Skip to main content

Browse the archive

https://github.com/facebookresearch/pythia
31 March 2023, 09:22:40 UTC
sort by:
Revision Author Date Message Commit Date
1ea94d2 ryan-qiyu-jiang 16 December 2021, 23:18:48 UTC Update on "[docs] Update UNITER project doc" Update doc to include VILLA citation and feature discrepancy explaination. Differential Revision: [D33064276](https://our.internmc.facebook.com/intern/diff/D33064276) [ghstack-poisoned] 16 December 2021, 23:18:48 UTC
5cb8004 ryan-qiyu-jiang 16 December 2021, 23:18:48 UTC Update base for Update on "[docs] Update UNITER project doc" Update doc to include VILLA citation and feature discrepancy explaination. Differential Revision: [D33064276](https://our.internmc.facebook.com/intern/diff/D33064276) [ghstack-poisoned] 16 December 2021, 23:18:48 UTC
1746def ryan-qiyu-jiang 16 December 2021, 22:22:48 UTC [docs] Add pytorchvideo docs (#1164) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1164 Add pytorchvideo tutorial docs for using a pytorchvideo model as an encoder through the TorchVideoEncoder class. Test Plan: Imported from OSS **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D32773734/V8/mmf/)| |**Modified Pages**| |[docs/tutorials/pytorchvideo](https://our.intern.facebook.com/intern/staticdocs/eph/D32773734/V8/mmf/docs/tutorials/pytorchvideo/)||[docs/tutorials/pytorchvideo](https://our.intern.facebook.com/intern/staticdocs/eph/D32773734/V7/mmf/docs/tutorials/pytorchvideo/)||[docs/tutorials/pytorchvideo](https://our.intern.facebook.com/intern/staticdocs/eph/D32773734/V6/mmf/docs/tutorials/pytorchvideo/)||[docs/tutorials/pytorchvideo](https://our.intern.facebook.com/intern/staticdocs/eph/D32773734/V5/mmf/docs/tutorials/pytorchvideo/)||[docs/tutorials/pytorchvideo](https://our.intern.facebook.com/intern/staticdocs/eph/D32773734/V4/mmf/docs/tutorials/pytorchvideo/)||[docs/tutorials/pytorchvideo](https://our.intern.facebook.com/intern/staticdocs/eph/D32773734/V3/mmf/docs/tutorials/pytorchvideo/)| Reviewed By: apsdehal Differential Revision: D32773734 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: d4d2d78d9366299b0ba923a4e0ac42d9c990cd61 16 December 2021, 22:24:59 UTC
68add70 ryan-qiyu-jiang 16 December 2021, 22:22:48 UTC [feat] Add pytorchvideo encoder wrapper (#1156) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1156 Add an encoder class that constructs any pytorchvideo model from config, and uses this model for its forward pass. Can load pretrained or random init models, based on config. Test Plan: Tested through unit tests on slowfast50 and mvit. Will be tested end-to-end when datasets and transformers are available in mmf ``` (torchvideo) ryanjiang@learnfair5083:~/copy/mmf$ pytest tests/models/test_mmf_transformer.py ================================================== test session starts ================================================== platform linux -- Python 3.7.11, pytest-6.2.5, py-1.10.0, pluggy-1.0.0 rootdir: /private/home/ryanjiang/copy/mmf plugins: forked-1.3.0, timeout-1.4.2, hydra-core-1.1.1, xdist-2.4.0, dash-2.0.0 collected 15 items tests/models/test_mmf_transformer.py ............... [100%] (torchvideo) ryanjiang@learnfair5083:~/copy/mmf$ pytest tests/modules/test_encoders.py ================================================== test session starts ================================================== platform linux -- Python 3.7.11, pytest-6.2.5, py-1.10.0, pluggy-1.0.0 rootdir: /private/home/ryanjiang/copy/mmf plugins: forked-1.3.0, timeout-1.4.2, hydra-core-1.1.1, xdist-2.4.0, dash-2.0.0 collected 12 items tests/modules/test_encoders.py ............ [100%] ``` Reviewed By: apsdehal Differential Revision: D32631207 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 6b549162f7ae9ccea162563e48ed910618a6da54 16 December 2021, 22:24:59 UTC
ee19bd9 ryan-qiyu-jiang 16 December 2021, 22:22:48 UTC [feat] Add standard poolers (#1173) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1173 Add identity, cls, and average poolers. Registered and instantiable through registry. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D33032556 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: ddccba75f75ba61972760600903cd75810545ff6 16 December 2021, 22:24:58 UTC
f3ede0e ryan-qiyu-jiang 16 December 2021, 16:27:39 UTC [chore] Update numpy version (#1177) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1177 Set numpy version range to [1.16.6, 1.21.4] This fixes website documentation deployment. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D33071993 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: fd2b1c8c8376516dab39b482472397313fd3d5e1 16 December 2021, 16:28:26 UTC
e7092de ryan-qiyu-jiang 14 December 2021, 21:46:29 UTC [fix] Fix UNITER unit tests on gpu (#1170) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1170 Send sample list to cuda device when available. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D33002411 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: ce3eccdc67e4ebdc4e3fd84a9d0335eaa32e7bab 14 December 2021, 21:47:29 UTC
6c443af ryan-qiyu-jiang 14 December 2021, 18:56:54 UTC [docs] Update NOTICES (#1167) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1167 ViLT, UNITER, Oscar, Scene Graph Benchmark License Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D32851104 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 62a0a9601d48a5f12f3c2d5215e6120c299421b9 14 December 2021, 18:57:46 UTC
262dc40 ryan-qiyu-jiang 13 December 2021, 17:33:19 UTC [docs] Update UNITER project doc Update doc to include VILLA citation and feature discrepancy explaination. [ghstack-poisoned] 13 December 2021, 17:33:19 UTC
8e3413c ryan-qiyu-jiang 13 December 2021, 17:33:16 UTC [fix] Update UNITER VILLA checksums Update UNITER and VILLA checksums for new checkpoints that include empty configs. [ghstack-poisoned] 13 December 2021, 17:33:16 UTC
d578f79 ryan-qiyu-jiang 11 December 2021, 02:00:34 UTC [feat] Add VinVL model wrapper (#1151) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1151 Add VinVL BaseModel for training and testing from MMF. This model defers to either the classification or pretraining model depending on its config. For an example config consult the project dir or unit tests. Test Plan: ### Unit Tests Tested BaseModel instantiation from config, and forward pass for classification and pretraining. Reviewed By: ebsmothers Differential Revision: D32574738 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 50f8396821effd778c6d5184cd940864fc1eb3b1 11 December 2021, 02:01:42 UTC
a51b977 ryan-qiyu-jiang 11 December 2021, 02:00:34 UTC [feat] Add VinVL classification and pretraining models (#1150) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1150 Add VinVL classification and pretraining models that use the VinVL BertImgModel trunk. These are nn.Module objects, usable outside of MMF. For example text preprocessing for pretraining, take a look at VinVLTextTokenizer in a later diff. Models forward returns dict with scores and losses. For example usage consult the unit tests or VinVL basemodel. Test Plan: ### Unit Tests Tested forward passes for classification and pretraining models. Pretraining model forward was tested in end-to-end on winoground dataset. Reviewed By: apsdehal Differential Revision: D32574735 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: fc1a58db421a33d941b1ddbb5f5a3f35e308e741 11 December 2021, 02:01:41 UTC
8e67391 ryan-qiyu-jiang 11 December 2021, 02:00:34 UTC [feat] Add VinVL three-way contrastive head (#1171) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1171 Add contrastive head for three-way vinvl loss, which sometimes randomly swaps captions or labels with the task of classifying whether image, text, label triples are correct, or have a swapped text/label. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D33001604 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: b49fa93a18b8b80f16d5fe28963bfa39100ebd0d 11 December 2021, 02:01:41 UTC
9f712e4 ryan-qiyu-jiang 11 December 2021, 02:00:34 UTC [feat] Add VinVL Base Model (#1149) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1149 Add BertImgModel model used in VinVL. The model is a nn.Module based off github.com/microsoft/Oscar/ modified for mmf. This model is usable without MMF trainer. BaseModel class and pretraining to come. Test Plan: ### Unit tests Tests the instantiation and forward pass of the BertImgModel. Used in VinVL classification and pretraining by models in later diffs. Reviewed By: apsdehal Differential Revision: D32574736 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: d43e1670cab9c91533d7e6dfc0cdfd7ff7cb59c3 11 December 2021, 02:01:41 UTC
7be6c7f ryan-qiyu-jiang 10 December 2021, 17:18:29 UTC [test] Fix pooler unit tests on gpu (#1165) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1165 Fix no attribute "cuda" runtime error when testing poolers with cuda. Unit tests pass on github actions as those machines run without gpu. Test Plan: Test unit tests with GPU on learnfair Reviewed By: ebsmothers Differential Revision: D32805045 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 5d2272fc99688d71bbf9cd3b7b71c051deb1de36 10 December 2021, 17:19:46 UTC
6f3f40f Sethu Sankaran 08 December 2021, 20:20:53 UTC ReFNet changes: Use encoders for refiner pooler inputs and fix edge case for MS loss calculation when no negative pairs found. (#1160) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1160 Loss value dict needs to be initialized and set to default values when no negatives per-batch (for MS loss calculation). Also, the refiner class should take in encoded layers rather than sequence output. Reviewed By: ebsmothers Differential Revision: D32696676 fbshipit-source-id: b52d8532445141499152353c7893fb83ef6142c4 08 December 2021, 20:22:09 UTC
cd6f58f Siyu Wang 08 December 2021, 00:16:11 UTC Update update_requirements Summary: from https://www.internalfb.com/intern/wiki/Pytorch_Ecosystem_Foundation_(EcoF)/PyTorch_Lightning/Operations/Sync_OSS_FBCode/ Reviewed By: ananthsub Differential Revision: D32933988 fbshipit-source-id: 60d9054d7c1f6951910a0892e3001f26930a16f5 08 December 2021, 00:17:43 UTC
2e1b8cb ryan-qiyu-jiang 02 December 2021, 19:14:30 UTC [docs] Add UNITER model to website (#1144) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1144 Add citation and training instructions under projects/uniter. Test Plan: Viewed doc changes on website. **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D32313967/V8/mmf/)| |**Modified Pages**| |[docs/projects/uniter](https://our.intern.facebook.com/intern/staticdocs/eph/D32313967/V8/mmf/docs/projects/uniter/)| |[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D32313967/V8/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D32313967/V7/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D32313967/V6/mmf/docs/notes/model_zoo/)| Reviewed By: ebsmothers Differential Revision: D32313967 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: e2faed15e396064b4f8944dc577d7838284da97e 02 December 2021, 19:15:44 UTC
