andreagemelli commited on
Commit
38ff506
·
verified ·
1 Parent(s): bdd4301

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +6 -13
README.md CHANGED
@@ -16,6 +16,8 @@ tags:
16
  base_model:
17
  - OpenGVLab/InternViT-300M-448px-V2_5
18
  - PleIAs/Baguettotron
 
 
19
  ---
20
 
21
  # baguettotron-internvit-alignment
@@ -36,9 +38,9 @@ Two checkpoints are published:
36
  — instruction-tuned on top of it. It goes past plain description and follows visual
37
  instructions, so prefer it for a richer chat experience.
38
 
39
- Apache 2.0.
40
- Source: [github.com/andreagemelli/baguettotron-vlm](https://github.com/andreagemelli/baguettotron-vlm) ·
41
- Write-up: [andreagemelli.me/posts/baguettotron-vlm](https://andreagemelli.me/posts/baguettotron-vlm/)
42
 
43
 
44
  > **Alignment checkpoint.** Only the MLP projector was trained (on LLaVA-CC3M-Pretrain-595K) — the ViT and the Baguettotron LLM are unmodified base weights. It writes short captions and nothing more. Published so the alignment stage can be reproduced; for actual use take [baguettotron-vision-vqa](https://huggingface.co/andreagemelli/baguettotron-vision-vqa).
@@ -103,15 +105,6 @@ Greedy, prompt `Describe the image concisely.` — verbatim output. Images from
103
 
104
  Trained on short image captions with no `<think>` traces. The processor emits a bare assistant prefix (`<|im_start|>assistant\n`) and the model completes the caption directly. Keep prompts simple ("Describe the image").
105
 
106
- ## Limitations
107
-
108
- - **Resolution ceiling.** One 448×448 crop → 256 visual tokens puts document text at
109
- roughly 2–4 px/char. OCR, charts and documents are out of reach by architecture, not
110
- by budget. Neither published checkpoint reads text in an image.
111
- - **Hallucinations**, especially on fine-grained or text-heavy questions.
112
- - **Multilingual capability is inherited, not verified.** The backbone covers six
113
- languages; the VLM was never evaluated on non-English benchmarks.
114
-
115
  > Tested against `transformers` 4.57. Newer major versions may need adjustments.
116
 
117
  **Contributions and suggestions are very welcome** — issues, PRs, and ideas for
@@ -137,4 +130,4 @@ If you use or extend Baguettotron-VLM in your research, please cite it:
137
 
138
  ## License
139
 
140
- Apache 2.0 — see the [GitHub repo](https://github.com/andreagemelli/baguettotron-vlm).
 
16
  base_model:
17
  - OpenGVLab/InternViT-300M-448px-V2_5
18
  - PleIAs/Baguettotron
19
+ datasets:
20
+ - liuhaotian/LLaVA-CC3M-Pretrain-595K
21
  ---
22
 
23
  # baguettotron-internvit-alignment
 
38
  — instruction-tuned on top of it. It goes past plain description and follows visual
39
  instructions, so prefer it for a richer chat experience.
40
 
41
+
42
+ - Code: [github.com/andreagemelli/baguettotron-vlm](https://github.com/andreagemelli/baguettotron-vlm) ·
43
+ - Blogpost: [andreagemelli.me/posts/baguettotron-vlm](https://andreagemelli.me/posts/baguettotron-vlm/)
44
 
45
 
46
  > **Alignment checkpoint.** Only the MLP projector was trained (on LLaVA-CC3M-Pretrain-595K) — the ViT and the Baguettotron LLM are unmodified base weights. It writes short captions and nothing more. Published so the alignment stage can be reproduced; for actual use take [baguettotron-vision-vqa](https://huggingface.co/andreagemelli/baguettotron-vision-vqa).
 
105
 
106
  Trained on short image captions with no `<think>` traces. The processor emits a bare assistant prefix (`<|im_start|>assistant\n`) and the model completes the caption directly. Keep prompts simple ("Describe the image").
107
 
 
 
 
 
 
 
 
 
 
108
  > Tested against `transformers` 4.57. Newer major versions may need adjustments.
109
 
110
  **Contributions and suggestions are very welcome** — issues, PRs, and ideas for
 
130
 
131
  ## License
132
 
133
+ Apache 2.0 — see the [GitHub repo](https://github.com/andreagemelli/baguettotron-vlm).