add checkpoint for msrvtt-caption

microsoft · Mar 1, 2023 · 9792913 · 9792913
1 parent 5c783cd
commit 9792913
Show file tree

Hide file tree

Showing 3 changed files with 6 additions and 0 deletions.
diff --git a/README.md b/README.md
@@ -61,13 +61,15 @@ This repo presents some example codes to reproduce some results in
     | GIT_BASE_VQAv2      | fine-tuned on VQAv2                                 | test-dev: 72.72         |
     | GIT_BASE_TEXTVQA    | fine-tuned on TextVQA                               | val/acc: 18.81          |
     | GIT_BASE_VATEX      | fine-tuned on VATEX for captioning                  | public/test/CIDEr: 60.0 |
+    | GIT_BASE_MSRVTT     | fine-tuned on MSRVTT for captioning                 | test/CIDEr: 57.8        |
     | GIT_BASE_MSRVTT_QA  | fine-tuned on MSRVTT for question answering         | acc: 41.0               |
     | GIT_LARGE           | pretrained on 14M images                            |                         |
     | GIT_LARGE_COCO      | fine-tuned on COCO                                  | CIDEr: 138.5            |
     | GIT_LARGE_TEXTCAPS  | fine-tuned on TextCaps for captioning               | val/CIDEr: 106.3        |
     | GIT_LARGE_VQAv2     | fine-tuned on VQAv2                                 | test-dev: 75.51         |
     | GIT_LARGE_TEXTVQA   | fine-tuned on TextVQA                               | val/acc: 37.47          |
     | GIT_LARGE_VATEX     | fine-tuned on VATEX for captioning                  | public/test/CIDEr: 72.5 |
+    | GIT_LARGE_MSRVTT    | fine-tuned on MSRVTT for captioning                 | test/CIDEr: 64.1        |
     | GIT_LARGE_MSRVTT_QA | fine-tuned on MSRVTT for question answering         | acc: 42.7               |
 
   - In the dataset of cc12m, the caption may contain some special tags to hide

diff --git a/aux_data/models/GIT_BASE_MSRVTT/parameter.yaml b/aux_data/models/GIT_BASE_MSRVTT/parameter.yaml
@@ -0,0 +1 @@
+num_image_with_embedding: 6
diff --git a/aux_data/models/GIT_LARGE_MSRVTT/parameter.yaml b/aux_data/models/GIT_LARGE_MSRVTT/parameter.yaml
@@ -0,0 +1,3 @@
+num_image_with_embedding: 6
+visual_feature_size: 1024
+image_encoder_type: CLIPViT_L_14