27d6db3 ryan-qiyu-jiang 02 December 2021, 19:14:30 UTC [feat] Add UNITER yaml config (#1128) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1128 Add configs for UNITER vqa2 finetuning, and README in projects. Test Plan: Similar configs were used end-to-end to run on another dataset Reviewed By: ebsmothers Differential Revision: D31768456 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 9083ef5eaa174eddc2efdcfc295b5f17bd97de45 02 December 2021, 19:15:43 UTC
7c3b527 ryan-qiyu-jiang 02 December 2021, 19:14:30 UTC [feat] Add UNITER model wrapper (#1127) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1127 Add UNITER BaseModel for training and testing from MMF. This model defers to either the classification or pretraining model depending on its config. For an example config consult the project dir or unit tests. Test Plan: ### Unit Tests Tested forward passes for classification and pertaining from mmf basemodel build from config. ### End to End Model tested end-to-end using butd extracted features on winoground. Will work on converting caffe feature extraction to pytorch so we can add Uniter and Villa checkpoints usefully. Reviewed By: ebsmothers Differential Revision: D31768457 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: b311419f4b1431a2cf8bb5322bd08d80e8a883c3 02 December 2021, 19:15:43 UTC
426de65 ryan-qiyu-jiang 02 December 2021, 19:14:30 UTC [feat] Add classifier and pretraining UNITER models (#1133) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1133 Add UNITER classification and pretraining models that use UNITERBaseModel. The pretraining model chooses a task per batch from MLM, ITM, MRC, MRFR, and WRA. These are nn.Modules usable outside MMF. Models forward returns dict with scores and losses. For example usage consult the unit tests or UNITER BaseModel. Test Plan: ### Unit Tests Tested forward passes for classification and pretraining models. Pretraining model forward was tested in end-to-end on winoground dataset Reviewed By: ebsmothers Differential Revision: D31865996 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 533e90fe4eec189af55de422f9ae5b2b866f6cac 02 December 2021, 19:15:43 UTC
8ba414a Jennifer Dai 02 December 2021, 09:23:39 UTC [PyTorchLightning/pytorch-lightning] Refactor progress bar initialization to avoid extra attribute set on Trainer (#10553) Summary: ### New commit log messages 2c7c4aab8 Refactor progress bar initialization to avoid extra attribute set on Trainer (#10553) Reviewed By: ananthsub Differential Revision: D32543265 fbshipit-source-id: 4a40350f412418230f6cafa71abd02daff1cbdce 02 December 2021, 09:24:39 UTC
f569969 Jennifer Dai 01 December 2021, 03:31:42 UTC [PyTorchLightning/pytorch-lightning] Fix loop examples after Accelerator API removals (#10514) Summary: ### New commit log messages 98de69b14 Fix loop examples after Accelerator API removals (#10514) Reviewed By: ananthsub Differential Revision: D32541936 fbshipit-source-id: 83a4d3d3f2b204cc7c626a7825134382f9e13d51 01 December 2021, 03:33:09 UTC
b672a74 ryan-qiyu-jiang 26 November 2021, 22:48:08 UTC [feat] Add UNITER text processor (#1132) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1132 Add uniter_text_tokenizer which adds 'input_ids_masked' to the sample list seperate from 'input_ids'. Test Plan: **Unit tests** Tests for the construction of the processor and processor output. We assert that the tokens are correct and contain the enhanced fields required for UNITER. Reviewed By: ebsmothers Differential Revision: D31865997 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: cf8f8b312aebe39bdfe3d5f5516831e99e27888b 26 November 2021, 22:49:27 UTC
9564983 ryan-qiyu-jiang 26 November 2021, 22:48:08 UTC [feat] Add UNITER pretraining heads (#1126) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1126 Add MRC, MRFR, WRA heads for UNITER pretraining. MRC = Masked Region Classification. MRFR = Masked Region Feature Regression. WRA = Word Region Alignment. Heads forward return a dict with `losses`. These heads can be used as pretraining tasks for other VL models. Details at https://arxiv.org/abs/1909.11740 Test Plan: **Unit tests** Test direct instantiation and forward pass for each head. Instantiation through build() and configs is tested in unit tests in later diffs by the models that use these heads. Tested as part of UNITER pretraining on masked COCO Reviewed By: ebsmothers Differential Revision: D31768455 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 9b48f81c472cd1859f32bc813484296208e206f5 26 November 2021, 22:49:27 UTC
6fba5f6 ryan-qiyu-jiang 26 November 2021, 22:48:08 UTC [feat] Add UNITER model (#1115) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1115 Add nn.Module model from github.com/ChenRocks/UNITER modified for mmf. Model is usable without MMF trainer. BaseModel class and pretraining to come. Test Plan: **Unit tests** Tests instantiation and forward pass of the UNITERModelBase. Used in UNITER finetuning and pretraining by models in later diffs. Reviewed By: ebsmothers Differential Revision: D31768458 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 499e610085c729e3573c5c12b0a81328580ab19a 26 November 2021, 22:49:27 UTC
d1df04c ryan-qiyu-jiang 24 November 2021, 05:22:08 UTC [docs] Add ViLT model to website (#1142) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1142 Test Plan: Imported from OSS **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D32103419/V6/mmf/)| |**Modified Pages**| |[docs/projects/vilt](https://our.intern.facebook.com/intern/staticdocs/eph/D32103419/V6/mmf/docs/projects/vilt/)| |[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D32103419/V6/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D32103419/V5/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D32103419/V4/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D32103419/V3/mmf/docs/notes/model_zoo/)| Reviewed By: apsdehal Differential Revision: D32103419 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 9a6f43eb3e575ede1a01cf684bbcd6122eb98e8d 24 November 2021, 05:23:05 UTC
9be685d ryan-qiyu-jiang 24 November 2021, 05:22:08 UTC [feat] Add pretraining support to ViLT (#1106) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1106 Add support for ITM and MLM pretraining. Test Plan: To finetune a pretrained model on vqa2 ```MMF_DATA_DIR=/private/home/asg/pythia_dev/data mmf_run config=projects/vilt/configs/vqa2/vit_b16_224.yaml run_type=train_val dataset=vqa2 model=vilt training.find_unused_parameters=True ``` To pretrain on masked coco ``` MMF_DATA_DIR=/checkpoint/ronghanghu/pythia/data mmf_run config=projects/vilt/configs/masked_coco/pretrain.yaml run_type=train_val dataset=masked_coco model=vilt training.find_unused_parameters=True training.batch_size=256 ``` **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D31117424/V20/mmf/)| |**Modified Pages**| |[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D31117424/V20/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D31117424/V19/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D31117424/V18/mmf/docs/notes/model_zoo/)||[docs/notes/model_zoo](https://our.intern.facebook.com/intern/staticdocs/eph/D31117424/V17/mmf/docs/notes/model_zoo/)| Reviewed By: apsdehal Differential Revision: D31117424 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 79ab08be81449e00a8b92648fb4b7c379c2eb10a 24 November 2021, 05:23:05 UTC
38a202b ryan-qiyu-jiang 24 November 2021, 05:22:08 UTC [feat] Add ViLT yaml configs (#1103) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1103 Add yaml configs for ViLT training on vqa2 Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D31117423 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 139041bfc291a8fbe3305e67fff460576e1ab2fe 24 November 2021, 05:23:05 UTC
e1eb0be ryan-qiyu-jiang 24 November 2021, 05:22:08 UTC [feat] Add ViLT model (#1102) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1102 Add ViLT model and unit tests. ViLT model from megaVLT, using ViT from huggingface. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D31117427 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 7616a8294562340ce506525ed167816f1bc421e0 24 November 2021, 05:23:05 UTC
f4a2ccc ryan-qiyu-jiang 24 November 2021, 05:22:08 UTC [feat] Add HeadsDict transformer head combiner (#1141) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1141 Add HeadsDict util class to correctly build and run a forward pass on many heads per task. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D32103420 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 478327140376494cfe9051dc424f184b1aa968ca 24 November 2021, 05:23:05 UTC
4a53fad ryan-qiyu-jiang 24 November 2021, 05:22:08 UTC [feat] Add ViLT image and text embeddings (#1101) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1101 Add ViT image and Bert text embedding encoders for ViLT model. Test Plan: Test plan is identical to ViTEncoder and ViT modules commits. Unit test, tests instantiation with pretrained weights of ViLT image and text embeddings and performs a forward pass. Backwards compatibility with older transformer versions is handled through ViT modules. Unit tests are skipped with older transformer versions. Reviewed By: apsdehal Differential Revision: D31117428 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 2732d56ee5ed78ab144572b67c1a74a0fee55ea1 24 November 2021, 05:23:04 UTC
c9ab349 Ayush Thakur 23 November 2021, 17:59:22 UTC [feat] Model version control using W&B Artifacts (#1137) Summary: 🚀 I have extended the `WandbLogger` with the ability to log the `current.pt` checkpoint as W&B Artifacts. Note that this PR is based on top of this [PR](https://github.com/facebookresearch/mmf/pull/1129). ### What is W&B Artifacts? > W&B Artifacts was designed to make it effortless to version your datasets and models, regardless of whether you want to store your files with us or whether you already have a bucket you want us to track. Once you've tracked your dataset or model files, W&B will automatically log each and every modification, giving you a complete and auditable history of changes to your files. Through this PR, W&B Artifacts can help save and organize machine learning models throughout a project's lifecycle. More details in the documentation [here](https://docs.wandb.ai/guides/artifacts/model-versioning). ### Modification This PR adds a `log_model_checkpoint` method to the `WandbLogger` class in the `utils/logger.py` file. This method is called in the `utils/checkpoint.py` file. ### Usage To use this, in the `config/defaults.yaml` do, `training.wandb.enabled=true` and `training.wandb.log_checkpoint=true`. ### Result The screenshot shows the `current.pt` checkpoints saved at intervals defined by `training.checkpoint_interval`. You can check out the logged artifacts page [here](https://wandb.ai/ayut/mmf/artifacts/model/run_ey9xextf_model/0dc64164acbdc300fd01/api). ![image](https://user-images.githubusercontent.com/31141479/139390462-d5c8445e-5c20-4fdd-85d0-51ef64846bf0.png) ### Superpowers With this small addition, now one can easily track different versions of the model, download a checkpoint of interest by using the API in the API tab, easily share the checkpoints with teammates, etc. ### Requests This is a draft PR as there are a few more things that can be improved here. * Is there a better way to access the path to the `current.pt` checkpoint? Rather is the modification made to `utils/checkpoint.py` an acceptable way of approaching this? * While logging a file as W&B artifacts we can also provide metadata associated with that file. In this case, we can add current iteration, training metrics, etc. as the metadata. Would love to get suggestions about the different data points that I should log as metadata alongside the checkpoints. * How to determine if a checkpoint is the best one? If a checkpoint is best I can add `best` as an alias for that checkpoint's artifact. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1137 Test Plan: Imported from GitHub, without a `Test Plan:` line. **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D32402090/V6/mmf/)| |**Modified Pages**| |[docs/notes/logger](https://our.intern.facebook.com/intern/staticdocs/eph/D32402090/V6/mmf/docs/notes/logger/)| Reviewed By: apsdehal Differential Revision: D32402090 Pulled By: ebsmothers fbshipit-source-id: 94b881ec55c4197301331d571bc926521e2feecc 23 November 2021, 18:00:14 UTC
