Datasets:
|
Download README.md from kleinbem/edge-llm-benchmarks: direct link, hf CLI and curl.
- Browser
- Download file 2.43 kB
-
https://huggingface.co/datasets/kleinbem/edge-llm-benchmarks/resolve/main/README.md
- Command line
-
hf download hf://datasets/kleinbem/edge-llm-benchmarks/README.md
-
curl -L -o README.md https://huggingface.co/datasets/kleinbem/edge-llm-benchmarks/resolve/main/README.md
2.43 kB
metadata
license: mit
task_categories:
- text-generation
tags:
- benchmark
- quantization
- kl-divergence
- gguf
- llama-cpp
- jetson-orin-nano
- energy-efficiency
- edge-ai
pretty_name: Edge & Local GGUF Quantization, KL-Divergence & Power Telemetry
size_categories:
- n<1K
configs:
- config_name: default
data_files:
- split: train
path: data/benchmarks.parquet
Edge & Local GGUF Quantization, KL-Divergence & Power Telemetry
Hardware throughput (pp512, tg128), exact FP16 logit Kullback-Leibler divergence (mean_kld, p99_kld, same_top1_pct), unified-memory OOM boundaries, and INA3221 / RAPL energy per token (Joules/token) collected reproducibly via github.com/kleinbem/hf-bench.
- Interactive Explorer Space:
spaces/kleinbem/edge-llm-pareto
Evaluated Base Models
base_models:
- Qwen/Qwen2.5-Coder-7B-Instruct
- Qwen/Qwen3-4B
Hardware Targets
orin-nano: NVIDIA Jetson Orin Nano (8 GiB unified LPDDR5, Ampere 1024-core Tegra GPUsm_87, NixOS + JetPack 6 CUDA 12.6, INA3221 hardware power rail sampling).nixos-nvme: 13th Gen Intel Core i3-1315U (x86_64Raptor Lake AVX2, 62 GiB DDR5, NixOS idle scheduler priority, Intel RAPL package energy sampling).
Schema Highlights
| Column | Type | Description |
|---|---|---|
model_name |
string |
Base model architecture name |
quant |
string |
llama.cpp quantization (Q8_0 .. IQ3_M, plus Q4_K_M-std uncalibrated control) |
imatrix_used |
bool |
Whether agentic + code importance matrix calibration was applied |
bpw |
float64 |
Effective bits per weight (weight_bytes * 8 / params) |
status |
string |
ok or oom when exceeding the device's unified memory cliff |
mean_kld |
float64 |
Mean Kullback-Leibler divergence (nats) vs exact FP16 base logits |
same_top1_pct |
float64 |
Percentage of evaluation tokens sharing the exact FP16 argmax prediction |
pp512_tok_s |
float64 |
512-token prompt prefill speed (tokens/sec) |
tg128_tok_s |
float64 |
128-token autoregressive decode speed (tokens/sec) |
avg_watts |
float64 |
Mean board/package power draw during inference (Watts) |
joules_per_token |
float64 |
Energy consumed per generated token (avg_watts / tg128_tok_s) |