Update README.md
Browse files
README.md
CHANGED
|
@@ -61,8 +61,14 @@ histopathology slides. Its LLM component is trained on a diverse set of medical
|
|
| 61 |
data, including radiology images, histopathology patches, ophthalmology images,
|
| 62 |
and dermatology images.
|
| 63 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
MedGemma 27B has been trained exclusively on medical text and optimized for
|
| 65 |
-
inference-time computation.
|
|
|
|
| 66 |
|
| 67 |
MedGemma variants have been evaluated on a range of clinically relevant
|
| 68 |
benchmarks to illustrate their baseline performance. These include both open
|
|
|
|
| 61 |
data, including radiology images, histopathology patches, ophthalmology images,
|
| 62 |
and dermatology images.
|
| 63 |
|
| 64 |
+
MedGemma 4B is available in both pre-trained (suffix: `-pt`) and
|
| 65 |
+
instruction-tuned (suffix `-it`) versions. The instruction-tuned version is a
|
| 66 |
+
better starting point for most applications. The pre-trained version notably
|
| 67 |
+
achieves better performance on MIMIC-style chest X-ray reporting.
|
| 68 |
+
|
| 69 |
MedGemma 27B has been trained exclusively on medical text and optimized for
|
| 70 |
+
inference-time computation. MedGemma 27B is only available as an
|
| 71 |
+
instruction-tuned model.
|
| 72 |
|
| 73 |
MedGemma variants have been evaluated on a range of clinically relevant
|
| 74 |
benchmarks to illustrate their baseline performance. These include both open
|