Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
title: EXD Benchmark Results
|
| 3 |
+
emoji: 📊
|
| 4 |
+
colorFrom: gray
|
| 5 |
+
colorTo: red
|
| 6 |
+
sdk: static
|
| 7 |
+
pinned: false
|
| 8 |
+
license: mit
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# EXD Benchmark Results
|
| 12 |
+
|
| 13 |
+
Performance data from inference sweeps on the Gigabyte AI TOP (DGX Spark-class, NVIDIA GB10).
|
| 14 |
+
|
| 15 |
+
## Contents
|
| 16 |
+
|
| 17 |
+
| File | Source | Description |
|
| 18 |
+
|------|--------|-------------|
|
| 19 |
+
| `episodes/Ep03/03_bench.txt` | Episode 3 | First benchmark run — Qwen2.5-7B, TTFT, TPOT, throughput |
|
| 20 |
+
| `benchmarks/qwen3.6-35b-a3b-llama-benchy.md` | Serve harness | Qwen3.6-35B-A3B with llama-benchy |
|
| 21 |
+
|
| 22 |
+
## Hardware
|
| 23 |
+
|
| 24 |
+
- **Machine:** Gigabyte AI TOP (DGX Spark variant)
|
| 25 |
+
- **SoC:** NVIDIA Grace-Blackwell (GB10)
|
| 26 |
+
- **Memory:** 128 GB unified (CPU + GPU shared)
|
| 27 |
+
- **Runtime:** vLLM inside Docker (NGC container)
|
| 28 |
+
|
| 29 |
+
## Related
|
| 30 |
+
|
| 31 |
+
- [Org hub](https://huggingface.co/EXD-AI)
|
| 32 |
+
- [GitHub](https://github.com/Ramshreyas/EXD)
|