Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -11,7 +11,14 @@ tags:
|
|
| 11 |
- homelab
|
| 12 |
- open-data
|
| 13 |
- consumer-hardware
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
pretty_name: Poor Paul's Benchmark Results
|
|
|
|
|
|
|
| 15 |
size_categories:
|
| 16 |
- 1K<n<10K
|
| 17 |
---
|
|
@@ -43,6 +50,16 @@ Current benchmark runners represented in the dataset include:
|
|
| 43 |
- `llama-server` — TTFT and ITL latency with real conversational prompts
|
| 44 |
- `llama-server-loadtest` — max concurrency discovery and concurrency curves
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
## Dataset Structure
|
| 47 |
|
| 48 |
Each row corresponds to one normalized benchmark result.
|
|
|
|
| 11 |
- homelab
|
| 12 |
- open-data
|
| 13 |
- consumer-hardware
|
| 14 |
+
- poor-pauls-benchmark
|
| 15 |
+
- gpu-benchmark
|
| 16 |
+
- quantization
|
| 17 |
+
- vram
|
| 18 |
+
- mcp
|
| 19 |
pretty_name: Poor Paul's Benchmark Results
|
| 20 |
+
language:
|
| 21 |
+
- en
|
| 22 |
size_categories:
|
| 23 |
- 1K<n<10K
|
| 24 |
---
|
|
|
|
| 50 |
- `llama-server` — TTFT and ITL latency with real conversational prompts
|
| 51 |
- `llama-server-loadtest` — max concurrency discovery and concurrency curves
|
| 52 |
|
| 53 |
+
## Ecosystem
|
| 54 |
+
|
| 55 |
+
| Component | Link |
|
| 56 |
+
|---|---|
|
| 57 |
+
| Benchmark tool | [poor-pauls-benchmark](https://github.com/paulplee/poor-pauls-benchmark) |
|
| 58 |
+
| MCP server | [ppb-mcp](https://github.com/paulplee/ppb-mcp) — lets LLMs query this dataset directly |
|
| 59 |
+
| Analytics dashboard | [poorpaul.dev/insights](https://poorpaul.dev/insights) |
|
| 60 |
+
|
| 61 |
+
Ask an LLM to query this dataset: connect any MCP client to `mcp.poorpaul.dev`.
|
| 62 |
+
|
| 63 |
## Dataset Structure
|
| 64 |
|
| 65 |
Each row corresponds to one normalized benchmark result.
|