OpenGVLab
/

InternVL-Chat-V1-5

Image-Text-to-Text

feature-extraction

Model card Files Files and versions

Metrics Training metrics Community

czczup commited on Apr 18, 2024

Commit

8c16a21

·

verified ·

1 Parent(s): a2fda74

Update README.md

Files changed (1) hide show

README.md +2 -2

README.md CHANGED Viewed

@@ -19,8 +19,8 @@ pipeline_tag: visual-question-answering
 - **Model Stats:**
   - Architecture: [InternViT-6B-448px-V1-5](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2) + MLP + [InternLM2-Chat-20B](https://huggingface.co/internlm/internlm2-chat-20b)
   - Params: 25.5B
-  - Image size: dynamic resolution, max to 24 tiles of 448 x 448 during inference.
-  - Number of visual tokens: 256 * number of tiles
 - **Training Strategy:**
   - Pretraining Stage

 - **Model Stats:**
   - Architecture: [InternViT-6B-448px-V1-5](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2) + MLP + [InternLM2-Chat-20B](https://huggingface.co/internlm/internlm2-chat-20b)
   - Params: 25.5B
+  - Image size: dynamic resolution, max to 40 tiles of 448 x 448 during inference.
+  - Number of visual tokens: 256 * (number of tiles + 1)
 - **Training Strategy:**
   - Pretraining Stage