Upload folder using huggingface_hub

.gitattributes

@@ -58,3 +58,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
 data.jsonl filter=lfs diff=lfs merge=lfs -text
+qwen3-32b-buckets.jsonl filter=lfs diff=lfs merge=lfs -text
+trace.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,121 @@
+---
+pretty_name: DataSys LLM Serving Trace
+license: other
+language:
+  - en
+tags:
+  - llm-serving
+  - inference
+  - request-trace
+  - token-reuse
+  - qwen3
+size_categories:
+  - 10M<n<100M
+configs:
+  - config_name: trace
+    data_files:
+      - split: train
+        path: trace.jsonl
+  - config_name: qwen3-32b-buckets
+    data_files:
+      - split: train
+        path: qwen3-32b-buckets.jsonl
+---
+
+# LLM Serving Trace
+
+This dataset contains anonymized LLM serving request metadata prepared for
+systems research on request patterns, token accounting, latency, and reusable
+input-token buckets. It is packaged as newline-delimited JSON.
+
+## Dataset Files
+
+| File | Rows | Size | Description |
+| --- | ---: | ---: | --- |
+| `trace.jsonl` | 16,329,237 | 6.98 GB | Request-level trace records with timestamps, status, model name, reported token counts, and generation parameters. |
+| `qwen3-32b-buckets.jsonl` | 3,994,435 | 4.62 GB | Qwen/Qwen3-32B subset augmented with token-bucket information for token reuse analysis. |
+
+## Loading
+
+```python
+from datasets import load_dataset
+
+trace = load_dataset("eth-easl/swissai-serving-trace", "trace", split="train")
+qwen_buckets = load_dataset(
+    "eth-easl/swissai-serving-trace",
+    "qwen3-32b-buckets",
+    split="train",
+)
+```
+
+For low-memory iteration, use streaming:
+
+```python
+from datasets import load_dataset
+
+rows = load_dataset(
+    "eth-easl/swissai-serving-trace",
+    "qwen3-32b-buckets",
+    split="train",
+    streaming=True,
+)
+
+first_row = next(iter(rows))
+```
+
+## Schema
+
+### `trace`
+
+Each row contains:
+
+| Field | Type | Description |
+| --- | --- | --- |
+| `id` | string | Request identifier. |
+| `status` | string | Request status, for example `DEFAULT` or `ERROR`. |
+| `created_at` | string | Request creation timestamp in ISO-8601 format. |
+| `finished_at` | string | Request completion timestamp in ISO-8601 format. |
+| `model` | string | Model identifier used for the request. |
+| `model_parameters` | object | Generation parameters supplied with the request. |
+| `reported_token_input` | int64 | Reported input-token count, or `-1` when unavailable. |
+| `reported_token_output` | int64 | Reported output-token count, or `-1` when unavailable. |
+
+`model_parameters` contains:
+
+| Field | Type |
+| --- | --- |
+| `temperature` | float64 |
+| `max_tokens` | string |
+| `top_p` | float64 |
+| `frequency_penalty` | float64 |
+| `presence_penalty` | float64 |
+| `seed` | int64 |
+
+### `qwen3-32b-buckets`
+
+This config contains the same request metadata fields as `trace`, plus:
+
+| Field | Type | Description |
+| --- | --- | --- |
+| `token_count` | int64 | Number of Qwen/Qwen3-32B input tokens retained for the request. |
+| `bucket_ids` | list[int64] | Deterministic token-bucket identifiers for the request input. Buckets were generated with 16-token buckets and right padding. |
+
+## Intended Uses
+
+This dataset is intended for research on LLM serving workloads, including:
+
+- request arrival and completion patterns;
+- latency and status analysis;
+- input/output token accounting;
+- cache locality and token reuse over bucketized input-token sequences.
+
+The bucketized Qwen subset is useful when the raw token IDs should not be
+distributed or when experiments only need stable identifiers for repeated token
+chunks.
+
+## Limitations
+
+The trace records contain serving metadata rather than original prompts or model
+outputs. Reported token counts may be unavailable for failed or incomplete
+requests, represented as `-1`. The `bucket_ids` are model- and preprocessing-
+specific and should not be interpreted as vocabulary token IDs.

datasys_trace.py

@@ -0,0 +1,127 @@
+"""Hugging Face Datasets loader for the DataSys LLM serving trace."""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+import datasets
+
+
+_DESCRIPTION = """\
+An anonymized LLM serving request trace for systems research on request
+patterns, token accounting, latency, and token-bucket reuse.
+"""
+
+_HOMEPAGE = ""
+_LICENSE = "other"
+
+
+class DataSysTraceConfig(datasets.BuilderConfig):
+    """Builder config for one JSONL file in the trace package."""
+
+    def __init__(self, *, filename: str, features: datasets.Features, **kwargs):
+        super().__init__(**kwargs)
+        self.filename = filename
+        self.features = features
+
+
+_MODEL_PARAMETERS = {
+    "temperature": datasets.Value("float64"),
+    "max_tokens": datasets.Value("string"),
+    "top_p": datasets.Value("float64"),
+    "frequency_penalty": datasets.Value("float64"),
+    "presence_penalty": datasets.Value("float64"),
+    "seed": datasets.Value("int64"),
+}
+
+_TRACE_FEATURES = datasets.Features(
+    {
+        "id": datasets.Value("string"),
+        "status": datasets.Value("string"),
+        "created_at": datasets.Value("string"),
+        "finished_at": datasets.Value("string"),
+        "model": datasets.Value("string"),
+        "model_parameters": _MODEL_PARAMETERS,
+        "reported_token_input": datasets.Value("int64"),
+        "reported_token_output": datasets.Value("int64"),
+    }
+)
+
+_QWEN_BUCKET_FEATURES = datasets.Features(
+    {
+        **_TRACE_FEATURES,
+        "token_count": datasets.Value("int64"),
+        "bucket_ids": datasets.Sequence(datasets.Value("int64")),
+    }
+)
+
+
+def _coerce_record(record: dict, include_buckets: bool) -> dict:
+    """Fill optional nested keys so strict feature schemas stay stable."""
+    model_parameters = record.get("model_parameters") or {}
+    record["model_parameters"] = {
+        "temperature": model_parameters.get("temperature"),
+        "max_tokens": (
+            None
+            if model_parameters.get("max_tokens") is None
+            else str(model_parameters.get("max_tokens"))
+        ),
+        "top_p": model_parameters.get("top_p"),
+        "frequency_penalty": model_parameters.get("frequency_penalty"),
+        "presence_penalty": model_parameters.get("presence_penalty"),
+        "seed": model_parameters.get("seed"),
+    }
+    if include_buckets:
+        record["bucket_ids"] = record.get("bucket_ids") or []
+    return record
+
+
+class DataSysTrace(datasets.GeneratorBasedBuilder):
+    """DataSys trace dataset builder."""
+
+    VERSION = datasets.Version("1.0.0")
+    DEFAULT_CONFIG_NAME = "trace"
+
+    BUILDER_CONFIGS = [
+        DataSysTraceConfig(
+            name="trace",
+            version=VERSION,
+            description="Full request-level trace metadata.",
+            filename="trace.jsonl",
+            features=_TRACE_FEATURES,
+        ),
+        DataSysTraceConfig(
+            name="qwen3-32b-buckets",
+            version=VERSION,
+            description="Qwen/Qwen3-32B subset with token-bucket IDs.",
+            filename="qwen3-32b-buckets.jsonl",
+            features=_QWEN_BUCKET_FEATURES,
+        ),
+    ]
+
+    def _info(self) -> datasets.DatasetInfo:
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=self.config.features,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+        )
+
+    def _split_generators(self, dl_manager: datasets.DownloadManager):
+        data_file = dl_manager.download_and_extract(self.config.filename)
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"filepath": data_file},
+            )
+        ]
+
+    def _generate_examples(self, filepath: str):
+        include_buckets = "bucket_ids" in self.config.features
+        with Path(filepath).open("r", encoding="utf-8") as handle:
+            for idx, line in enumerate(handle):
+                line = line.strip()
+                if not line:
+                    continue
+                yield idx, _coerce_record(json.loads(line), include_buckets)

qwen3-32b-buckets.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58b2265b98a10ec3bb2a0b455c520731c1296967906c7adc7b6d3710a013e402
+size 4617817342

trace.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79c92fd40752e40a6cac1fc897491ea2f54f37939dc4dcc48c96ea06383c81b7
+size 6981478609