Refer dataset guidance to docs and record v3.2.0 training selections
Browse files
README.md
CHANGED
|
@@ -29,7 +29,7 @@ Run directories include checkpoints with optimizer/resume state, configs, model
|
|
| 29 |
kwargs, histories, training logs, TensorBoard streams and final plots.
|
| 30 |
All runs use seed 12345 and training revision
|
| 31 |
[`6b80b983`](https://github.com/jpata/particleflow/commit/6b80b983477321cf708e5043fb1d2ec15dd398bc).
|
| 32 |
-
|
| 33 |
|
| 34 |
## Training and validation
|
| 35 |
|
|
@@ -141,9 +141,25 @@ for study summaries, archived code, datasets and citation guidance.
|
|
| 141 |
|
| 142 |
## Datasets
|
| 143 |
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
|
|
|
|
| 149 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
kwargs, histories, training logs, TensorBoard streams and final plots.
|
| 30 |
All runs use seed 12345 and training revision
|
| 31 |
[`6b80b983`](https://github.com/jpata/particleflow/commit/6b80b983477321cf708e5043fb1d2ec15dd398bc).
|
| 32 |
+
See [training datasets](#datasets-used-for-v320-training) for the TFDS selections.
|
| 33 |
|
| 34 |
## Training and validation
|
| 35 |
|
|
|
|
| 141 |
|
| 142 |
## Datasets
|
| 143 |
|
| 144 |
+
The [dataset catalog](https://jpata.github.io/particleflow/datasets/catalog) is the
|
| 145 |
+
main reference for dataset definitions, schema compatibility and published availability.
|
| 146 |
+
Follow the [download guide](https://jpata.github.io/particleflow/datasets/download)
|
| 147 |
+
for retrieval and verification.
|
| 148 |
|
| 149 |
+
### Datasets used for v3.2.0 training
|
| 150 |
|
| 151 |
+
Both backbones use the selections below, as recorded in each model directory's
|
| 152 |
+
`train-config.yaml`. All runs combine configuration partitions **1–10**.
|
| 153 |
+
|
| 154 |
+
| Detector / inputs | TFDS training datasets | TFDS version |
|
| 155 |
+
| --- | --- | --- |
|
| 156 |
+
| [CMS tracks/clusters](https://jpata.github.io/particleflow/datasets/cms) | `cms_pf_{ttbar,qcd,ztt}` and `cms_pf_{ttbar,qcd,ztt}_nopu` | 2.8.0 |
|
| 157 |
+
| [CLIC tracks/clusters](https://jpata.github.io/particleflow/datasets/clic) | `clic_edm_{ttbar,ww_fullhad,qq}_pf` | 3.3.0 |
|
| 158 |
+
| [CLIC hits](https://jpata.github.io/particleflow/datasets/clic-hits) | `clic_edm_{ttbar,ww_fullhad,qq}_hits` | 3.2.2 |
|
| 159 |
+
| [CLD tracks/clusters](https://jpata.github.io/particleflow/datasets/cld) | `cld_edm_{ttbar,ww_fullhad,qq}_pf` | 3.3.0 |
|
| 160 |
+
| [CLD hits](https://jpata.github.io/particleflow/datasets/cld-hits) | `cld_edm_{ttbar,ww_fullhad,qq}_hits` | 3.2.2 |
|
| 161 |
+
|
| 162 |
+
TFDS schema versions and model release versions are independent. The current CMS
|
| 163 |
+
production guide describes schema 3.2.0; these checkpoints were trained on 2.8.0.
|
| 164 |
+
The catalog's [published-availability section](https://jpata.github.io/particleflow/datasets/catalog#published-availability)
|
| 165 |
+
tracks downloadable selections, including partition coverage.
|