File size: 1,034 Bytes
13f1e59
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Barist.1 — Benchmark

Generated by `scripts/benchmark.py`.

## Hardware

- Device: `AMD Radeon Graphics` (cuda)
- torch: `2.13.0+rocm7.1`
- Python: `3.12.12`
- OS: `Linux-7.1.7-100.fc43.x86_64-x86_64-with-glibc2.43`

## Model

| Property | Value |
| --- | --- |
| Parameters | 903,552 |
| Checkpoint size | 3.45 MB |
| Vocabulary | 2048 |
| Context | 4096 tokens |
| Input limit | 4000 characters |
| Layers / d_model / heads / d_ff | 3 / 128 / 4 / 384 |

## Inference

| Metric | Value |
| --- | --- |
| Single text latency | 1.04 ms |
| Single text throughput | 958.1 texts/s |
| Batch of 8 latency | 1.23 ms |
| Batched throughput | 6490.0 texts/s |
| Peak VRAM (inference) | 84.7 MB |

## Accuracy

| Split | Rows | Accuracy | F1 | Precision | Recall |
| --- | --- | --- | --- | --- | --- |
| val | 6585 | 0.8535 | 0.8175 | 0.8498 | 0.7875 |
| test | 6496 | 0.8570 | 0.8225 | 0.8519 | 0.7950 |
| probe (hand-written) | 97 | 0.9897 | 0.9882 | 0.9767 | 1.0000 |

The probe set is hand-written and absent from all training data.