Skip to content
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.

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

TT-Buda-Demos Updates for TT-Buda Release (8/30) #123

Merged
merged 14 commits into from
Sep 24, 2024
Merged

Conversation

milank94
Copy link
Collaborator

Updated model configs and demos for 8/30 TT-Buda release.

Models Support Table

Model e75 e150 n150 n300 (single-chip) n300 (dual-chip) Supported Release
ALBERT ✔️ ✔️ ✔️ ✔️ v0.19.1
Autoencoder (convolutional) ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
Autoencoder (linear) ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
BeiT ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
BERT ✔️ ✔️ ✔️ ✔️ v0.19.1
CLIP ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
CodeGen ✔️ ✔️ ✔️ ✔️ v0.19.1
DeiT ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
DenseNet ✔️ ✔️ ✔️ ✔️ v0.19.1
DistilBERT ✔️ ✔️ ✔️ ✔️ v0.19.1
DLA ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
DPR ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
EfficientNet-Lite ✔️ ✔️ v0.19.1
Falcon-7B ✔️ ✔️ v0.19.1
FLAN-T5 ✔️ ✔️ ✔️ ✔️ v0.19.1
Fuyu-8B TBD
GhostNet ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
GoogLeNet ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
GPT-2 ✔️ ✔️ ✔️ ✔️ v0.19.1
GPT Neo ✔️ ✔️ ✔️ ✔️ v0.19.1
Hand Landmark ✔️ ✔️ v0.19.1
HardNet ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
HRNet ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
Inception-v4 ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
MLP-Mixer ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
MobileNetSSD ✔️ ✔️ v0.19.1
MobileNetV1 ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
MobileNetV2 ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
MobileNetV3 ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
Monodle ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
OpenPose ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
OPT ✔️ ✔️ ✔️ ✔️ v0.19.1
Pose Landmark ✔️ ✔️ v0.19.1
Perceiver IO ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
Phi2 ✔️ ✔️ v0.19.1
Qwen1.5 ✔️ ✔️ v0.19.1
ResNet ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
ResNeXt ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
RetinaNet ✔️ ✔️ ✔️ ✔️ v0.19.1
RoBERTa ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
SegFormer ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
SqueezeBERT ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
SSD300 ResNet50 ✔️ ✔️ ✔️ ✔️ v0.19.1
Stable Diffusion ✔️ ✔️ v0.19.1
T5 ✔️ ✔️ ✔️ ✔️ v0.19.1
U-Net ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
VGG ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
ViLT ✔️ ✔️ ✔️ ✔️ v0.19.1
ViT ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
VoVNet ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
WideResNet ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
Whisper ✔️ ✔️ ✔️ ✔️ v0.19.1
Xception ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
XGLM ✔️ ✔️ ✔️ ✔️ v0.19.1
YOLOv3 ✔️ ✔️ ✔️ ✔️ v0.19.1
YOLOv5 ✔️ ✔️ ✔️ ✔️ v0.19.1
YOLOv6 ✔️ ✔️ ✔️ ✔️ ✔️ v0.19.1
YOLOX ✔️ ✔️ ✔️ v0.19.1

kamalrajkannan78 and others added 10 commits August 29, 2024 00:03
* Qwen1.5 0.5B pybuda implementation

* remove unneeded requirement

* rename "acceleration" to "accelerate"

* Update env vars and compiler configs

* remove undefined device_map

* Remove misleading and unnecessary environment variables

* remove qwen from phi branch

* Add Phi 2

* Update requirements.txt

* Standardize Phi2 demo and added tests

* Remove old phi2 demo

* fix missing quote in pyproject.toml

* fix

* Fix test saying qwen1_5 instead of phi2
* Qwen1.5 0.5B pybuda implementation

* remove unneeded requirement

* Update env vars and compiler configs

* remove undefined device_map

* Remove misleading and unnecessary environment variables

* Refine qwen solution

* Rename qwen file

* Rename qwen filename and added qwen1.5-chat

* Add qwen1.5 test case

* Fix typo in pyproject.toml

* Disable dynamic caching

* Add extra whitespace below model title commment

* Fix typo "moadl" to "model"
Copy link

Our automated SPDX license verification process has discovered that the following files are missing a license header:

./model_demos/cv_demos/yolo_x/pytorch_yolox.py
./model_demos/tests/test_pytorch_yolox.py

Please ensure each indicated file includes a valid SPDX license identifier. This is essential for maintaining licensing compliance. Your attention and cooperation in updating these files are greatly appreciated. Thank you.

Copy link

Our automated SPDX license verification process has discovered that the following files are missing a license header:

./model_demos/cv_demos/yolo_x/pytorch_yolox.py
./model_demos/tests/test_pytorch_yolox.py

Please ensure each indicated file includes a valid SPDX license identifier. This is essential for maintaining licensing compliance. Your attention and cooperation in updating these files are greatly appreciated. Thank you.

1 similar comment
Copy link

Our automated SPDX license verification process has discovered that the following files are missing a license header:

./model_demos/cv_demos/yolo_x/pytorch_yolox.py
./model_demos/tests/test_pytorch_yolox.py

Please ensure each indicated file includes a valid SPDX license identifier. This is essential for maintaining licensing compliance. Your attention and cooperation in updating these files are greatly appreciated. Thank you.

Copy link
Contributor

@anirudTT anirudTT left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!
Changes reviewed and cross references models support table with internal excel sheet.

@JushBJJ
Copy link
Contributor

JushBJJ commented Sep 13, 2024

Curious, what were the problems Qwen and Phi were having with Wormhole chips?

@milank94
Copy link
Collaborator Author

Curious, what were the problems Qwen and Phi were having with Wormhole chips?

@JushBJJ here are the errors on Wormhole n150:

Phi2

FATAL    | Always          - The overlay blob for chip_0__y_18__x_22 (epoch:95,chip:0,y:18,x:22) does not fit, the max size is 65408, however we tried to allocate 78776
FATAL    | Always          - The overlay blob for chip_0__y_18__x_25 (epoch:5,chip:0,y:18,x:25) does not fit, the max size is 65408, however we tried to allocate 78776
RuntimeError: Backend compile failed!

Qwen1.5

FATAL    | Always          - The overlay blob for chip_0__y_21__x_25 (epoch:42,chip:0,y:21,x:25) does not fit, the max size is 65408, however we tried to allocate 76444
FATAL    | Always          - The overlay blob for chip_0__y_21__x_25 (epoch:30,chip:0,y:21,x:25) does not fit, the max size is 65408, however we tried to allocate 76444
FATAL    | Always          - The overlay blob for chip_0__y_23__x_18 (epoch:12,chip:0,y:23,x:18) does not fit, the max size is 65408, however we tried to allocate 76444
FATAL    | Always          - The overlay blob for chip_0__y_21__x_25 (epoch:36,chip:0,y:21,x:25) does not fit, the max size is 65408, however we tried to allocate 76444
RuntimeError: Backend compile failed!

Both of them fail during the backend compilation.

@anirudTT anirudTT merged commit 1f4dfc0 into main Sep 24, 2024
2 checks passed
@milank94 milank94 deleted the mkordic/rc_20240830 branch September 25, 2024 00:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants