Hugging Face's logo

menglc
/
SliMM-DeepStackE-Qwen2VL-2B

Image-Text-to-Text
Transformers
Safetensors
qwen2_vl
text-generation
multimodal
conversational
text-generation-inference
Model card Files
xet
Community
menglc commited on
Commit
d72dba2
·
verified ·
1 Parent(s): b2a7ef6

update readme

Browse files
Files changed (1) hide show
  1. README.md +3 -3
README.md CHANGED
@@ -89,12 +89,12 @@ print(output_text)
89
 
90
  | Model | MMMU (Val) | ChartQA (Test) | AI2D (test) | DocVQA (val)
91
  |----------------------------------------------------------|------------|----------------|-------------|-------------|
92
- |Qwen2VL-2B (official evaluation) |41.1 | 73.5 |74.7 |90.1 |
93
  |Qwen2VL-2B (our evaluation, 1024 max vistokens to LLM) |39.4 | 75.6 |70.7 |90.4 |
94
  |SliMM-DeepStackE-Qwen2VL-0.5B (256 max vistokens to LLM) |40.7 | 74.5 |74.7 |85.4 |
95
  |SliMM-DeepStackE-Qwen2VL-0.5B (400 max vistokens to LLM) |41.2 | 76.8 |74.9 |88.0 |
96
-
97
-
98
 
99
  <p align="left">
100
  <img src="https://cdn-uploads.huggingface.co/production/uploads/64d852a4bab152b2470bf96e/dtVzPkcIp40oH8sg7MG_u.png" alt="Trade-off between N Vistokens for LLM and Acc" style="width:500px;" > <br>
 
89
 
90
  | Model | MMMU (Val) | ChartQA (Test) | AI2D (test) | DocVQA (val)
91
  |----------------------------------------------------------|------------|----------------|-------------|-------------|
92
+ |Qwen2VL-2B (official evaluation) |41.1 | 73.5 |74.7 |90.1* |
93
  |Qwen2VL-2B (our evaluation, 1024 max vistokens to LLM) |39.4 | 75.6 |70.7 |90.4 |
94
  |SliMM-DeepStackE-Qwen2VL-0.5B (256 max vistokens to LLM) |40.7 | 74.5 |74.7 |85.4 |
95
  |SliMM-DeepStackE-Qwen2VL-0.5B (400 max vistokens to LLM) |41.2 | 76.8 |74.9 |88.0 |
96
+
97
+ <code>*</code> indicates the performance on DocVQA test set
98
 
99
  <p align="left">
100
  <img src="https://cdn-uploads.huggingface.co/production/uploads/64d852a4bab152b2470bf96e/dtVzPkcIp40oH8sg7MG_u.png" alt="Trade-off between N Vistokens for LLM and Acc" style="width:500px;" > <br>