TristanBehrens
commited on
Commit
•
17b266c
1
Parent(s):
c0aac03
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
tags:
|
5 |
+
- NLP
|
6 |
+
license: mit
|
7 |
+
datasets:
|
8 |
+
- TristanBehrens/bach_garland_2024-100K
|
9 |
+
base_model: None
|
10 |
+
---
|
11 |
+
|
12 |
+
# bach_garland_mamba - An xLSTM Model
|
13 |
+
|
14 |
+
![Trained with Helibrunna](banner.jpg)
|
15 |
+
|
16 |
+
Trained with [Helibrunna](https://github.com/AI-Guru/helibrunna) by [Dr. Tristan Behrens](https://de.linkedin.com/in/dr-tristan-behrens-734967a2).
|
17 |
+
|
18 |
+
## Configuration
|
19 |
+
|
20 |
+
```
|
21 |
+
training:
|
22 |
+
model_name: bach_garland_mamba
|
23 |
+
batch_size: 28
|
24 |
+
lr: 0.001
|
25 |
+
lr_warmup_steps: 1428
|
26 |
+
lr_decay_until_steps: 14285
|
27 |
+
lr_decay_factor: 0.001
|
28 |
+
weight_decay: 0.1
|
29 |
+
amp_precision: bfloat16
|
30 |
+
weight_precision: float32
|
31 |
+
enable_mixed_precision: true
|
32 |
+
num_epochs: 8
|
33 |
+
output_dir: output/bach_garland_mamba
|
34 |
+
save_every_step: 500
|
35 |
+
log_every_step: 10
|
36 |
+
wandb_project: bach_garland
|
37 |
+
torch_compile: false
|
38 |
+
model:
|
39 |
+
type: mamba
|
40 |
+
d_model: 64
|
41 |
+
n_layers: 4
|
42 |
+
context_length: 4096
|
43 |
+
vocab_size: 178
|
44 |
+
dataset:
|
45 |
+
hugging_face_id: TristanBehrens/bach_garland_2024-100K
|
46 |
+
tokenizer:
|
47 |
+
type: whitespace
|
48 |
+
fill_token: '[EOS]'
|
49 |
+
|
50 |
+
```
|