Spaces:
Configuration error
Configuration error
Update README.md
Browse files
README.md
CHANGED
|
@@ -7,8 +7,6 @@ We build things that are:
|
|
| 7 |
- layered 🧠
|
| 8 |
- occasionally overengineered 😏
|
| 9 |
|
| 10 |
-
---
|
| 11 |
-
|
| 12 |
## 🍞 What’s on the menu?
|
| 13 |
|
| 14 |
### 🧠 DaMorph
|
|
@@ -17,12 +15,9 @@ A collection of experimental models exploring **morphological segmentation for D
|
|
| 17 |
👉 Because Danish words are long… and deserve to be sliced properly.
|
| 18 |
|
| 19 |
**Includes:**
|
| 20 |
-
-
|
| 21 |
-
-
|
| 22 |
-
-
|
| 23 |
-
- Tokenizers (yes, we slice at every layer)
|
| 24 |
-
|
| 25 |
-
---
|
| 26 |
|
| 27 |
### 🏥 DaMedSum
|
| 28 |
Danish medical summarization models trained on **LUMI HPC**.
|
|
@@ -33,16 +28,12 @@ Danish medical summarization models trained on **LUMI HPC**.
|
|
| 33 |
- T5-large
|
| 34 |
- large / base / small variants
|
| 35 |
|
| 36 |
-
---
|
| 37 |
-
|
| 38 |
### 🔪 Tokenizers (a.k.a. precision slicing)
|
| 39 |
Because no good sandwich starts without proper slicing.
|
| 40 |
|
| 41 |
- Morphological tokenizers for Danish
|
| 42 |
- Built to explore how structure impacts understanding
|
| 43 |
|
| 44 |
-
---
|
| 45 |
-
|
| 46 |
## 🥓 Philosophy
|
| 47 |
|
| 48 |
Why “OpenFacedSandwich”?
|
|
@@ -55,8 +46,6 @@ Because:
|
|
| 55 |
👉 We don’t just build models
|
| 56 |
👉 we **assemble sandwiches**
|
| 57 |
|
| 58 |
-
---
|
| 59 |
-
|
| 60 |
## 😂 What to expect
|
| 61 |
|
| 62 |
- Serious experiments 🤓
|
|
@@ -64,16 +53,12 @@ Because:
|
|
| 64 |
- Danish NLP in all its glory 🇩🇰
|
| 65 |
- Things that *probably* shouldn’t work… but do
|
| 66 |
|
| 67 |
-
---
|
| 68 |
-
|
| 69 |
## 🧑🍳 Contribute
|
| 70 |
|
| 71 |
Got a model? A tokenizer? A weird idea?
|
| 72 |
|
| 73 |
👉 Add a layer.
|
| 74 |
|
| 75 |
-
---
|
| 76 |
-
|
| 77 |
## ⚡ Slogan
|
| 78 |
|
| 79 |
**Open source never tasted this good.**
|
|
|
|
| 7 |
- layered 🧠
|
| 8 |
- occasionally overengineered 😏
|
| 9 |
|
|
|
|
|
|
|
| 10 |
## 🍞 What’s on the menu?
|
| 11 |
|
| 12 |
### 🧠 DaMorph
|
|
|
|
| 15 |
👉 Because Danish words are long… and deserve to be sliced properly.
|
| 16 |
|
| 17 |
**Includes:**
|
| 18 |
+
- DaMorph
|
| 19 |
+
- DaMorph Tokenizers (yes, we slice at every layer)
|
| 20 |
+
- DaMedSum
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
### 🏥 DaMedSum
|
| 23 |
Danish medical summarization models trained on **LUMI HPC**.
|
|
|
|
| 28 |
- T5-large
|
| 29 |
- large / base / small variants
|
| 30 |
|
|
|
|
|
|
|
| 31 |
### 🔪 Tokenizers (a.k.a. precision slicing)
|
| 32 |
Because no good sandwich starts without proper slicing.
|
| 33 |
|
| 34 |
- Morphological tokenizers for Danish
|
| 35 |
- Built to explore how structure impacts understanding
|
| 36 |
|
|
|
|
|
|
|
| 37 |
## 🥓 Philosophy
|
| 38 |
|
| 39 |
Why “OpenFacedSandwich”?
|
|
|
|
| 46 |
👉 We don’t just build models
|
| 47 |
👉 we **assemble sandwiches**
|
| 48 |
|
|
|
|
|
|
|
| 49 |
## 😂 What to expect
|
| 50 |
|
| 51 |
- Serious experiments 🤓
|
|
|
|
| 53 |
- Danish NLP in all its glory 🇩🇰
|
| 54 |
- Things that *probably* shouldn’t work… but do
|
| 55 |
|
|
|
|
|
|
|
| 56 |
## 🧑🍳 Contribute
|
| 57 |
|
| 58 |
Got a model? A tokenizer? A weird idea?
|
| 59 |
|
| 60 |
👉 Add a layer.
|
| 61 |
|
|
|
|
|
|
|
| 62 |
## ⚡ Slogan
|
| 63 |
|
| 64 |
**Open source never tasted this good.**
|