Commit
•
4e6364b
1
Parent(s):
d6dbafc
upload new gpt weights
Browse files- micro_gpt_config.yaml +47 -0
- micro_gpt_weights.pt +3 -0
micro_gpt_config.yaml
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
attn_pdrop: 0.1
|
| 2 |
+
b1: 0.9
|
| 3 |
+
b2: 0.95
|
| 4 |
+
batch_size: 1024
|
| 5 |
+
blocks: 6
|
| 6 |
+
d_model: 128
|
| 7 |
+
embd_pdrop: 0.1
|
| 8 |
+
epochs: 500000
|
| 9 |
+
heads: 4
|
| 10 |
+
lr: 0.0003
|
| 11 |
+
model_type: gpt
|
| 12 |
+
num_workers: 4
|
| 13 |
+
resid_pdrop: 0.1
|
| 14 |
+
stoi:
|
| 15 |
+
' ': 1
|
| 16 |
+
'''': 2
|
| 17 |
+
'-': 3
|
| 18 |
+
.: 0
|
| 19 |
+
a: 4
|
| 20 |
+
b: 5
|
| 21 |
+
c: 6
|
| 22 |
+
d: 7
|
| 23 |
+
e: 8
|
| 24 |
+
f: 9
|
| 25 |
+
g: 10
|
| 26 |
+
h: 11
|
| 27 |
+
i: 12
|
| 28 |
+
j: 13
|
| 29 |
+
k: 14
|
| 30 |
+
l: 15
|
| 31 |
+
m: 16
|
| 32 |
+
n: 17
|
| 33 |
+
o: 18
|
| 34 |
+
p: 19
|
| 35 |
+
q: 20
|
| 36 |
+
r: 21
|
| 37 |
+
s: 22
|
| 38 |
+
t: 23
|
| 39 |
+
u: 24
|
| 40 |
+
v: 25
|
| 41 |
+
w: 26
|
| 42 |
+
x: 27
|
| 43 |
+
y: 28
|
| 44 |
+
z: 29
|
| 45 |
+
vocab: 30
|
| 46 |
+
weight_decay: 0.1
|
| 47 |
+
window: 32
|
micro_gpt_weights.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6584b4706a04079483a29e40028bb22945471961b01194d584806575ea8cd8df
|
| 3 |
+
size 4859909
|