b6a5804 Ananth Subramaniam 18 November 2021, 18:42:49 UTC [PyTorchLightning/pytorch-lightning] remove deprecated train_loop (#10482) Summary: ### New commit log messages fa0ed17f8 remove deprecated train_loop (#10482) Reviewed By: kandluis Differential Revision: D32454980 fbshipit-source-id: a35237dde06cc9ddac5373b75992ce88a6771c76 18 November 2021, 18:44:25 UTC
2af7d2c Alisson Gusatti Azzolini 17 November 2021, 06:12:15 UTC [PyTorchLightning/pytorch-lightning] Remove deprecated accelerator pass through functions in Accelerator (#10403) Summary: ### New commit log messages f9b9cdb0d Remove deprecated accelerator pass through functions in Accelerator (#10403) Reviewed By: edward-io Differential Revision: D32261339 fbshipit-source-id: c6696154be5e349cd1de1796ba396325ae06b831 17 November 2021, 06:13:22 UTC
b0b2af7 Evan Smothers 15 November 2021, 17:33:09 UTC [chore][cleanup] Add ufmt formatter and format code (#1139) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1139 Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D32036762 Pulled By: ebsmothers fbshipit-source-id: 7e97d28357961f315038fe30c92d05e0973f0d2f 15 November 2021, 17:34:19 UTC
974d4ad Zohar Barzelay 14 November 2021, 06:56:36 UTC [MMF] assert when processor isn't registered (#1143) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1143 when a processor isn't registered, raise an error specifying its name Reviewed By: apsdehal Differential Revision: D31903309 fbshipit-source-id: 7024baf1a1a73ed8f68984b99d234fe69eed9df7 14 November 2021, 06:57:38 UTC
e6f5116 Edward Wang (EcoF) 10 November 2021, 23:47:14 UTC [PyTorchLightning/pytorch-lightning] Remove deprecated dataloader arguments in Trainer methods (#10325) (#575) Summary: Pull Request resolved: https://github.com/facebookresearch/ReAgent/pull/575 ### New commit log messages 412f0a4d2 Remove deprecated dataloader arguments in Trainer methods (#10325) Reviewed By: tangbinh Differential Revision: D32261342 fbshipit-source-id: 0dc24bb64eeb186f722ba147aa569d2b8af63f84 10 November 2021, 23:48:47 UTC
5817cff ryan-qiyu-jiang 04 November 2021, 23:31:44 UTC [feat] Add UNITER embeddings (#1114) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1114 Add image embeddings over image features used by the UNITER V-L model. Test Plan: **UNITer Tests** haha Unit test instantiates uniter image embedding module and checks for a successful forward pass. **UNITER training** UNITER finetuning on vqa2 and pretraining on masked_coco seem run with decreasing loss. Haven't attempted to replicated paper numbers yet. Reviewed By: ebsmothers Differential Revision: D31768454 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: d70f69677c5c5809833336853e779f151c9f5976 04 November 2021, 23:33:23 UTC
bf17d14 ryan-qiyu-jiang 02 November 2021, 17:16:55 UTC [feat] Add ViT encoder (#1100) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1100 Add encoder for loading pretrained ViTModel weights from huggingface. Wrapper around ViTModel construction and loading pretrained weights from huggingface. Test Plan: **Unit Test.** Tests ViTEncoder, which loads pretrained weights for the `google/vit-base-patch16-224` ViTModel from huggingface tests that the forward pass is successful. Unit test is skipped in older transformer versions, preserving backwards compatibility for unit tests. **Modules used in ViLT Training** ViTEncoder is used in the ViLT model which successfully performs pertaining and finetuning, ViLT model tested using mmf_run on VQA2. **Tested Backwards Compatibility** ViTEncoder backward compatibility comes from ViTModel backwards compatibility, and throws useful exceptions to update transformers when instantiated with transformers version < 4.5.0. Reviewed By: ebsmothers, apsdehal Differential Revision: D31117425 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: bec474dc6ccc35e260b2bc810e7cac48f1358544 02 November 2021, 17:18:00 UTC
c8c3d6d ryan-qiyu-jiang 02 November 2021, 16:04:54 UTC [feat] Add ViT modules (#1099) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1099 Add modified huggingface ViT modules from megaVLT supporting attention masks Test Plan: **Unit Test.** Tests ViTModel, which uses all ViT submodules. Test checks if ViTModel can successfully perform a forward pass. Unit test is skipped in older transformer versions, preserving backwards compatibility for unit tests. **Modules used in ViLT Training** ViT modules are used in the ViLT model which successfully performs pretraining and finetuning, ViLT model tested using mmf_run on VQA2. **Tested Backwards Compatibility** ViT modules backward compatibility with previous versions of transformers tested directly with * 3.4.0 (throws helpful upgrade transformers version for ViT related modules) * 4.5.0 (__spec missing error patched in previous transformers version bump), and * 4.10.1 Reviewed By: ebsmothers Differential Revision: D31117429 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 9a473eb5d7e0183711d8d19dcd8f2b77b7792e03 02 November 2021, 16:06:23 UTC
5c904c2 Sethu Sankaran 29 October 2021, 14:24:36 UTC [feat] Add a refiner head to MMFT (#1136) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1136 We add a refiner module that adds a shallow decoder to reconstruct target embeddings per modality (based on e.g. modality specific pooling). This is based off this paper - https://arxiv.org/abs/2104.03435 Reviewed By: ebsmothers Differential Revision: D31936897 fbshipit-source-id: 6f272b28b9bff704bf82b66ac4c21a8af10f14be 29 October 2021, 14:30:29 UTC
dabf95f Sethu Sankaran 26 October 2021, 19:18:43 UTC Revert D30704069: [feat] Add a refiner head that can be used with MMFT Differential Revision: D30704069 (https://github.com/facebookresearch/mmf/commit/efa98757ccbecc4eb6a17532a38209679b7eb60b) Original commit changeset: 94e31778bd7b fbshipit-source-id: e6c6ffe345861b8bf05556c5d8f87613fc1263ba 26 October 2021, 19:19:31 UTC
efa9875 Sethu Sankaran 26 October 2021, 17:20:19 UTC [feat] Add a refiner head that can be used with MMFT Summary: We add a refiner module that adds a shallow decoder to reconstruct target embeddings per modality (based on e.g. modality specific pooling). This is based off this paper - https://arxiv.org/abs/2104.03435 Reviewed By: ebsmothers Differential Revision: D30704069 fbshipit-source-id: 94e31778bd7b622700a4a52921a2e52e48be228b 26 October 2021, 17:21:32 UTC
b5cbc5d Shashank Chaudhry 24 October 2021, 03:34:34 UTC [NOOP][clangformat][codemod] Enable CLANGFORMAT Reviewed By: zertosh Differential Revision: D31877463 fbshipit-source-id: 2c0bb28e7b86ff9ad51dcdbae2566f7ef2a3e094 24 October 2021, 03:35:40 UTC
347d2a8 Animesh Sinha 22 October 2021, 14:45:54 UTC [fix] Remove SizeProportionalIterationStrategy for val and test splits (#1130) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1130 Pull Request resolved: https://github.com/fairinternal/mmf-internal/pull/191 We'd like to have the same iteration strategy for all dataset splits (train, val, test). Test reporter is used to iterate over datasets one by one rather than using any iteration strategy. Reviewed By: apsdehal Differential Revision: D31740149 fbshipit-source-id: 87a413dfcdb098a15a1eca84ae5637a0414bc408 22 October 2021, 14:47:10 UTC
26485eb ryan-qiyu-jiang 20 October 2021, 14:20:43 UTC [fix] Undo replace_with_jit in unit tests (#1107) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1107 Fix bug where monkey patch on transformers leaked into other unit tests resulting in ViLT unit tests passing in isolation but failing due to unexpected signatures when run with other tests. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D31149844 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: ae015110924888a43efe24f1311166a8dbd1436d 20 October 2021, 14:21:36 UTC
c172069 ryan-qiyu-jiang 20 October 2021, 01:06:51 UTC [chore] Bump transformers version to 4.5.1 (#1098) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1098 Bumping transformers version for ViT module. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D31117430 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 5cee5f4f344e809b0c8cfb854d627dab2571a300 20 October 2021, 01:07:55 UTC
ca25dc3 ryan-qiyu-jiang 20 October 2021, 00:03:38 UTC [feat] Add ViLT image text processors (#1097) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1097 Add ViT image processor and bert text processor. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D31117422 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 978cb2d45857cd75331802b51504d4fa6a7be39c 20 October 2021, 00:04:34 UTC
4930ffb Emmanuel Ndubuisi 19 October 2021, 06:29:16 UTC [Docs] Update Broken Installation Link (#1125) Summary: Update the broken installation link in README.md from https://mmf.sh/docs/getting_started/installation to https://mmf.sh/docs/. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1125 Reviewed By: ebsmothers Differential Revision: D31737841 Pulled By: apsdehal fbshipit-source-id: 448a42a293fab54774d08d459909a7a3b4739515 19 October 2021, 06:30:05 UTC
ab137a6 Amanpreet Singh 16 October 2021, 01:22:43 UTC [fix,chores] Lint and wandb (#1123) Summary: - Fixes lint issues introduced by recent PR - Fixes wandb issue as reported and fixed https://github.com/facebookresearch/mmf/issues/1088 as author is unresponsive. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1123 Test Plan: CI **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D31663376/V3/mmf/)| |**Modified Pages**| |[docs/notes/concepts](https://our.intern.facebook.com/intern/staticdocs/eph/D31663376/V3/mmf/docs/notes/concepts/)||[docs/notes/concepts](https://our.intern.facebook.com/intern/staticdocs/eph/D31663376/V2/mmf/docs/notes/concepts/)| Reviewed By: ebsmothers Differential Revision: D31663376 Pulled By: apsdehal fbshipit-source-id: d8aa4acadbc462824e0290c81eda3e39ce135e09 16 October 2021, 01:23:39 UTC
