initial commit

ET-Instruct/FPO/fpo_annotation.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d06f4a87470842485cf2c1cd9775260134d947d7b7628f3a3f427284dd8fed58
+size 276539205

ET-Instruct/FPO/tokenized_fpo_annotation.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cedd16c036e06a291b4242e42f5255476d423301f4104f068cd456fb07b9e56a
+size 655216316

ET-Instruct/evi.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d07e22bacb09971db4be35c3e76f10e42d774f0707c3d023f38941520d3450dc
+size 80971635

ETBench/evi/dvc_hirest.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d20daff23548aed21a2ab7d5d259c368aca3480d6df593f6ce68aa04b0e7ff8
+size 53431

ETBench/evi/dvc_youcook2.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12653eee50b89827511df9865350573be05bbe8e4d3d5c08c863b3168848639e
+size 374612

ETBench/evi/eca_charades_sta.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27863910ddec932803a7931a01eab304d13dc2b76a86f20ddc0b4a91a6639c61
+size 316203

ETBench/evi/epm_eg4d_nlq.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcf08d0d7119c273b14e2a95b71af5bc63b6a9bff1ae6dafde9a7ee1ff2ee242
+size 227773

ETBench/evi/evs_summe.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:201ee0ced9b3b6c8d3130cde94acf659051828b443595a605859eb4411f6eae0
+size 16417

ETBench/evi/evs_tvsum.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52a3c8b4b343deb676eb4d31082e81babc65dd668c7b95add0a979856f79da8f
+size 29503

ETBench/evi/gvq_qa_ego4d.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fdc75007942fe46825b04168dbbbf05fd37ff5c5181d50ac169c217084decec
+size 227378

ETBench/evi/rar_perception_test.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84f3951010bb1f00249400a85e77d83e9ef55a9f669ff8e865c1d0afb5ac19b2
+size 377201

ETBench/evi/rvq_star.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2973235b07573ec95986ad486eb0f3b9796e15688d968548846ef14c69700ad
+size 402009

ETBench/evi/slc_cross_task.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82177300230e265443522582b2b038d9235fd679fc158dc1060b6106b64b9d36
+size 198424

ETBench/evi/slc_ht_step.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:531b318aa2c138758378343c009799f535731bf018cf972511b22a2c9d1716c9
+size 343154

ETBench/evi/tal_perception_test.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:463a7a98ae326827fa34cb8cce512cae3222ff854bf802ff26e74a0206d95490
+size 262613

ETBench/evi/tal_thumos14.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49ba6d6e7fc6c07a2b15831e2864e7a03696adc2789891ec7a234dd1a702fb92
+size 74363

ETBench/evi/tal_thumos15.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a0c8c5b82615b83cc4a5b5f9074589c06ad097721f7fe74b371a58c5b2ca88a
+size 152944

ETBench/evi/tem_perception_test.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66d3989f8d30952d881d28c6fcad9b43265f974ba728ba9f85d411dd318c8a8e
+size 247676

ETBench/evi/tem_qvhighlights.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fec56b0e19ed096491410034db4a105f1bcf956f038b8c6b0c1128dcbea7b53
+size 221839

ETBench/evi/tvg_charades_sta.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c626ed59a94e00bf5979e079feb386ff569e2b70c6ef9a7382cbd73a4f2c7fec
+size 201680

ETBench/evi/tvg_qvhighlights.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0392f7abd3b15082d0fe76572bb95413f86c5368b5c48faf48644b21cc46ec2
+size 220427

ETBench/evi/vhd_qvhighlights.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:997c81f8fcc744695665e4fac89fd932c903d01ea3d88c934b7c34a23ed829ee
+size 235422

ETBench/evi/vhd_youtube_highlights.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f223f221cfc1b1f916dfe79062c02cd119f4cbadea0bef8877c2de1b24204448
+size 144337

README.md

@@ -1,3 +1,75 @@
----
-license: apache-2.0
----
+# Dataset used in D2VLM
+
+Here we provided the instruction for dataset preparation. We provided the transformed annotation (or steps for transformation) tailor for D2VLM, due to the new training objective design.
+
+
+| Dataset   Name |         Role         | Source |
+|----------------|:--------------------:|:------:|
+| E.T. Bench     | Evaluation Benchmark |    [Data](https://github.com/PolyU-ChenLab/ETBench/blob/main/docs/BENCHMARK.md)   |
+| Charades-STA   | Evaluation Benchmark |    [Data](https://prior.allenai.org/projects/charades), [Annotation](https://huggingface.co/datasets/ShuhuaiRen/TimeIT)   | 
+| Youcook2       | Evaluation Benchmark |    [Data](http://youcook2.eecs.umich.edu/download), [Annotation](https://huggingface.co/datasets/ShuhuaiRen/TimeIT)   |
+| E.T. Instruct  |    Finetuning data   |    [Data](https://github.com/PolyU-ChenLab/ETBench/blob/main/docs/DATASET.md)   |
+
+
+
+## E.T. Bench
+1. Download and process the data following [official document](https://github.com/PolyU-ChenLab/ETBench/blob/main/docs/BENCHMARK.md).
+
+2. Use the annotations provided in this Hugging Face repo (i.e., `D2VLM-Dataset/ETBench/evi`).
+
+
+## Charades-STA
+1. Follow the provided link in the table above to download the raw videos and the annotations. 
+The folder is organized as follows.
+
+```
+xx/charades
+├─ Charades_v1 (Downloaded and extracted raw videos from Charades_v1.zip)
+│  ├─ 001TG.mp4
+│  ├─ 003WS.mp4
+│  └─ ...
+└─ test.caption_coco_format.json
+```
+2. Run the post-processing scripts (in project repo). Remember to update paths in the .py files referenced by the script below.
+
+  ```shell
+  bash other_benchmark_organize/charades/run.sh
+  ```
+## Youcook2
+1. Follow the provided link in the table above to download the raw videos and the annotations. 
+The folder is organized as follows.
+
+```
+xx/youcook2
+├─ raw_videos (Downloaded and extracted raw videos from raw_videos.tar.gz)
+│  ├─ testing
+│  ├─ training
+│  └─ validation
+└─ val.caption_coco_format.json
+```
+
+
+2. Run the post-processing scripts (in project repo). Remember to update paths in the .py files referenced by the script below.
+
+ ```shell
+  bash other_benchmark_organize/youcook2/run.sh
+  ```
+
+## E.T. Instruct
+1. Download and process the data following [official document](https://github.com/PolyU-ChenLab/ETBench/blob/main/docs/DATASET.md).
+
+2. Use the annotations provided in this Hugging Face repo.
+
+- `D2VLM-Dataset/ET-Instruct/evi.json` --> For supervised finetuning (SFT).
+
+- `/D2VLM-Dataset/ET-Instruct/FPO/tokenized_fpo_annotation.pt` --> For FPO. 
+
+>> For FPO, we applied an optimization: the dispreferred sample is appended after the preferred (positive) sample within the same sequence, and we adjust the positional encodings and attention mask accordingly. This makes training more efficient than naively training paired samples as separate batch items. One may check the `fpo_anno_gen/run.sh` in the project repo to learn more.
+
+
+
+
+
+ 
+
+