth init
rakkit Sep 23, 2025
51756a7
improve readability of depth init
rakkit Sep 23, 2025
54ed8f1
rename it to disco.py
rakkit Oct 22, 2025
81f8e55
from upstream torchtitan/datasets/hf_datasets.py -> torchtitan/hf_dat…
rakkit Nov 19, 2025
9b5934f
fix pyproject for pre-commit
rakkit Nov 19, 2025
2ff168a
fix train.py and args after rebase. and refactor args
rakkit Nov 19, 2025
bdca787
disable compile in lmo's normlize
rakkit Nov 19, 2025
9043220
formating
rakkit Nov 19, 2025
5dfef2b
formating
rakkit Nov 19, 2025
e46db6e
refactor llama3 about args & config
rakkit Nov 19, 2025
e671ca5
refactor moe about args and config. taking upstream changes on moe
rakkit Nov 19, 2025
0321e3b
make sure PP not fail when splites accidently set to [ ]
rakkit Nov 19, 2025
1e54d4b
fix a strange import error and train spec cannot find modules due to …
rakkit Nov 19, 2025
9944c38
fix a strange import error and train spec cannot find modules due to …
rakkit Nov 19, 2025
416bb86
decouple moe's auxloss to make main loss be only CE, and rename aux l…
rakkit Nov 19, 2025
22384d3
refactor model init to allow explictly set head-dim and ffn/mode's di…
rakkit Nov 20, 2025
6e8c85d
stop github cli test
rakkit Nov 20, 2025
9b120e5
disable rope scaling for llama3 by default
rakkit Nov 26, 2025
67b846f
record models' dim-in and dim-out
rakkit Nov 26, 2025
fecaa44
save HF's format of config when convrt DCP to HF ckpt
rakkit Nov 26, 2025
b118d11
add MoEllama's DCP <-> HF convert
rakkit Nov 26, 2025
d937084
add HF assests for llama and MoEllama, when convert DCP to HF automat…
rakkit Nov 26, 2025
15cea0f
set checkpoint.last_save_model_only to *False* by default, to avoid w…
rakkit Nov 26, 2025
9e3da55
add data mixing scheduler
rakkit Nov 27, 2025
d0849fa
fix a bug that attention fail for 2x query head
rakkit Nov 27, 2025
a3eb33e
formating
rakkit Nov 27, 2025
0c3abd8
improving logging params and flops
rakkit Nov 28, 2025
55f247d
make prepare momentum works faster
rakkit Nov 28, 2025
1b9a56c
add proxy model
rakkit Dec 1, 2025
647d3e8
improve logging
rakkit Dec 1, 2025
ce9d75e
fix entropy logging
rakkit Dec 5, 2025
5c670a4
fix moe's ddp parallel func
rakkit Dec 6, 2025
8ad768e
disable skip weights init to makesure non persistent can be init from…
rakkit Dec 6, 2025
cdffae5
improve HF convert
rakkit Dec 9, 2025
0c6a480
fix load balance aux loss
rakkit Dec 10, 2025
6531a40
fix scion norm with compile, see https://github.com/SDLAML/disco/pull/1
rakkit Dec 11, 2025
add6147
refactor dataset a bit to clean legacy code
rakkit Dec 11, 2025
d56deb0
add sft dataset and patch for plelas synt data
rakkit Dec 13, 2025
8519e32
fix hf adapter for moe
rakkit Dec 15, 2025
2cef47a
better data mix logger
rakkit Dec 16, 2025
156c8c7
fix logger in dataset
rakkit Dec 17, 2025
436443e
improve and fix data sample
rakkit Dec 22, 2025
c2d3b10
add few more logging
rakkit Dec 22, 2025
839f8f2
cleanup sft dataset, will move to experiment later
rakkit Dec 24, 2025
b320fa0
revert compile in disco
rakkit Dec 24, 2025
dccec36
allow to set pre-fetch
rakkit Dec 24, 2025
e418299
improve and fix seed for data mix
rakkit Dec 24, 2025
305a020
fix sum_data_sampled all-reduce
rakkit Dec 25, 2025
bedf495
fix and improve logic of exhuasted datasets
rakkit Dec 31, 2025
6c09335
fix dataset state and log out the actual sample ratio
rakkit Dec 31, 2025
4290c95
fix and improve datalaoder
rakkit Jan 3, 2026
64cc998
fix on num_sampled_per_dataset state in loader
rakkit Jan 6, 2026
de0ce0e
improve hf model convert
rakkit Jan 6, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
12 changes: 10 additions & 2 deletions .ci/docker/build.sh
Original file line number Diff line number Diff line change
Expand Up @@ -13,14 +13,20 @@ shift
echo "Building ${IMAGE_NAME} Docker image"

