---
license: cc-by-nc-4.0
pretty_name: CounterStrike-1K (720p shards)
task_categories:
- video-classification
- reinforcement-learning
- video-to-video
tags:
- counter-strike-2
- world-model
- video-prediction
- action-conditioned-video
- multi-view
- webdataset
- audio
- esports
size_categories:
- 10K<n<100K
configs:
- config_name: default
  data_files:
  - split: train
    path: shards-720p/**/*.tar
---

# CounterStrike-1K — 720p WebDataset shards

This repo contains the **720p** shards of [CounterStrike-1K](https://huggingface.co/datasets/ArnieRamesh/CounterStrike-1K). Use the [main repo](https://huggingface.co/datasets/ArnieRamesh/CounterStrike-1K) to browse the manifest, schema, and subsets.

## Quickstart

Start a fresh `uv` project and add the loader:

```bash
mkdir cs1k-demo && cd cs1k-demo
uv init
uv add datasets "counterstrike1k @ git+https://github.com/AnirudhhRamesh/counterstrike1k"
```

<details>
<summary>Using pip instead</summary>

```bash
mkdir cs1k-demo && cd cs1k-demo
python -m venv .venv && source .venv/bin/activate
pip install datasets "counterstrike1k @ git+https://github.com/AnirudhhRamesh/counterstrike1k"
```
</details>

```python
from datasets import Video, load_dataset
from counterstrike1k import decode_sample

shards = load_dataset(
    "ArnieRamesh/CounterStrike-1K-720-wds", split="train", streaming=True,
).cast_column("mp4", Video(decode=False))
sample = decode_sample(next(iter(shards)))

print(sample["actions"].shape, sample["state"].shape, len(sample["video"]))
```

`decode_sample(...)` returns:

- `video`: mp4 bytes (H.264 + AAC, 1280×720 @ 32 FPS, synchronized stereo audio)
- `actions`: structured numpy array (per-frame `tick`, `delta_pitch`, `delta_yaw`, 12-button bitmask)
- `state`: structured numpy array (per-frame view, position, weapon, ammo, HP, money, score, …)
- `events`: list of sparse round/kill/bomb events
- `metadata`: public sample metadata sidecar

## Size

- 396 shards, ~1.5 TB total
- One round = 10 synchronized POV samples sharing a `round_id`
- Splits (in `manifest.parquet` from the main repo): 1,341.7 / 74.5 / 74.5 POV-hours train/val/test

## Filtering by subset

Most users want a smaller training run. Filter the manifest first, then stream only the matching shards:

```python
from datasets import Video, load_dataset
import pandas as pd
from huggingface_hub import hf_hub_download

manifest = pd.read_parquet(hf_hub_download(
    "ArnieRamesh/CounterStrike-1K", "manifest.parquet", repo_type="dataset",
))
keys = set(manifest[manifest["split"] == "train"]["sample_key"])

shards = load_dataset(
    "ArnieRamesh/CounterStrike-1K-720-wds", split="train", streaming=True,
).cast_column("mp4", Video(decode=False))
for raw in shards:
    if raw["__key__"] in keys:
        sample = decode_sample(raw)
        # ... use sample
```

## License & citation

CC BY-NC 4.0. Citation in the [main dataset card](https://huggingface.co/datasets/ArnieRamesh/CounterStrike-1K).