Updated hf_dataset_card.md

origin_hardware_baselines/resident_evil_4/hf_dataset_card.md

@@ -1,17 +1,3 @@
----
-license: gpl-3.0
-task_categories:
-- time-series-forecasting
-tags:
-- neuromorphic
-- snn
-- mixture-of-experts
-- gaming
-- hardware-telemetry
-- gpu
-pretty_name: Metis SMoE Latent Telemetry (Gaming)
----
-
 # Metis SMoE Latent Telemetry
 ## Neuromorphic Hardware Telemetry from demanding Gaming Workloads
 
@@ -19,8 +5,7 @@ This dataset provides high-fidelity, high-frequency (5ms interval) hardware tele
 
 ### Context
 The telemetry data was recorded using a custom Rust-based data collector via the NVIDIA Management Library (NVML) on a Fedora 43 Linux system. Workloads represent highly transient rendering applications including:
-- **Resident Evil 4 (Remake)** (with rendering complexities)
-- **Cyberpunk 2077** (Path Tracing, DLSS 4.0)
+- **Resident Evil 4 (Remake)** (Path Tracing, DLSS 4.0) at around 11.3 GB of Vram usage
 
 This system provides the rich, high-frequency time-series data required to train **Spiking Neural Networks (SNNs)** and **Liquid State Machines (LSMs)**.