binesh commited on
Commit
4136cc2
·
verified ·
1 Parent(s): 9c2651d

Card: single model

Browse files
Files changed (1) hide show
  1. README.md +4 -25
README.md CHANGED
@@ -29,22 +29,7 @@ sensitivity in medical vision-language models originates*.
29
  > **Scope.** A research probe, **not a clinical model**. It answers binary presence
30
  > questions about 14 chest findings. Not a medical device; not for clinical use.
31
 
32
- ## Two variants
33
-
34
- | | default (this repo root) | `probe-1841/` |
35
- |---|---|---|
36
- | Purpose | grounded model that **transfers to unseen hospitals** | the **controlled probe** behind the dissertation's causal experiments |
37
- | Training data | 107k per-finding-balanced questions (NIH + PadChest) | 1,841 questions (MIMIC-CXR + PadChest) |
38
- | Tokenizer | MedGemma SentencePiece, pruned to 141 pieces | word-level, 734 words |
39
- | Paraphrases | 48 templates across 4 phenomena | 4.8 generated per question |
40
- | Grounding token | yes (MedSigLIP pooled embedding) | no |
41
-
42
- The dissertation's Section 5.5 now describes the **default** model. `probe-1841/` is
43
- kept for reproducing the earlier published figures (augmentation lever 29.5% -> 9.5%)
44
- and the experiments not re-run at scale (divergence trajectory, decoder-depth sweep,
45
- vision-dropout grounding sweep).
46
-
47
- ## What the default model does
48
 
49
  Trained on NIH + PadChest, evaluated on **MIMIC and VinDr held out entirely** (no
50
  image, and no hospital, seen in training). Every split is balanced per finding, so a
@@ -108,22 +93,16 @@ logits = model(input_ids=input_ids, vision_features=vision_features,
108
  print(model.config.id2label[int(logits.argmax(-1))]) # "yes" or "no"
109
  ```
110
 
111
- The dissertation probe loads the same way with `subfolder="probe-1841"` (it takes no
112
- `ground` argument). `encode_images` needs gated access to `google/medsiglip-448`.
113
- `feature_cache/` and `checkpoints/` in this repo belong to the **probe** variant.
114
 
115
  ## Repository structure
116
 
117
  ```
118
- config.json model.safetensors modeling_babymedgemma.py scaled grounded model (default)
119
- probe-1841/ the 1,841-question dissertation probe
120
- feature_cache/medsiglip_feats.pt 896-pixel MedSigLIP cache (probe variant)
121
- checkpoints/{B,C,D}/ per-run BabyGemmaVLM state_dicts (probe variant)
122
  ```
123
 
124
  Training and analysis code, and the exact scripts that produced every number here, are on
125
- GitHub: [thedatasense/babymedgemma](https://github.com/thedatasense/babymedgemma). Load
126
- either model with the same call; the probe needs `subfolder="probe-1841"`.
127
 
128
  ## Citation
129
 
 
29
  > **Scope.** A research probe, **not a clinical model**. It answers binary presence
30
  > questions about 14 chest findings. Not a medical device; not for clinical use.
31
 
32
+ ## What it does
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
 
34
  Trained on NIH + PadChest, evaluated on **MIMIC and VinDr held out entirely** (no
35
  image, and no hospital, seen in training). Every split is balanced per finding, so a
 
93
  print(model.config.id2label[int(logits.argmax(-1))]) # "yes" or "no"
94
  ```
95
 
96
+ `encode_images` needs gated access to `google/medsiglip-448`.
 
 
97
 
98
  ## Repository structure
99
 
100
  ```
101
+ config.json model.safetensors modeling_babymedgemma.py the model + self-contained wrapper
 
 
 
102
  ```
103
 
104
  Training and analysis code, and the exact scripts that produced every number here, are on
105
+ GitHub: [thedatasense/babymedgemma](https://github.com/thedatasense/babymedgemma).
 
106
 
107
  ## Citation
108