eyad-silx commited on
Commit
875ec1c
·
verified ·
1 Parent(s): 04701f7

Add HFA validation results: validation_report.json

Browse files
Files changed (1) hide show
  1. validation_report.json +54 -0
validation_report.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "executive_summary": {
3
+ "total_tests": 7,
4
+ "hfa_wins": 2,
5
+ "standard_wins": 0,
6
+ "key_findings": [
7
+ "HFA outperforms standard attention on pattern_recognition by 97.8%",
8
+ "HFA outperforms standard attention on computational_efficiency by 36335.1%"
9
+ ]
10
+ },
11
+ "detailed_results": {
12
+ "pattern_recognition": {
13
+ "hfa_performance": 0.29534883408391155,
14
+ "standard_performance": 0.1493469230324704,
15
+ "improvement_ratio": 1.977602404434526,
16
+ "sequence_lengths_tested": [
17
+ 32,
18
+ 64,
19
+ 128
20
+ ],
21
+ "metadata": {
22
+ "pattern_type": "alternating_with_markers"
23
+ }
24
+ },
25
+ "computational_efficiency": {
26
+ "hfa_performance": 0.20173163414001466,
27
+ "standard_performance": 0.000553673505783081,
28
+ "improvement_ratio": 364.3512503902424,
29
+ "sequence_lengths_tested": [
30
+ 32,
31
+ 64,
32
+ 128,
33
+ 256
34
+ ],
35
+ "metadata": {
36
+ "hfa_speed": "586",
37
+ "standard_speed": "451076",
38
+ "unit": "tokens_per_second"
39
+ }
40
+ }
41
+ },
42
+ "visualizations": [],
43
+ "reproducibility": {
44
+ "torch_version": "2.7.1+cu128",
45
+ "cuda_version": "12.8",
46
+ "random_seed": 42,
47
+ "hardware_info": {
48
+ "gpu_name": "NVIDIA GeForce RTX 3090 Ti",
49
+ "gpu_memory": 25294995456,
50
+ "cpu_count": 32,
51
+ "ram_total": 405372833792
52
+ }
53
+ }
54
+ }