Update README.md
Browse files
README.md
CHANGED
@@ -18,33 +18,6 @@ The goal is to adapt a powerful **multimodal vision-language model** for **medic
|
|
18 |
|
19 |
---
|
20 |
|
21 |
-
## Model Details
|
22 |
-
|
23 |
-
- **Base model:** Qwen/Qwen2.5-VL-3B-Instruct
|
24 |
-
- **Adapter type:** LoRA (PEFT)
|
25 |
-
- **Training objective:** Supervised fine-tuning (SFT) on chest X-ray reports
|
26 |
-
- **Dataset:** [MIMIC-CXR](https://physionet.org/content/mimic-cxr/2.0.0/) (radiology images + reports)
|
27 |
-
- **Languages:** English (medical reporting domain)
|
28 |
-
- **Frameworks:** `transformers`, `peft`, `trl`
|
29 |
-
|
30 |
-
---
|
31 |
-
|
32 |
-
## Intended Uses
|
33 |
-
|
34 |
-
### Direct Use
|
35 |
-
- Generating radiology-style reports from chest X-ray images.
|
36 |
-
- Research on applying large multimodal models to medical imaging tasks.
|
37 |
-
|
38 |
-
### Downstream Use
|
39 |
-
- Medical text generation tasks where radiological image context is available.
|
40 |
-
- Adaptation for other healthcare VQA (Visual Question Answering) tasks.
|
41 |
-
|
42 |
-
### Out-of-Scope Use
|
43 |
-
⚠️ **Not for clinical decision-making.**
|
44 |
-
This model is intended **for research purposes only**. Do not use it in medical practice without proper validation and regulatory approval.
|
45 |
-
|
46 |
-
---
|
47 |
-
|
48 |
## How to Use
|
49 |
|
50 |
```python
|
@@ -99,3 +72,29 @@ sample = [
|
|
99 |
output = generate_text_from_sample(model, processor, sample)
|
100 |
print(output)
|
101 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
18 |
|
19 |
---
|
20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
## How to Use
|
22 |
|
23 |
```python
|
|
|
72 |
output = generate_text_from_sample(model, processor, sample)
|
73 |
print(output)
|
74 |
```
|
75 |
+
---
|
76 |
+
|
77 |
+
## Model Details
|
78 |
+
|
79 |
+
- **Base model:** Qwen/Qwen2.5-VL-3B-Instruct
|
80 |
+
- **Adapter type:** LoRA (PEFT)
|
81 |
+
- **Training objective:** Supervised fine-tuning (SFT) on chest X-ray reports
|
82 |
+
- **Dataset:** [MIMIC-CXR](https://physionet.org/content/mimic-cxr/2.0.0/) (radiology images + reports)
|
83 |
+
- **Languages:** English (medical reporting domain)
|
84 |
+
- **Frameworks:** `transformers`, `peft`, `trl`
|
85 |
+
|
86 |
+
---
|
87 |
+
|
88 |
+
## Intended Uses
|
89 |
+
|
90 |
+
### Direct Use
|
91 |
+
- Generating radiology-style reports from chest X-ray images.
|
92 |
+
- Research on applying large multimodal models to medical imaging tasks.
|
93 |
+
|
94 |
+
### Downstream Use
|
95 |
+
- Medical text generation tasks where radiological image context is available.
|
96 |
+
- Adaptation for other healthcare VQA (Visual Question Answering) tasks.
|
97 |
+
|
98 |
+
### Out-of-Scope Use
|
99 |
+
⚠️ **Not for clinical decision-making.**
|
100 |
+
This model is intended **for research purposes only**. Do not use it in medical practice without proper validation and regulatory approval.
|