e2da181 Sethu Sankaran 14 October 2021, 19:11:35 UTC Pooler functions of transformer encoder layers to support Refiner (#1122) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1122 Modules for different pooler functions from transformer encoder layers to support Refiner (D30704069) and possibly other applications. Reviewed By: apsdehal Differential Revision: D31507765 fbshipit-source-id: 787eed551560088d531486601cf97c985688210a 14 October 2021, 19:12:49 UTC
bbbd99f Amanpreet Singh 14 October 2021, 01:39:08 UTC [chores] Rename master to main everywhere (#190) Summary: Pull Request resolved: https://github.com/fairinternal/mmf-internal/pull/190 Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1121 Renaming master references to main at most places as part of terminology shift. Reviewed By: ebsmothers Differential Revision: D31623714 fbshipit-source-id: a8052d9606373f5bfb7c9ad6bb78628bbfa220e7 14 October 2021, 02:40:40 UTC
582c719 ryan-qiyu-jiang 08 October 2021, 23:49:36 UTC [feat] Add ViLT transformer heads (#1096) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1096 Add multilayer_mlp head for ViLT. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D31117426 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 94b364b37e421643bd2e88911059b6a0a41c54b5 08 October 2021, 23:50:57 UTC
9bb60da Danielle Pintz 30 September 2021, 16:46:37 UTC [PyTorchLightning/pytorch-lightning] Remove `ABC` from `LightningModule` (#9517) Summary: ### New commit log messages 3aba9d16a Remove `ABC` from `LightningModule` (#9517) Reviewed By: ananthsub Differential Revision: D31296721 fbshipit-source-id: a9992486c61a6f86fb251f2733bbc9311d93f293 30 September 2021, 16:47:44 UTC
57d29b0 Amanpreet Singh 29 September 2021, 23:02:06 UTC [fix] bypassing the checkpoint mismatch (#1112) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1112 This was broken in recent diffs. Fixing it to support this option again. Reviewed By: ebsmothers Differential Revision: D31068271 fbshipit-source-id: c03a955b3c2425372b1b2412a9686ed1823aac30 29 September 2021, 23:03:17 UTC
842c4d1 ryan-qiyu-jiang 29 September 2021, 13:13:50 UTC [fix] Update zoo urls (#1111) Summary: Update mmimdb url and skip flickr30k in unit-tests. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1111 Reviewed By: ebsmothers Differential Revision: D31223313 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 096bfac81c5b77a0a166ca770a8e69912af476db 29 September 2021, 13:14:49 UTC
7b733ed Ronghang Hu 27 September 2021, 23:20:50 UTC [fix] fix TPU checkpoint saving, parameter broadcasting, and misc stuff (#954) Summary: Fix the following on TPU training: - instead of using `xm.save`, add a custom `save_xla_ckpt` to avoid crashing on lr_scheduler state dict; **note that if we later switch back to xm.save, we need to set `global_master=True` in xm.save** - only delete old checkpoint from master process when using XLA - only open file in "wb" mode from the master process when using XLA. (Otherwise, concurrent writing to the same file in `with PathManager.open(ckpt_filepath, "wb")` from multiple processes causes permission denied error). - only save `config.yaml` from the global master process. - skip batch size validation in XLA mode in evaluation loop (since data loader already automatically drops the last batch in XLA mode) - On TPUs, there isn't an equivalent to DistributedDaraParallel yet, so adding a manual parameter broadcasting at the beginning of training (`broadcast_xla_master_model_param`). Otherwise we will have different parameters on each process since we are explicitly setting different seeds in different processes in MMF. - set `drop_last=True` in DistributedSampler for XLA to correctly drop the last batch in XLA mode (the current codebase only set `drop_last=True` in DataLoader, which is not enough). Pull Request resolved: https://github.com/facebookresearch/mmf/pull/954 Test Plan: tested on GCP. Reviewed By: apsdehal Differential Revision: D29415896 Pulled By: ronghanghu fbshipit-source-id: 2826feb8774b5bfc557e2495608a55e73afe54a4 27 September 2021, 23:21:53 UTC
8c1bf37 Diana Liskovich 20 September 2021, 21:30:04 UTC Update reference from master to main elsewhere in fbcode Summary: Update reference from master to main elsewhere in fbcode Reviewed By: alexeib Differential Revision: D30938472 fbshipit-source-id: 243b98550207f241c9d3265bf3d4060350aaf0a8 20 September 2021, 22:30:23 UTC
fe63eb0 ryan-qiyu-jiang 15 September 2021, 14:20:32 UTC [docs] Add loss tutorial doc (#1087) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1087 Test Plan: Imported from GitHub, without a `Test Plan:` line. **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D30876325/V4/mmf/)| |**Modified Pages**| |[docs/tutorials/losses](https://our.intern.facebook.com/intern/staticdocs/eph/D30876325/V4/mmf/docs/tutorials/losses/)| Reviewed By: apsdehal Differential Revision: D30876325 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: d0d2ab47a2149a3585d080c601066c7375a77024 15 September 2021, 14:23:33 UTC
357ee45 ryan-qiyu-jiang 15 September 2021, 14:19:05 UTC [docs] Add metrics tutorial doc (#1083) Summary: Add a tutorial doc for custom metrics. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1083 Test Plan: Imported from GitHub, without a `Test Plan:` line. **Static Docs Preview: mmf** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D30851769/V4/mmf/)| |**Modified Pages**| |[docs/tutorials/metrics](https://our.intern.facebook.com/intern/staticdocs/eph/D30851769/V4/mmf/docs/tutorials/metrics/)| Reviewed By: apsdehal Differential Revision: D30851769 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: deb1c2811ff3e8e9a5668e7b0586c478e9c5aa39 15 September 2021, 14:20:11 UTC
610c142 ryan-qiyu-jiang 15 September 2021, 14:11:49 UTC [feat] Add loss support for dict return type (#1086) Summary: Add loss support for classes that return a dict of losses. All loss dictionaries are flattened in the final losses dict. Added bce_kl, mse_mae loss classes as examples and for unit testing. This new loss class returns a dict, containing the bce and kl div tensors as values with string keys. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1086 Reviewed By: apsdehal Differential Revision: D30872944 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: 4116ff92bcb24439a20d4720c3813d3c3db6ced0 15 September 2021, 14:12:49 UTC
ce870b1 ryan-qiyu-jiang 13 September 2021, 13:29:35 UTC [feat] Add metric support for dict return type (#1075) Summary: Add metric support for classes that return a dict of metrics. All metric dictionaries are flattened in the final metrics dict. Add example f1_precision_recall metric class as an example, and for unit testing. This new metric class returns a dict, containing the f1, precision, and recall tensors as values with string keys. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1075 Reviewed By: apsdehal Differential Revision: D30778427 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: a7443eab8a0e9a2ee5d361883d84dc8880312fbf 13 September 2021, 13:30:50 UTC
7fa7607 Mudit Dhawan 09 September 2021, 22:57:18 UTC [feat] Weights and Biases Logging Feature (#1060) Summary: Add a WandbLogger class to support wandb logging. Update config options to enable wandb logger and integrate with the trainer. To enable wandb logging use the following options `training.wandb=True env.wandb_projectname=<ProjectName> env.wandb_runname=<RunName>` **Examples:** Screenshot of the wandb dashboard for two runs of m4c project on textvqa dataset. ![image](https://user-images.githubusercontent.com/42584330/131163071-50eb1405-aebb-4b61-982c-71000c29466f.png) 1. command line code for _run1_: `mmf_run config=projects/m4c/configs/textvqa/defaults.yaml datasets=textvqa model=m4c run_type=train_val training.batch_size=2 training.log_interval=2 training.wandb=True training.max_updates=10 env.wandb_projectname=Try1 env.wandb_runname=run1 ` 2. command line code for _run2_: `mmf_run config=projects/m4c/configs/textvqa/defaults.yaml datasets=textvqa model=m4c run_type=train_val training.batch_size=2 training.log_interval=1 training.wandb=True training.max_updates=10 env.wandb_projectname=Try1 env.wandb_runname=run2 ` Linked Issue: Wandb Logger in mmf https://github.com/facebookresearch/mmf/issues/588 Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1060 Test Plan: Imported from GitHub, without a `Test Plan:` line. |[Site Preview: mmf](https://our.intern.facebook.com/intern/staticdocs/eph/D30690086/V3/mmf/) Reviewed By: apsdehal Differential Revision: D30690086 Pulled By: vedanuj fbshipit-source-id: 684d0091b696747ed9535f4e4204e10dacdbf72b 09 September 2021, 22:58:13 UTC
15fa630 Sasha Sheng 09 September 2021, 02:58:17 UTC [refactor][PL][9/N] change the name of the _load_checkpoint to be more descriptive to contain test (#1030) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1030 Test Plan: Imported from OSS Reviewed By: simran2905 Differential Revision: D29888294 Pulled By: ytsheng fbshipit-source-id: 54919dc4e8448a1240b2cc72a364702aa7cf49c2 09 September 2021, 02:59:18 UTC
be5b032 Sasha Sheng 09 September 2021, 02:48:09 UTC [feat][PL][8/N] save config with checkpoint (#1024) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1024 Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D29670421 Pulled By: ytsheng fbshipit-source-id: ee97bd6b45f1ab8099403896e73f2d0c3a98ca96 09 September 2021, 02:49:07 UTC
e903a45 Sasha Sheng 09 September 2021, 02:23:52 UTC [feat][PL][7/N] trainer checkpoint from mmf to lightning update (#1023) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1023 Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D29670422 Pulled By: ytsheng fbshipit-source-id: 3d6387999634b19abd4b7895b6648f299f860772 09 September 2021, 02:25:22 UTC
5bb56b2 Sasha Sheng 09 September 2021, 02:10:36 UTC [feat][PL][6/N] pretrained state mapping for PL (#1022) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1022 Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D29670417 Pulled By: ytsheng fbshipit-source-id: f135cb85df212d6f86916a30fff9d90f20714844 09 September 2021, 02:11:28 UTC
c1d7495 Amanpreet Singh 09 September 2021, 00:56:52 UTC [fix,feat] Support other pathmanager prefixes for config (#1079) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1079 This will allow us to support configs present in manifold. Reviewed By: BruceChaun Differential Revision: D30798178 fbshipit-source-id: 9a22e0184db37ba3a8190527df042a6cc1007066 09 September 2021, 00:59:03 UTC
f5f1ef0 Amanpreet Singh 08 September 2021, 20:46:17 UTC [fix] Make demjson dep optional (#1080) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1080 Allows fixing issues with CI due to demjson. Reviewed By: ytsheng Differential Revision: D30813103 fbshipit-source-id: 3bfd113fd3a0b0438648305919f4fcf7e6497ea9 08 September 2021, 20:47:40 UTC
