docs: remove dangling repo reference, generalize voice

README.md

@@ -1,5 +1,5 @@
 ---
-pretty_name: MLX Benchmarks
+pretty_name: "MLX Benchmarks"
 license: apache-2.0
 language:
 - en
@@ -15,74 +15,31 @@ tags:
 - reasoning
 - math
 size_categories:
-- n<1K
+- "n<1K"
 configs:
 - config_name: default
   data_files:
   - split: train
-    path: data/train-*
-dataset_info:
-  features:
-  - name: suite
-    dtype: string
-  - name: model
-    dtype: string
-  - name: git_sha
-    dtype: string
-  - name: timestamp
-    dtype: string
-  - name: trigger
-    dtype: string
-  - name: schema_version
-    dtype: string
-  - name: pr_number
-    dtype: int64
-  - name: skipped
-    dtype: bool
-  - name: os
-    dtype: string
-  - name: chip
-    dtype: string
-  - name: memory_gb
-    dtype: int64
-  - name: vllm_mlx_version
-    dtype: string
-  - name: runner
-    dtype: string
-  - name: errors_json
-    dtype: string
-  - name: metric_name
-    dtype: string
-  - name: metric_metric
-    dtype: string
-  - name: metric_value
-    dtype: float64
-  - name: metric_unit
-    dtype: string
-  - name: tags_json
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 43190
-    num_examples: 140
-  download_size: 13945
-  dataset_size: 43190
+    path: "data/*.parquet"
 ---
 
 # MLX Benchmarks
 
 Structured benchmark results for **MLX-quantized** and other **locally-hosted
-LLMs** on Apple Silicon, covering throughput, time-to-first-token, tool-calling,
+LLMs** on Apple Silicon. Covers throughput, time-to-first-token, tool-calling,
 code generation, reasoning, knowledge, and math suites.
 
-Produced by the sweep harness in
-[`JacobPEvans/mlx-benchmarks` (GitHub)](https://github.com/JacobPEvans/mlx-benchmarks)
-which wires upstream tools against a local `vllm-mlx` inference server:
+Results are produced by a sweep harness that wires upstream evaluation tools
+against a local `vllm-mlx` inference server:
 
-- [`lm-evaluation-harness`](https://github.com/EleutherAI/lm-evaluation-harness) — coding/reasoning/knowledge/math
-- [`MLXBench`](https://github.com/linusvwe/MLXBench) — throughput and time-to-first-token
-- [`vllm benchmark_serving`](https://docs.vllm.ai/en/latest/performance/benchmarks.html) — perf second opinion
-- [`lighteval`](https://github.com/huggingface/lighteval) — broader task coverage
+- [EleutherAI/lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness) — coding, reasoning, knowledge, math
+- [linusvwe/MLXBench](https://github.com/linusvwe/MLXBench) — throughput and time-to-first-token
+- [vllm `benchmark_serving`](https://docs.vllm.ai/en/latest/performance/benchmarks.html) — performance second opinion
+- [huggingface/lighteval](https://github.com/huggingface/lighteval) — broader task coverage
+
+All data here is generated on Apple Silicon hardware (MINISFORUM MS-A2 / M4 Max
+class), stored in flat columnar Parquet for easy querying, and appended to via
+unique-filename commits so historical shards are never overwritten.
 
 ## Quickstart
 
@@ -92,27 +49,33 @@ from datasets import load_dataset
 ds = load_dataset("JacobPEvans/mlx-benchmarks")
 print(ds)
 
-# Example: average throughput per model on the throughput suite
+# Example: average throughput per model
 import pandas as pd
 df = ds["train"].to_pandas()
 throughput_rows = df[df.suite == "throughput"]
-print(throughput_rows.groupby("model")["metric_value"].mean().sort_values(ascending=False))
+print(
+    throughput_rows.groupby("model")["metric_value"]
+    .mean()
+    .sort_values(ascending=False)
+)
 ```
 
 Raw Parquet fetch (token-optimal for agents):
 
 ```bash
-curl -sSL https://huggingface.co/datasets/JacobPEvans/mlx-benchmarks/resolve/main/data/train-00000-of-00001.parquet -o run.parquet
+curl -sSL \
+  https://huggingface.co/datasets/JacobPEvans/mlx-benchmarks/resolve/main/data/train-00000-of-00001.parquet \
+  -o run.parquet
 ```
 
 ## Schema
 
-Each input JSON envelope (see `schema.json` for the authoritative v1 spec) is
-**exploded row-wise** into flat scalar columns — one row per metric entry in
-the envelope's `results[]` array. Skipped envelopes become a single sentinel
-row with null metric columns and `skipped=true`. This mirrors the columnar
-layout used by
-[`open-llm-leaderboard/contents`](https://huggingface.co/datasets/open-llm-leaderboard/contents).
+Each input benchmark run produces a JSON envelope (see `schema.json` in this
+repo for the authoritative v1 spec). The envelope is **exploded row-wise** into
+flat scalar columns — one row per entry in the envelope's `results[]` array.
+Skipped runs become a single sentinel row with null metric columns and
+`skipped=true`. This mirrors the columnar layout used by the
+[Open LLM Leaderboard contents dataset](https://huggingface.co/datasets/open-llm-leaderboard/contents).
 
 | Column | Type | Notes |
 | --- | --- | --- |
@@ -128,11 +91,11 @@ layout used by
 | `chip` | string | CPU/chip identifier |
 | `memory_gb` | int64 | Total system RAM |
 | `vllm_mlx_version` | string | Backend version if captured |
-| `runner` | string | GitHub Actions runner label or `local` |
+| `runner` | string | Runner label or `local` |
 | `metric_name` | string | Individual test/measurement name |
 | `metric_metric` | string | Metric family (e.g. `throughput`, `latency`, `score`) |
 | `metric_value` | float64 | Numeric value |
-| `metric_unit` | string | Unit (`tok/s`, `seconds`, `ratio`, etc.) |
+| `metric_unit` | string | Unit (`tok/s`, `seconds`, `ratio`, ...) |
 | `tags_json` | string | JSON-serialized tag dict (per-suite custom metadata) |
 | `errors_json` | string | JSON-serialized list of non-fatal errors from the run |
 
@@ -142,11 +105,11 @@ JSON-serialized strings so no information is lost — rehydrate with
 
 ## Update cadence
 
-New rows are appended on every sweep via `HfApi.create_commit` with unique
-filenames (`data/run-{timestamp}-{sha}-{suite}-{model}.parquet`). Historical
-shards are never overwritten. `load_dataset()` concatenates all
-`data/*.parquet` files into a single `train` split at load time.
+New rows are appended on every sweep via a unique-filename commit pattern
+(`data/run-{timestamp}-{sha}-{suite}-{model}.parquet`). Historical shards are
+never overwritten. `load_dataset()` concatenates all `data/*.parquet` files
+into a single `train` split at load time.
 
 ## License
 
-Apache 2.0 — same as the generator repo and the underlying upstream tools.
+Apache 2.0 — same as the underlying upstream evaluation tools.