Hugging Face's logo

inference-optimization
/
Llama-3.1-8B-Instruct-QKV-Cache-FP8-Per-Head

Safetensors
llama
compressed-tensors
Model card Files
xet
Community
Llama-3.1-8B-Instruct-QKV-Cache-FP8-Per-Head
File size: 2,188 Bytes 
ca74a77
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
193e4d3
a75de39
 
a5ead57
 
ca74a77
4b13fac
281a2cc
ca74a77
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
281a2cc
ca74a77
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
 
---
license: apache-2.0
base_model:
- meta-llama/Llama-3.1-8B-Instruct
---



### Accuracy
<table>
  <thead>
    <tr>
      <th>Task</th>
      <th>Context Length</th>
      <th>meta-llama/<br>Llama-3.1-8B-Instruct</th>
      <th>Llama-3.1-8B-Instruct-<br>FP8-dynamic-<br>QKV-Cache-FP8-<br>Per-Head</th>
      <th>Llama-3.1-8B-Instruct-<br>FP8-dynamic-<br>QKV-Cache-FP8-<br>Per-Tensor</th>
      <th>Llama-3.1-8B-Instruct-<br>QKV-Cache-FP8-<br>Per-Head</th>
      <th>Llama-3.1-8B-Instruct-<br>QKV-Cache-FP8-<br>Per-Tensor</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td rowspan="5"><b>NIAH<br>Single 2</b></td>
      <td>4096</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>16384</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>32768</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>65536</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>131072</td>
      <td>99.2</td>
      <td>99.6</td>
      <td>99.4</td>
      <td>99.4</td>
      <td>99.0</td>
    </tr>
    <br>
    <!-- <tr>
      <td rowspan="5"><b>NIAH<br>Single 3</b></td>
      <td>4096</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>16384</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>32768</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>65536</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr>
    <tr>
      <td>131072</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
      <td>100.00</td>
    </tr> -->
  </tbody>
</table>