75ea3ab Ryan Jiang 02 September 2021, 19:29:28 UTC [fix] Add zoo requirements to coco masked config (#1072) Summary: Add same zoo_requirements as coco dataset to masked_coco. Prevents file not found error while loading datasets. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1072 Reviewed By: apsdehal Differential Revision: D30727857 Pulled By: Ryan-Qiyu-Jiang fbshipit-source-id: b96b26063156d2bb463fa6d19e254a9f703f9242 02 September 2021, 19:31:59 UTC
fd74eca evilshipper 02 September 2021, 05:14:54 UTC [docs] Changed the installation link which was broken in README.md (#1056) Summary: Update the broken link for installation in README.md from https://mmf.sh/docs/getting_started/installation to https://mmf.sh/docs/. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1056 Reviewed By: vedanuj Differential Revision: D30568779 Pulled By: apsdehal fbshipit-source-id: 2f66de70a8c2745932747415f8926279402992ea 02 September 2021, 05:30:48 UTC
6ad3335 dependabot[bot] 02 September 2021, 05:14:37 UTC Bump tar from 6.1.5 to 6.1.11 in /website (#1070) Summary: Bumps [tar](https://github.com/npm/node-tar) from 6.1.5 to 6.1.11. <details> <summary>Commits</summary> <ul> <li><a href="https://github.com/npm/node-tar/commit/e573aeea19d4d650908b7f6bf0a1ad8dce9f1736"><code>e573aee</code></a> 6.1.11</li> <li><a href="https://github.com/npm/node-tar/commit/edb8e9a3fa5869cfb935479a262f6f61b0a2ec57"><code>edb8e9a</code></a> fix: perf regression on hot string munging path</li> <li><a href="https://github.com/npm/node-tar/commit/a9d9b05fc974425089922bf66e8e7751ea8bfbbc"><code>a9d9b05</code></a> chore(test): Avoid spurious failures packing node_modules/.cache</li> <li><a href="https://github.com/npm/node-tar/commit/24b8bdadf37118182496ecf81fa7a872196fb38b"><code>24b8bda</code></a> fix(test): use posix path for testing path reservations</li> <li><a href="https://github.com/npm/node-tar/commit/e5a223c6b293f1cb2ebb2dd97668f787e8983ca4"><code>e5a223c</code></a> fix(test): make unpack test pass on case-sensitive fs</li> <li><a href="https://github.com/npm/node-tar/commit/188baddc1d0e6ef5140c7a788f03fc2a6c3df2ea"><code>188badd</code></a> 6.1.10</li> <li><a href="https://github.com/npm/node-tar/commit/23312ce7db8a12c78d0fba96d7664a01619266a3"><code>23312ce</code></a> drop dirCache for symlink on all platforms</li> <li><a href="https://github.com/npm/node-tar/commit/4f1f4a21fca64f3089da0e83ceea775c66b55052"><code>4f1f4a2</code></a> 6.1.9</li> <li><a href="https://github.com/npm/node-tar/commit/875a37e3ec031186fc6599f6807341f56c584598"><code>875a37e</code></a> fix: prevent path escape using drive-relative paths</li> <li><a href="https://github.com/npm/node-tar/commit/b6162c7fafe797f856564ef37f4b82747f051455"><code>b6162c7</code></a> fix: reserve paths properly for unicode, windows</li> <li>Additional commits viewable in <a href="https://github.com/npm/node-tar/compare/v6.1.5...v6.1.11">compare view</a></li> </ul> </details> <br /> [![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility_score?dependency-name=tar&package-manager=npm_and_yarn&previous-version=6.1.5&new-version=6.1.11)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `dependabot rebase`. [//]: # (dependabot-automerge-start) [//]: # (dependabot-automerge-end) --- <details> <summary>Dependabot commands and options</summary> <br /> You can trigger Dependabot actions by commenting on this PR: - `dependabot rebase` will rebase this PR - `dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `dependabot merge` will merge this PR after your CI passes on it - `dependabot squash and merge` will squash and merge this PR after your CI passes on it - `dependabot cancel merge` will cancel a previously requested merge and block automerging - `dependabot reopen` will reopen this PR if it is closed - `dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually - `dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself) - `dependabot use these labels` will set the current labels as the default for future PRs for this repo and language - `dependabot use these reviewers` will set the current reviewers as the default for future PRs for this repo and language - `dependabot use these assignees` will set the current assignees as the default for future PRs for this repo and language - `dependabot use this milestone` will set the current milestone as the default for future PRs for this repo and language You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/facebookresearch/mmf/network/alerts). </details> Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1070 Test Plan: Imported from GitHub, without a `Test Plan:` line. |[Site Preview: mmf](https://our.intern.facebook.com/intern/staticdocs/eph/D30690090/V2/mmf/) Reviewed By: apsdehal Differential Revision: D30690090 Pulled By: vedanuj fbshipit-source-id: 176e05fb9a99fa782b402ec7203f8d61b0e0787b 02 September 2021, 05:30:48 UTC
bbfd9bc tranvinhhung99 02 September 2021, 05:13:16 UTC [fix] Fix typo in script feature extraction (#945) Summary: `files` datatype should be set, so the function to remove an obj is `remove`, not `pop`. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/945 Reviewed By: vedanuj Differential Revision: D28703139 Pulled By: apsdehal fbshipit-source-id: c43035334258b3d61e1844b08a3c30a56d93dcfd 02 September 2021, 05:30:48 UTC
4658d90 Amanpreet Singh 02 September 2021, 00:34:00 UTC [fix,refactor] Allow proper teardown for SummaryWriter (#1074) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1074 This diff does following things: - Refactors TensorboardLogger to be more clean and support cleanup method - Adds a teardown hook support for callbacks - Cleans up the trainer a little to have single teardown call. Reviewed By: vedanuj Differential Revision: D30680073 fbshipit-source-id: f300cdae02f9f9f2cff4524797440bb8c1a1e4ce 02 September 2021, 01:31:23 UTC
453c39e Amanpreet Singh 01 September 2021, 21:52:17 UTC [fix] Only return cuda as current device if cuda is actually initialized (#942) Summary: This causes issues when GPU is available but person still want to run on CPU. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/942 Test Plan: Tested locally Reviewed By: vedanuj Differential Revision: D28917011 Pulled By: apsdehal fbshipit-source-id: e1cb1829140f152d589eaa1d613ec05f59269f7c 01 September 2021, 22:28:15 UTC
b4e2098 Rui Hou 01 September 2021, 21:22:30 UTC [feat] merge no_token_embedding_backend to main mmf (#188) Summary: Pull Request resolved: https://github.com/fairinternal/mmf-internal/pull/188 Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1071 We can upstream the pytext_backend_without_token_embeddings to main mmf to avoid duplicate code. Reviewed By: apsdehal Differential Revision: D30673593 fbshipit-source-id: f0b4b852049e1258cd7a0de29dc28bf626edc7a2 01 September 2021, 21:31:04 UTC
592289f Albert Lin 01 September 2021, 04:16:37 UTC [fix] Fix OmegaConf depreciation warning around register_resolver (#1069) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1069 The diff aims to fix the depreciation warning around register_resolver function by replacing it with the register_new_resolver. Reviewed By: vedanuj, apsdehal Differential Revision: D30671569 fbshipit-source-id: 743a51d936fda7060c1a257222a762676935be63 01 September 2021, 04:28:03 UTC
0d24224 PotatoSpudowski 31 August 2021, 02:44:04 UTC [feat] ALBEF Vision Transformer Encoder (#1063) Summary: PR adds ALBEF's Vision Transformer to MMF. - Adds Vision Transformer model - Adds Vision Transformer encoder - Adds tests for Vision Transformer encoder Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1063 Reviewed By: ronghanghu Differential Revision: D30646130 Pulled By: apsdehal fbshipit-source-id: 3daf5b9f8232203aa6d544e8975f6fa7e964dbaf 31 August 2021, 02:45:06 UTC
baa1f70 dependabot[bot] 28 August 2021, 01:20:05 UTC Bump url-parse from 1.5.1 to 1.5.3 in /website (#1046) Summary: Bumps [url-parse](https://github.com/unshiftio/url-parse) from 1.5.1 to 1.5.3. <details> <summary>Commits</summary> <ul> <li><a href="https://github.com/unshiftio/url-parse/commit/ad444931666a30bad11472d89a216461cf16cae2"><code>ad44493</code></a> [dist] 1.5.3</li> <li><a href="https://github.com/unshiftio/url-parse/commit/c7984617e235892cc22e0f47bb5ff1c012e6e39f"><code>c798461</code></a> [fix] Fix host parsing for file URLs (<a href="https://github-redirect.dependabot.com/unshiftio/url-parse/issues/210">https://github.com/facebookresearch/mmf/issues/210</a>)</li> <li><a href="https://github.com/unshiftio/url-parse/commit/201034b8670c2aa382d7ec410ee750ac6f2f9c38"><code>201034b</code></a> [dist] 1.5.2</li> <li><a href="https://github.com/unshiftio/url-parse/commit/2d9ac2c94067742b2116332c1e03be9f37371dff"><code>2d9ac2c</code></a> [fix] Sanitize only special URLs (<a href="https://github-redirect.dependabot.com/unshiftio/url-parse/issues/209">https://github.com/facebookresearch/mmf/issues/209</a>)</li> <li><a href="https://github.com/unshiftio/url-parse/commit/fb128af4f43fa17f351d50cf615c7598c751f50a"><code>fb128af</code></a> [fix] Use <code>'null'</code> as <code>origin</code> for non special URLs</li> <li><a href="https://github.com/unshiftio/url-parse/commit/fed6d9e338ea39de2d68bb66607066d71328c62f"><code>fed6d9e</code></a> [fix] Add a leading slash only if the URL is special</li> <li><a href="https://github.com/unshiftio/url-parse/commit/94872e7ab9103ee69b958959baa14c9e682a7f10"><code>94872e7</code></a> [fix] Do not incorrectly set the <code>slashes</code> property to <code>true</code></li> <li><a href="https://github.com/unshiftio/url-parse/commit/81ab967889b08112d3356e451bf03e6aa0cbb7e0"><code>81ab967</code></a> [fix] Ignore slashes after the protocol for special URLs</li> <li><a href="https://github.com/unshiftio/url-parse/commit/ee22050a48a67409aa5f7c87947284156d615bd1"><code>ee22050</code></a> [ci] Use GitHub Actions</li> <li><a href="https://github.com/unshiftio/url-parse/commit/d2979b586d8c7751e0c77f127d9ce1b2143cc0c9"><code>d2979b5</code></a> [fix] Special case the <code>file:</code> protocol (<a href="https://github-redirect.dependabot.com/unshiftio/url-parse/issues/204">https://github.com/facebookresearch/mmf/issues/204</a>)</li> <li>Additional commits viewable in <a href="https://github.com/unshiftio/url-parse/compare/1.5.1...1.5.3">compare view</a></li> </ul> </details> <br /> [![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility_score?dependency-name=url-parse&package-manager=npm_and_yarn&previous-version=1.5.1&new-version=1.5.3)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `dependabot rebase`. [//]: # (dependabot-automerge-start) [//]: # (dependabot-automerge-end) --- <details> <summary>Dependabot commands and options</summary> <br /> You can trigger Dependabot actions by commenting on this PR: - `dependabot rebase` will rebase this PR - `dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `dependabot merge` will merge this PR after your CI passes on it - `dependabot squash and merge` will squash and merge this PR after your CI passes on it - `dependabot cancel merge` will cancel a previously requested merge and block automerging - `dependabot reopen` will reopen this PR if it is closed - `dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually - `dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself) - `dependabot use these labels` will set the current labels as the default for future PRs for this repo and language - `dependabot use these reviewers` will set the current reviewers as the default for future PRs for this repo and language - `dependabot use these assignees` will set the current assignees as the default for future PRs for this repo and language - `dependabot use this milestone` will set the current milestone as the default for future PRs for this repo and language You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/facebookresearch/mmf/network/alerts). </details> Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1046 Test Plan: Imported from GitHub, without a `Test Plan:` line. |[Site Preview: mmf](https://our.intern.facebook.com/intern/staticdocs/eph/D30392262/V3/mmf/) Reviewed By: apsdehal Differential Revision: D30392262 Pulled By: vedanuj fbshipit-source-id: 5741d20e13b5269d60d214d56207c3f59c31c270 28 August 2021, 01:20:55 UTC
