datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
rocmrocm-vlm-ocr-awq-configs
ROCm VLM/OCR AWQ+BF16 Serving Config Bundle
A4 @ 200 DPI coverage verified (max_model_len >= 8192, visual tokens ~3800 + text ~4096).
Model Profiles
Profile
Model
Quant
max_model_len
Concurrent VRAM
vllm-qwen3-vl-awq.json
cyankiwi/Qwen3-VL-8B-Instruct-AWQ-4bit
AWQ W4A16
8192
0.30
vllm-bizonai-bf16.json
ONTHEIT/BizOnAI-OCR
BF16
8192
0.26
sglang-qwen3-vl-awq.json
cyankiwi/Qwen3-VL-8B-Instruct-AWQ-4bit
AWQ W4A16
8192
0.30
A4 @ 200 DPI Token… See the full description on the dataset page: https://huggingface.co/datasets/yakdoli/rocm-vlm-ocr-awq-configs.deepseek-v4-flash-rocm-vllm-repro
Reproducing DeepSeek-V4-Flash on AMD ROCm with vLLM: 32K Correctness and TopK Sweep
This article summarizes an engineering reproduction of
deepseek-ai/DeepSeek-V4-Flash on an AMD ROCm ModelScope DSW instance. The work
focuses on a practical question: can a complex, fast-moving DeepSeek-V4-Flash
serving path be turned into a reproducible ROCm baseline with explicit
correctness gates?
The answer from this run is yes, with an important boundary: the current setup
is a fallback-heavy… See the full description on the dataset page: https://huggingface.co/datasets/lyydfys/deepseek-v4-flash-rocm-vllm-repro.installamacpp-rocmwan2.2-rocm-profiles
Wan2.2 Sequence Parallel ROCm Profiles (MI300X)
This dataset contains PyTorch/Perfetto traces, offline execution logs, serving benchmarks, and comparative reports for Wan2.2-T2V-A14B sequence-parallel runs on AMD ROCm (gfx942, 8x MI300X node).
Dataset Directory Structure
reports/ / Root:
wan22_rocm_sp_sweep_analysis.md: 3-way sequence parallel topology comparison report.
wan22_profile_u4_r1_analysis.md: Detailed analysis of the Ulysses-4 topology.… See the full description on the dataset page: https://huggingface.co/datasets/Akshat/wan2.2-rocm-profiles.rocm-agent-sftrocm-build-env-a1roc_masked_middlekernel-grader-mi300x-rocm-trainium-nki
Grading a kernel on an MI300X, and two agents that skipped it entirely
I put 67,108,864 skewed keys into 4096 bins on an AMD Instinct MI300X VF, gfx942, under ROCm 7.2.4. One global atomic per key takes 86.3898 ms. Privatising the table into local data share per block, with one merge at the end, takes 0.1015 ms. That is 851.13 times, measured on the same card in the same run.
Two cheating agents then scored full marks against my own grader.
What ran
Measured time
Score… See the full description on the dataset page: https://huggingface.co/datasets/LaelaZorana/kernel-grader-mi300x-rocm-trainium-nki.OFFFELLIA_llama_ROCmFPX
ΩFFFΣLLIa • llama.cpp • OFFFELLIA_llama_ROCmFPX
██████╗ ███████╗███████╗███████╗██╗ ██╗ ██╗ █████╗
██╔═══██╗██╔════╝██╔════╝██╔════╝██║ ██║ ██║██╔══██╗
██║ ██║█████╗ █████╗ █████╗ ██║ ██║ ██║███████║
██║ ██║██╔══╝ ██╔══╝ ██╔══╝ ██║ ██║ ██║██╔══██║
╚██████╔╝██║ ██║ ███████╗███████╗███████╗██║██║ ██║OFFFELLIA_llama_ROCmFPX
╚═════╝ ╚═╝ ╚═╝ ╚══════╝╚══════╝╚══════╝╚═╝╚═╝ ╚═╝
High-Performance LLM /… See the full description on the dataset page: https://huggingface.co/datasets/Brunobkr/OFFFELLIA_llama_ROCmFPX.cuda-to-rocm-wavefront-bugs
CUDA → ROCm Wavefront Bug Dataset
170 expert-curated examples of GPU kernel bugs that survive mechanical hipify translation and only manifest on AMD MI300X hardware (gfx942, wavefront-64).
Built for the ROCmPort AI project — a multi-agent pipeline that ports and optimizes CUDA kernels for AMD GPUs.
Why This Dataset Exists
hipify-perl and hipify-clang do a great job of mechanical API renaming (CUDA → HIP). But they cannot detect semantic bugs caused by AMD's larger… See the full description on the dataset page: https://huggingface.co/datasets/tazwarrrr/cuda-to-rocm-wavefront-bugs.rocmpilot-agent-sft
ROCmPilot Agent SFT
This dataset contains seed supervised fine-tuning examples for ROCmPilot, a multi-agent tool that helps developers migrate PyTorch and vLLM workloads from CUDA/NVIDIA assumptions to AMD ROCm readiness.
The examples teach ROCmPilot's production-facing agent behaviors:
repo_doctor: scan repository evidence for CUDA/NVIDIA assumptions
migration_planner: identify CUDA/NVIDIA migration blockers and recommend ROCm-safe fixes
patch_planner: convert findings into scoped… See the full description on the dataset page: https://huggingface.co/datasets/Shivam311/rocmpilot-agent-sft.cuda-rocm-pairsrocm-agent-traj-v0rocm-agent-traj-v1