Fix minor typo
#45
by
pcuenq
HF staff
- opened
README.md
CHANGED
@@ -163,7 +163,7 @@ Falcon-7B is a causal decoder-only model trained on a causal language modeling t
|
|
163 |
|
164 |
The architecture is broadly adapted from the GPT-3 paper ([Brown et al., 2020](https://arxiv.org/abs/2005.14165)), with the following differences:
|
165 |
|
166 |
-
* **
|
167 |
* **Attention:** multiquery ([Shazeer et al., 2019](https://arxiv.org/abs/1911.02150)) and FlashAttention ([Dao et al., 2022](https://arxiv.org/abs/2205.14135));
|
168 |
* **Decoder-block:** parallel attention/MLP with a single layer norm.
|
169 |
|
|
|
163 |
|
164 |
The architecture is broadly adapted from the GPT-3 paper ([Brown et al., 2020](https://arxiv.org/abs/2005.14165)), with the following differences:
|
165 |
|
166 |
+
* **Positional embeddings:** rotary ([Su et al., 2021](https://arxiv.org/abs/2104.09864));
|
167 |
* **Attention:** multiquery ([Shazeer et al., 2019](https://arxiv.org/abs/1911.02150)) and FlashAttention ([Dao et al., 2022](https://arxiv.org/abs/2205.14135));
|
168 |
* **Decoder-block:** parallel attention/MLP with a single layer norm.
|
169 |
|