d772fd1 dependabot[bot] 27 August 2021, 21:39:18 UTC Bump tar from 6.1.0 to 6.1.5 in /website (#1039) Summary: Bumps [tar](https://github.com/npm/node-tar) from 6.1.0 to 6.1.5. <details> <summary>Commits</summary> <ul> <li><a href="https://github.com/npm/node-tar/commit/bd4691c90478f41b2649a97048199e34927dc046"><code>bd4691c</code></a> 6.1.5</li> <li><a href="https://github.com/npm/node-tar/commit/d694c4f810d864badf223efa35d24a000d780179"><code>d694c4f</code></a> ci: test on node 16</li> <li><a href="https://github.com/npm/node-tar/commit/84acbd31288541100910a528e437f901f8012214"><code>84acbd3</code></a> fix(unpack): fix hang on large file on open() fail</li> <li><a href="https://github.com/npm/node-tar/commit/97c46fcee7e4e7849ea3432086c4537fb6197025"><code>97c46fc</code></a> fix(unpack): always resume parsing after an entry error</li> <li><a href="https://github.com/npm/node-tar/commit/488ab8c01de69379406d937419fa3e5550e651c0"><code>488ab8c</code></a> chore: WriteEntry cleaner write() handling</li> <li><a href="https://github.com/npm/node-tar/commit/be89aafd95296e9721e124b77eee7c745e1c1e97"><code>be89aaf</code></a> WriteEntry backpressure</li> <li><a href="https://github.com/npm/node-tar/commit/ba73f5eea55f9cf65048b4d9578462ec8f80bc5e"><code>ba73f5e</code></a> chore: track fs state on WriteEntry class, not in arguments</li> <li><a href="https://github.com/npm/node-tar/commit/bf693837b3dcfeb76878b212310302dc5dc3d3dc"><code>bf69383</code></a> 6.1.4</li> <li><a href="https://github.com/npm/node-tar/commit/06cbde5935aa7643f578f874de84a7da2a74fe3a"><code>06cbde5</code></a> Avoid an unlikely but theoretically possible redos</li> <li><a href="https://github.com/npm/node-tar/commit/0b78386c53b00dce422742e19de94f2a4d9389f3"><code>0b78386</code></a> 6.1.3</li> <li>Additional commits viewable in <a href="https://github.com/npm/node-tar/compare/v6.1.0...v6.1.5">compare view</a></li> </ul> </details> <br /> [![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility_score?dependency-name=tar&package-manager=npm_and_yarn&previous-version=6.1.0&new-version=6.1.5)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `dependabot rebase`. [//]: # (dependabot-automerge-start) [//]: # (dependabot-automerge-end) --- <details> <summary>Dependabot commands and options</summary> <br /> You can trigger Dependabot actions by commenting on this PR: - `dependabot rebase` will rebase this PR - `dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `dependabot merge` will merge this PR after your CI passes on it - `dependabot squash and merge` will squash and merge this PR after your CI passes on it - `dependabot cancel merge` will cancel a previously requested merge and block automerging - `dependabot reopen` will reopen this PR if it is closed - `dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually - `dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself) - `dependabot use these labels` will set the current labels as the default for future PRs for this repo and language - `dependabot use these reviewers` will set the current reviewers as the default for future PRs for this repo and language - `dependabot use these assignees` will set the current assignees as the default for future PRs for this repo and language - `dependabot use this milestone` will set the current milestone as the default for future PRs for this repo and language You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/facebookresearch/mmf/network/alerts). </details> Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1039 Test Plan: Imported from GitHub, without a `Test Plan:` line. |[Site Preview: mmf](https://our.intern.facebook.com/intern/staticdocs/eph/D30392234/V3/mmf/) Reviewed By: apsdehal Differential Revision: D30392234 Pulled By: vedanuj fbshipit-source-id: 47546c55a093dc3a6d228228fa32e3cebccb6548 27 August 2021, 21:40:02 UTC
1947bd4 dependabot[bot] 27 August 2021, 21:21:02 UTC Bump prismjs from 1.23.0 to 1.24.0 in /website (#989) Summary: Bumps [prismjs](https://github.com/PrismJS/prism) from 1.23.0 to 1.24.0. <details> <summary>Release notes</summary> <p><em>Sourced from <a href="https://github.com/PrismJS/prism/releases">prismjs's releases</a>.</em></p> <blockquote> <h2>v1.24.0</h2> <p>Release 1.24.0</p> </blockquote> </details> <details> <summary>Changelog</summary> <p><em>Sourced from <a href="https://github.com/PrismJS/prism/blob/master/CHANGELOG.md">prismjs's changelog</a>.</em></p> <blockquote> <h2>1.24.0 (2021-06-27)</h2> <h3>New components</h3> <ul> <li><strong>CFScript</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2771">#2771</a>) <a href="https://github.com/PrismJS/prism/commit/b0a6ec85"><code>b0a6ec85</code></a></li> <li><strong>ChaiScript</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2706">#2706</a>) <a href="https://github.com/PrismJS/prism/commit/3f7d7453"><code>3f7d7453</code></a></li> <li><strong>COBOL</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2800">#2800</a>) <a href="https://github.com/PrismJS/prism/commit/7e5f78ff"><code>7e5f78ff</code></a></li> <li><strong>Coq</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2803">#2803</a>) <a href="https://github.com/PrismJS/prism/commit/41e25d3c"><code>41e25d3c</code></a></li> <li><strong>CSV</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2794">#2794</a>) <a href="https://github.com/PrismJS/prism/commit/f9b69528"><code>f9b69528</code></a></li> <li><strong>DOT (Graphviz)</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2690">#2690</a>) <a href="https://github.com/PrismJS/prism/commit/1f91868e"><code>1f91868e</code></a></li> <li><strong>False</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2802">#2802</a>) <a href="https://github.com/PrismJS/prism/commit/99a21dc5"><code>99a21dc5</code></a></li> <li><strong>ICU Message Format</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2745">#2745</a>) <a href="https://github.com/PrismJS/prism/commit/bf4e7ba9"><code>bf4e7ba9</code></a></li> <li><strong>Idris</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2755">#2755</a>) <a href="https://github.com/PrismJS/prism/commit/e9314415"><code>e9314415</code></a></li> <li><strong>Jexl</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2764">#2764</a>) <a href="https://github.com/PrismJS/prism/commit/7e51b99c"><code>7e51b99c</code></a></li> <li><strong>KuMir (КуМир)</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2760">#2760</a>) <a href="https://github.com/PrismJS/prism/commit/3419fb77"><code>3419fb77</code></a></li> <li><strong>Log file</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2796">#2796</a>) <a href="https://github.com/PrismJS/prism/commit/2bc6475b"><code>2bc6475b</code></a></li> <li><strong>Nevod</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2798">#2798</a>) <a href="https://github.com/PrismJS/prism/commit/f84c49c5"><code>f84c49c5</code></a></li> <li><strong>OpenQasm</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2797">#2797</a>) <a href="https://github.com/PrismJS/prism/commit/1a2347a3"><code>1a2347a3</code></a></li> <li><strong>PATROL Scripting Language</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2739">#2739</a>) <a href="https://github.com/PrismJS/prism/commit/18c67b49"><code>18c67b49</code></a></li> <li><strong>Q#</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2804">#2804</a>) <a href="https://github.com/PrismJS/prism/commit/1b63cd01"><code>1b63cd01</code></a></li> <li><strong>Rego</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2624">#2624</a>) <a href="https://github.com/PrismJS/prism/commit/e38986f9"><code>e38986f9</code></a></li> <li><strong>Squirrel</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2721">#2721</a>) <a href="https://github.com/PrismJS/prism/commit/fd1081d2"><code>fd1081d2</code></a></li> <li><strong>URI</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2708">#2708</a>) <a href="https://github.com/PrismJS/prism/commit/bbc77d19"><code>bbc77d19</code></a></li> <li><strong>V</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2687">#2687</a>) <a href="https://github.com/PrismJS/prism/commit/72962701"><code>72962701</code></a></li> <li><strong>Wolfram language</strong> &amp; <strong>Mathematica</strong> &amp; <strong>Mathematica Notebook</strong> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2921">#2921</a>) <a href="https://github.com/PrismJS/prism/commit/c4f6b2cc"><code>c4f6b2cc</code></a></li> </ul> <h3>Updated components</h3> <ul> <li>Fixed problems reported by <code>regexp/no-dupe-disjunctions</code> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2952">#2952</a>) <a href="https://github.com/PrismJS/prism/commit/f471d2d7"><code>f471d2d7</code></a></li> <li>Fixed some cases of quadratic worst-case runtime (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2922">#2922</a>) <a href="https://github.com/PrismJS/prism/commit/79d22182"><code>79d22182</code></a></li> <li>Fixed 2 cases of exponential backtracking (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2774">#2774</a>) <a href="https://github.com/PrismJS/prism/commit/d85e30da"><code>d85e30da</code></a></li> <li><strong>AQL</strong> <ul> <li>Update for ArangoDB 3.8 (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2842">#2842</a>) <a href="https://github.com/PrismJS/prism/commit/ea82478d"><code>ea82478d</code></a></li> </ul> </li> <li><strong>AutoHotkey</strong> <ul> <li>Improved tag pattern (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2920">#2920</a>) <a href="https://github.com/PrismJS/prism/commit/fc2a3334"><code>fc2a3334</code></a></li> </ul> </li> <li><strong>Bash</strong> <ul> <li>Accept hyphens in function names (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2832">#2832</a>) <a href="https://github.com/PrismJS/prism/commit/e4ad22ad"><code>e4ad22ad</code></a></li> <li>Fixed single-quoted strings (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2792">#2792</a>) <a href="https://github.com/PrismJS/prism/commit/e5cfdb4a"><code>e5cfdb4a</code></a></li> </ul> </li> <li><strong>C++</strong> <ul> <li>Added support for generic functions and made <code>::</code> punctuation (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2814">#2814</a>) <a href="https://github.com/PrismJS/prism/commit/3df62fd0"><code>3df62fd0</code></a></li> <li>Added missing keywords and modules (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2763">#2763</a>) <a href="https://github.com/PrismJS/prism/commit/88fa72cf"><code>88fa72cf</code></a></li> </ul> </li> <li><strong>Dart</strong> <ul> <li>Improved support for classes &amp; generics (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2810">#2810</a>) <a href="https://github.com/PrismJS/prism/commit/d0bcd074"><code>d0bcd074</code></a></li> </ul> </li> <li><strong>Docker</strong> <ul> <li>Improvements (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2720">#2720</a>) <a href="https://github.com/PrismJS/prism/commit/93dd83c2"><code>93dd83c2</code></a></li> </ul> </li> <li><strong>Elixir</strong> <ul> <li>Added missing keywords (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2958">#2958</a>) <a href="https://github.com/PrismJS/prism/commit/114e4626"><code>114e4626</code></a></li> <li>Added missing keyword and other improvements (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2773">#2773</a>) <a href="https://github.com/PrismJS/prism/commit/e6c0d298"><code>e6c0d298</code></a></li> <li>Added <code>defdelagate</code> keyword and highlighting for function/module names (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2709">#2709</a>) <a href="https://github.com/PrismJS/prism/commit/59f725d7"><code>59f725d7</code></a></li> </ul> </li> <li><strong>F#</strong></li> </ul> </blockquote> <p>... (truncated)</p> </details> <details> <summary>Commits</summary> <ul> <li><a href="https://github.com/PrismJS/prism/commit/3432b4b1e4440d6592ed82b6b5b9e72f660e43a8"><code>3432b4b</code></a> 1.24.0</li> <li><a href="https://github.com/PrismJS/prism/commit/46d07207687fa747018b6676250507a486c0117f"><code>46d0720</code></a> Updated <code>.npmignore</code> (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2971">#2971</a>)</li> <li><a href="https://github.com/PrismJS/prism/commit/aef7f08df6d6fe1e027ee3ab347c2f391c0c1045"><code>aef7f08</code></a> Changelog for v1.24.0 (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2965">#2965</a>)</li> <li><a href="https://github.com/PrismJS/prism/commit/e9477d8369bc59cacc99d1d81abfe3e20b7df258"><code>e9477d8</code></a> Markdown: Improved code snippets (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2967">#2967</a>)</li> <li><a href="https://github.com/PrismJS/prism/commit/4b55bd6af88559d430fc195fbe5845364ade8df1"><code>4b55bd6</code></a> Made Match Braces and Custom Class compatible (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2947">#2947</a>)</li> <li><a href="https://github.com/PrismJS/prism/commit/e8d3b50330a325a8291f20d63f60e68a985ae738"><code>e8d3b50</code></a> ESLint: Added <code>regexp/strict</code> rule (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2944">#2944</a>)</li> <li><a href="https://github.com/PrismJS/prism/commit/bfd7fded29755510571e3abf0846f2a9edf44ef6"><code>bfd7fde</code></a> GraphQL: Fixed <code>definition-query</code> and <code>definition-mutation</code> tokens (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2964">#2964</a>)</li> <li><a href="https://github.com/PrismJS/prism/commit/14e3868f05f84d05885f5465264c5c72e6ca9490"><code>14e3868</code></a> Fixed reST test</li> <li><a href="https://github.com/PrismJS/prism/commit/a7656de67a07e6415fe0c7149708c8613ff73c12"><code>a7656de</code></a> reST: Fixed <code>inline</code> pattern (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2946">#2946</a>)</li> <li><a href="https://github.com/PrismJS/prism/commit/b4ac0618156a13ab04ff685c5091cb436e8a13a4"><code>b4ac061</code></a> ESLint: Use cache (<a href="https://github-redirect.dependabot.com/PrismJS/prism/issues/2959">#2959</a>)</li> <li>Additional commits viewable in <a href="https://github.com/PrismJS/prism/compare/v1.23.0...v1.24.0">compare view</a></li> </ul> </details> <details> <summary>Maintainer changes</summary> <p>This version was pushed to npm by <a href="https://www.npmjs.com/~rundevelopment">rundevelopment</a>, a new releaser for prismjs since your current version.</p> </details> <br /> [![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility_score?dependency-name=prismjs&package-manager=npm_and_yarn&previous-version=1.23.0&new-version=1.24.0)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `dependabot rebase`. [//]: # (dependabot-automerge-start) [//]: # (dependabot-automerge-end) --- <details> <summary>Dependabot commands and options</summary> <br /> You can trigger Dependabot actions by commenting on this PR: - `dependabot rebase` will rebase this PR - `dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `dependabot merge` will merge this PR after your CI passes on it - `dependabot squash and merge` will squash and merge this PR after your CI passes on it - `dependabot cancel merge` will cancel a previously requested merge and block automerging - `dependabot reopen` will reopen this PR if it is closed - `dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually - `dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself) - `dependabot use these labels` will set the current labels as the default for future PRs for this repo and language - `dependabot use these reviewers` will set the current reviewers as the default for future PRs for this repo and language - `dependabot use these assignees` will set the current assignees as the default for future PRs for this repo and language - `dependabot use this milestone` will set the current milestone as the default for future PRs for this repo and language You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/facebookresearch/mmf/network/alerts). </details> Pull Request resolved: https://github.com/facebookresearch/mmf/pull/989 Test Plan: Imported from GitHub, without a `Test Plan:` line. |[Site Preview: mmf](https://our.intern.facebook.com/intern/staticdocs/eph/D29494387/V3/mmf/) Reviewed By: apsdehal Differential Revision: D29494387 Pulled By: vedanuj fbshipit-source-id: d717296fbab2c085c531084dbd2f54ca06c3e197 27 August 2021, 21:28:50 UTC
e019df6 dependabot[bot] 27 August 2021, 21:19:13 UTC Bump color-string from 1.5.3 to 1.5.5 in /website (#984) Summary: Bumps [color-string](https://github.com/Qix-/color-string) from 1.5.3 to 1.5.5. <details> <summary>Release notes</summary> <p><em>Sourced from <a href="https://github.com/Qix-/color-string/releases">color-string's releases</a>.</em></p> <blockquote> <h2>1.5.5 (Patch/Security Release) - hwb() ReDos patch (low-severity)</h2> <blockquote> <p>Release notes copied verbatim from the commit message, which can be found here: 0789e21284c33d89ebc4ab4ca6f759b9375ac9d3</p> </blockquote> <pre><code>Discovered by Yeting Li, c/o Colin Ife via Snyk.io. <p>A ReDos (Regular Expression Denial of Service) vulnerability was responsibly disclosed to me via email by Colin on Mar 5 2021 regarding an exponential time complexity for linearly increasing input lengths for <code>hwb()</code> color strings.</p> <p>Strings reaching more than 5000 characters would see several milliseconds of processing time; strings reaching more than 50,000 characters began seeing 1500ms (1.5s) of processing time.</p> <p>The cause was due to a the regular expression that parses hwb() strings - specifically, the hue value - where the integer portion of the hue value used a 0-or-more quantifier shortly thereafter followed by a 1-or-more quantifier.</p> <p>This caused excessive backtracking and a cartesian scan, resulting in exponential time complexity given a linear increase in input length.</p> <p>Thank you Yeting Li and Colin Ife for bringing this to my attention in a secure, responsible and professional manner.</p> <p>A CVE will not be assigned for this vulnerability. </code></pre></p> <h2>1.5.4 (Patch Release)</h2> <ul> <li>Removes rounding of alpha values in RGBA hex (<code>#rrggbbaa</code>) and condensed-hex (<code>#rgba</code>) parsers, which caused certain unique inputs to result in identical outputs (see <a href="https://github.com/qix-/color/issues/174">https://github.com/qix-/color/issues/174</a>).</li> </ul> </blockquote> </details> <details> <summary>Commits</summary> <ul> <li>See full diff in <a href="https://github.com/Qix-/color-string/commits/1.5.5">compare view</a></li> </ul> </details> <br /> [![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility_score?dependency-name=color-string&package-manager=npm_and_yarn&previous-version=1.5.3&new-version=1.5.5)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `dependabot rebase`. [//]: # (dependabot-automerge-start) [//]: # (dependabot-automerge-end) --- <details> <summary>Dependabot commands and options</summary> <br /> You can trigger Dependabot actions by commenting on this PR: - `dependabot rebase` will rebase this PR - `dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `dependabot merge` will merge this PR after your CI passes on it - `dependabot squash and merge` will squash and merge this PR after your CI passes on it - `dependabot cancel merge` will cancel a previously requested merge and block automerging - `dependabot reopen` will reopen this PR if it is closed - `dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually - `dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself) - `dependabot use these labels` will set the current labels as the default for future PRs for this repo and language - `dependabot use these reviewers` will set the current reviewers as the default for future PRs for this repo and language - `dependabot use these assignees` will set the current assignees as the default for future PRs for this repo and language - `dependabot use this milestone` will set the current milestone as the default for future PRs for this repo and language You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/facebookresearch/mmf/network/alerts). </details> Pull Request resolved: https://github.com/facebookresearch/mmf/pull/984 Test Plan: Imported from GitHub, without a `Test Plan:` line. |[Site Preview: mmf](https://our.intern.facebook.com/intern/staticdocs/eph/D29494369/V3/mmf/) Reviewed By: apsdehal Differential Revision: D29494369 Pulled By: vedanuj fbshipit-source-id: fb0c3d676bca6f29b542d072094f4e835c60e922 27 August 2021, 21:22:39 UTC
1ca5979 dependabot[bot] 27 August 2021, 21:18:02 UTC Bump path-parse from 1.0.6 to 1.0.7 in /website (#1045) Summary: Bumps [path-parse](https://github.com/jbgutierrez/path-parse) from 1.0.6 to 1.0.7. <details> <summary>Commits</summary> <ul> <li>See full diff in <a href="https://github.com/jbgutierrez/path-parse/commits/v1.0.7">compare view</a></li> </ul> </details> <br /> [![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility_score?dependency-name=path-parse&package-manager=npm_and_yarn&previous-version=1.0.6&new-version=1.0.7)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `dependabot rebase`. [//]: # (dependabot-automerge-start) [//]: # (dependabot-automerge-end) --- <details> <summary>Dependabot commands and options</summary> <br /> You can trigger Dependabot actions by commenting on this PR: - `dependabot rebase` will rebase this PR - `dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `dependabot merge` will merge this PR after your CI passes on it - `dependabot squash and merge` will squash and merge this PR after your CI passes on it - `dependabot cancel merge` will cancel a previously requested merge and block automerging - `dependabot reopen` will reopen this PR if it is closed - `dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually - `dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself) - `dependabot use these labels` will set the current labels as the default for future PRs for this repo and language - `dependabot use these reviewers` will set the current reviewers as the default for future PRs for this repo and language - `dependabot use these assignees` will set the current assignees as the default for future PRs for this repo and language - `dependabot use this milestone` will set the current milestone as the default for future PRs for this repo and language You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/facebookresearch/mmf/network/alerts). </details> Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1045 Test Plan: Imported from GitHub, without a `Test Plan:` line. |[Site Preview: mmf](https://our.intern.facebook.com/intern/staticdocs/eph/D30392250/V3/mmf/) Reviewed By: apsdehal Differential Revision: D30392250 Pulled By: vedanuj fbshipit-source-id: 6fc81e34b6806a3929f9f29bcc526515bfcf8264 27 August 2021, 21:19:19 UTC
87df959 Sasha Sheng 26 August 2021, 23:09:40 UTC [feat][PL][5/N] load model checkpoint directly + model zoo (#1021) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1021 So for each of the diffs, the test should make it clear what it does. But here is the overview. This diff does the following two: - the lightning trainer to read directly from an mmf zoo checkpoint. - added the test for loading the mmf zoo checkpoint using the lightning trainer Notes: lightning models use the provided API load_from_checkpoint api. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D29670423 Pulled By: ytsheng fbshipit-source-id: 8a07e875abaff9db7cd17b85f80cb15cb9fad6fe 26 August 2021, 23:10:48 UTC
c191cd4 Sasha Sheng 26 August 2021, 22:53:15 UTC [feat][PL][4/N] best ckpt (#1020) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1020 Add the summary as a comment since every time I export the PR from github it overrides the summary. This PR does the following things: - Adds the code to do best checkpoint monitoring. The monitor callback is an instance of the ModelCheckpoint callback, same as the trainingCallback. - Adds the test for the best checkpoint monitoring; Adjust the test to make it the cleanest for handle best checkpoint monitoring. Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D29670420 Pulled By: ytsheng fbshipit-source-id: 4a72196869c205bfe4cc5e4a6bfda21d74934065 26 August 2021, 22:54:19 UTC
e4ef8e9 Ronghang Hu 26 August 2021, 17:16:39 UTC [feat] allow extra commandline args in sweep scripts (#1057) Summary: This PR adds `--extra_args` in sweep lib to allow passing additional MMF command-line arguments, like `mmf_run` and `mmf_predict`. Usage: one can append to the sweep script extra command line arguments such as ``` python tools/sweeps/sweep_visual_bert.py \ <gpu, node, or other arguments> \ --extra_args training.checkpoint_interval 10000 training.batch_size 128 ``` Note that it also support the other format with equal sign `=` in args such as `--extra_args training.batch_size=128` Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1057 Test Plan: tested on FAIR cluster Reviewed By: apsdehal Differential Revision: D30568684 Pulled By: ronghanghu fbshipit-source-id: 88f85be01b4aa3aa32d616ae72e4b84ce4e51aa5 26 August 2021, 17:17:31 UTC
92c167d Ronghang Hu 20 August 2021, 23:58:48 UTC [fix] add `weight_decay` parameter to `get_bert_configured_parameters` (#1034) Summary: Allow specifiying custom weight decay in `get_bert_configured_parameters`. The behavior of existing codebase is kept the same for backward compatibility. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1034 Test Plan: tested in custom projects Reviewed By: apsdehal Differential Revision: D30459987 Pulled By: ronghanghu fbshipit-source-id: 9bca0775e3dfc095716e0aa53dc1b691f3b33730 20 August 2021, 23:59:48 UTC
8c4189b Sasha Sheng 18 August 2021, 01:36:45 UTC [fix][PL][3/N] minor change in base model (#1019) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1019 Two things: 1. make sure to not `should_update_loss` when updating the meter from the report. This is also the current behavior in `evaluation_loop` with mmf trainer. 2. Remove some functions in test, since they should just be using the parent class's functions. Test Plan: Imported from OSS Reviewed By: simran2905 Differential Revision: D29670419 Pulled By: ytsheng fbshipit-source-id: 52cc155907d25ade1be1f6d675f7cdb6ea6ff658 18 August 2021, 01:37:47 UTC
b4211e2 Sasha Sheng 18 August 2021, 01:36:45 UTC [feat][PL][2/N] PL train checkpoint (#1018) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1018 Test Plan: Imported from OSS Reviewed By: apsdehal Differential Revision: D29670418 Pulled By: ytsheng fbshipit-source-id: 54626b9367d42c1e6d0148a0afacbcf3d9248495 18 August 2021, 01:37:47 UTC
90b39ab Animesh Sinha 18 August 2021, 00:50:48 UTC Fix typo in BaseAlign's config Summary: Fix typo in BaseAlign's config Reviewed By: BruceChaun Differential Revision: D30378765 fbshipit-source-id: 5283977c8a3406d20a2823abb4c1b651fb46136d 18 August 2021, 00:51:51 UTC
5d35649 Luis Perez 16 August 2021, 22:34:12 UTC [Lighting] Upgrade `requirements.txt` for downstream projects to latest fbcode revision. Summary: ^ Reviewed By: yifuwang Differential Revision: D30346110 fbshipit-source-id: 154e69f233132635e947ddbd252ffbd957ead6f1 16 August 2021, 22:35:21 UTC
8c80a00 Amanpreet Singh 11 August 2021, 07:32:54 UTC [MMF][fix] Only save config in master process Summary: In large-scale distributed jobs, multiple processes writing config simultaneously to the manifold can cause issues with consistency among the keys. This PR makes sure that only master process writes the config. Reviewed By: vedanuj Differential Revision: D30193487 fbshipit-source-id: 060fe341a7dfd13f81627ded948e35a669ecbde6 11 August 2021, 07:33:53 UTC
c745af5 Ronghang Hu 11 August 2021, 05:32:59 UTC [feat] use persistent workers in data loader by default (#1041) Summary: Switch to use persistent workers in MMF data loaders, so that one doesn't need to keep forking and joining worker processes upon each iteration. This update reduces the overhead and potential OOM error in worker forking. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1041 Test Plan: tested locally Reviewed By: vedanuj Differential Revision: D30238770 Pulled By: ronghanghu fbshipit-source-id: d2ba9fa10d9be4468cf69004104824833428a45b 11 August 2021, 05:34:44 UTC
d37fffc Luis Perez 27 July 2021, 04:43:34 UTC Synchronize PyTorchLightning/pytorch-lightning (revision 000fbe63@000fbe63) to github/third-party/PyTorchLightning/pytorch-lightning Summary: # Manual Changes - Migrate callsites of `_extract_batch_size` to `extract_batch_size` (as per https://fburl.com/code/4q7n8fs9). - Remove unnecessary unit tests in `test_hive_writing_callback.py` ### New commit log messages 000fbe63 Expose `extract_batch_size` method and add corresponding tests. (#8357) Reviewed By: yifuwang Differential Revision: D29834484 fbshipit-source-id: 219a3d40401d9b2c35d3a74b75f2394c4f57d61b 27 July 2021, 04:44:34 UTC
7fabdb6 Luis Perez 27 July 2021, 04:43:34 UTC Synchronize PyTorchLightning/pytorch-lightning (revision eb6d9912@eb6d9912) to github/third-party/PyTorchLightning/pytorch-lightning Summary: # Note After this PR, gradients are clipped/normalized in between `on_after_backward` and `on_before_optimizer_step`. Changed the test to perform the gradient clipping check `on_before_optimizer_step` rather than `on_after_backward`. - Changed `test_grad_clipping.py` to use `on_before_optimizer_step` for the callback. ### New commit log messages eb6d9912 Refactor plugins backward (#8328) Reviewed By: yifuwang Differential Revision: D29834093 fbshipit-source-id: efa284ea92ec715d4fa2155f72f969e6399febca 27 July 2021, 04:44:34 UTC
1056bdd Ronghang Hu 25 July 2021, 22:41:40 UTC [feat] check NaNs in losses during training, and exit when it happens (#1027) Summary: Add config option `training.exit_on_nan_losses` (Default to True). When `training.exit_on_nan_losses` is True, check for NaNs in the loss dict and exit the training when NaN happens. Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1027 Test Plan: tested locally; a new test case `test_exit_on_nan_losses` is added. Reviewed By: apsdehal Differential Revision: D29855002 Pulled By: ronghanghu fbshipit-source-id: e5c8ea3c17c5ed7140a417317062c0f4c2ea45fa 25 July 2021, 22:42:43 UTC
77de437 Sasha Sheng 24 July 2021, 00:25:19 UTC [fix][PL][1/N]: clean up lightning test code (#1017) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1017 ≈ Test Plan: Imported from OSS Reviewed By: apsdehal, simran2905 Differential Revision: D29670416 Pulled By: ytsheng fbshipit-source-id: 8fb5e71cd31ea8133be8f868dd132f2553e04e15 24 July 2021, 00:26:14 UTC
cef7440 Rui Hou 23 July 2021, 05:15:13 UTC [perf] tune NCCL params for acceleration (#185) Summary: Pull Request resolved: https://github.com/fairinternal/mmf-internal/pull/185 Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1026 Following suggestions from post https://fb.workplace.com/notes/yi-wang/faster-distributed-training-in-2-lines/271004654689757 and add 2 more NCCL parameters if it's distributed training on FB cluster. Generally, we are seeing around 1.5 ~ 2X speedup in MMF. Reviewed By: vedanuj, apsdehal Differential Revision: D28306852 fbshipit-source-id: d41d44b6708e43e084aa061afeee9f1981bf5bec 23 July 2021, 05:16:44 UTC
c5c7930 Jun Chen 22 July 2021, 20:23:52 UTC [MMF][feat] Add callback in registry (#1029) Summary: Pull Request resolved: https://github.com/facebookresearch/mmf/pull/1029 Add callback register so that user can add their customized callback function in MMF training. # Usage ``` training: ... callbacks: - type: my_callback params: foo: bar ``` Reviewed By: vedanuj Differential Revision: D29810950 fbshipit-source-id: 36c03cd9aa29a9e6f1c8fda61ad3be95088dbca8 22 July 2021, 20:24:59 UTC
back to top