OS=ubuntu
OS_VERSION=20.04
CLANG_VERSION=""
PYTHON_VERSION=3.11
PYTHON_VERSION=3.12
MINICONDA_VERSION=24.3.0-0

case "${IMAGE_NAME}" in
torchtitan-ubuntu-20.04-clang12)
OS_VERSION=20.04
CLANG_VERSION=12
BASE_IMAGE=nvidia/cuda:12.4.1-cudnn-runtime-ubuntu${OS_VERSION}
;;
torchtitan-rocm-ubuntu-22.04-clang12)
OS_VERSION=22.04
CLANG_VERSION=12
BASE_IMAGE=rocm/dev-ubuntu-${OS_VERSION}:latest
;;
*)
echo "Invalid image name ${IMAGE_NAME}"
Expand All @@ -30,6 +36,7 @@ esac
docker build \
--no-cache \
--progress=plain \
--build-arg "BASE_IMAGE=${BASE_IMAGE}" \
--build-arg "OS_VERSION=${OS_VERSION}" \
--build-arg "CLANG_VERSION=${CLANG_VERSION}" \
--build-arg "PYTHON_VERSION=${PYTHON_VERSION}" \
Expand All @@ -38,3 +45,4 @@ docker build \
-f "${OS}"/Dockerfile \
"$@" \
.

1 change: 1 addition & 0 deletions .ci/docker/common/install_conda.sh
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,7 @@ install_pip_dependencies() {
pip_install -r /opt/conda/requirements-dev.txt
pip_install -r /opt/conda/requirements.txt
pip_install -r /opt/conda/requirements-flux.txt
pip_install -r /opt/conda/requirements-vlm.txt
popd
}

Expand Down
1 change: 0 additions & 1 deletion .ci/docker/requirements-dev.txt
Original file line number Diff line number Diff line change
Expand Up @@ -2,5 +2,4 @@ expecttest==0.1.6
pytest==7.3.2
pytest-cov
pre-commit
tomli-w >= 1.1.0
transformers
2 changes: 2 additions & 0 deletions .ci/docker/requirements-vlm.txt
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
einops
pillow
2 changes: 2 additions & 0 deletions .ci/docker/requirements.txt
Original file line number Diff line number Diff line change
@@ -1,10 +1,12 @@
torchdata >= 0.8.0
datasets >= 3.6.0
tomli >= 1.1.0 ; python_version < "3.11"
tomli-w >= 1.1.0
tensorboard
tabulate
wandb
fsspec
tyro
tokenizers >= 0.15.0
safetensors
psutil
7 changes: 4 additions & 3 deletions .ci/docker/ubuntu/Dockerfile
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
ARG OS_VERSION
ARG BASE_IMAGE

FROM nvidia/cuda:12.4.1-cudnn-runtime-ubuntu${OS_VERSION}
FROM ${BASE_IMAGE}

ARG OS_VERSION

Expand Down Expand Up @@ -32,10 +32,11 @@ ENV PATH /opt/conda/envs/py_$PYTHON_VERSION/bin:/opt/conda/bin:$PATH
COPY requirements-dev.txt /opt/conda/
COPY requirements.txt /opt/conda/
COPY requirements-flux.txt /opt/conda/
COPY requirements-vlm.txt /opt/conda/
COPY conda-env-ci.txt /opt/conda/
COPY ./common/install_conda.sh install_conda.sh
COPY ./common/utils.sh utils.sh
RUN bash ./install_conda.sh && rm install_conda.sh utils.sh /opt/conda/requirements-dev.txt /opt/conda/requirements.txt /opt/conda/requirements-flux.txt /opt/conda/conda-env-ci.txt
RUN bash ./install_conda.sh && rm install_conda.sh utils.sh /opt/conda/requirements-dev.txt /opt/conda/requirements.txt /opt/conda/requirements-flux.txt /opt/conda/requirements-vlm.txt /opt/conda/conda-env-ci.txt

USER ci-user
CMD ["bash"]
2 changes: 1 addition & 1 deletion .github/CODEOWNERS
Original file line number Diff line number Diff line change
Expand Up @@ -10,4 +10,4 @@
/torchtitan/experiments/

# codeowners for experiments/forge
/torchtitan/experiments/forge/* @ebsmothers @pbontrager @joecummings @allenwang28 @tianyu-l @wwwjn
/torchtitan/experiments/forge/* @ebsmothers @pbontrager @joecummings @allenwang28 @tianyu-l @wwwjn @fegin
40 changes: 0 additions & 40 deletions .github/workflows/build_whl_and_publish.yaml

This file was deleted.

64 changes: 0 additions & 64 deletions .github/workflows/docker-builds.yml

This file was deleted.

52 changes: 0 additions & 52 deletions .github/workflows/integration_test_8gpu_features.yaml

This file was deleted.

52 changes: 0 additions & 52 deletions .github/workflows/integration_test_8gpu_flux.yaml

This file was deleted.

56 changes: 0 additions & 56 deletions .github/workflows/integration_test_8gpu_h100.yaml

This file was deleted.

Loading