Views
No views yet
1dtype: float32
2merge_method: passthrough
3slices:
4- sources:
5 - layer_range: [0, 20]
6 model: "jondurbin_bagel-34b-v0.2"
7 parameters:
8 scale:
9 - filter: q_proj
10 value: 0.7071067812
11 - filter: k_proj
12 value: 0.7071067812
13 - value: 1
14- sources:
15 - layer_range: [10, 30]
16 model: "jondurbin_bagel-34b-v0.2"
17 parameters:
18 scale:
19 - filter: q_proj
20 value: 0.7071067812
21 - filter: k_proj
22 value: 0.7071067812
23 - value: 1
24- sources:
25 - layer_range: [20, 40]
26 model: "jondurbin_bagel-34b-v0.2"
27 parameters:
28 scale:
29 - filter: q_proj
30 value: 0.7071067812
31 - filter: k_proj
32 value: 0.7071067812
33 - value: 1
34- sources:
35 - layer_range: [30, 50]
36 model: "jondurbin_bagel-34b-v0.2"
37 parameters:
38 scale:
39 - filter: q_proj
40 value: 0.7071067812
41 - filter: k_proj
42 value: 0.7071067812
43 - value: 1
44- sources:
45 - layer_range: [40, 60]
46 model: "jondurbin_bagel-34b-v0.2"
47 parameters:
48 scale:
49 - filter: q_proj
50 value: 0.7071067812
51 - filter: k_proj
52 value: 0.7071067812
53 - value: 1
54name: 2xbagel_fp32
55---
56dtype: bfloat16
57merge_method: passthrough
58slices:
59- sources:
60 - layer_range: [0, 100]
61 model: 2xbagel_fp32
62name: bagel_new
63