DeepSeek V4 Flash 284B on AMD Strix Halo (gfx1151) — up to 32 tok/s decode & ~250 tok/s prefill via ROCmFPX, DSpark & ROCm 7.2
-
Updated
Jul 24, 2026 - C
#
dspark
Here are 5 public repositories matching this topic...
DeepSeek-V4-Flash + DSpark speculative decoding on a pair of NVIDIA DGX Sparks (vLLM TP=2 over RoCE) — tuned recipe, overlays that halve multi-turn TTFT, contamination-guarded benchmarks, ops runbook
-
Updated
Jul 3, 2026 - Python
Lossless inference speedup benchmarking suite for local LLMs using DeepSeek DSpark speculative draft heads.
pytorch streamlit inference-acceleration ollama speculative-decoding deepseek qwen3 dspark deepspec llm-speedup
-
Updated
Jun 29, 2026 - Python
Multi-Ensemble Memory-Elastic Token Prediction — a numpy implementation of DeepSeek's DSpark speculative decoding + the MEMTP elastic-ensemble extension, on AMD Strix Halo (gfx1151).
amd numpy mtp rocm kv-cache llm-inference speculative-decoding deepseek strix-halo multi-token-prediction gfx1151 dspark
-
Updated
Jun 29, 2026 - Python
MiniMax M3 NVFP4 on 4x DGX Spark with NVIDIA DSpark, native multi-node vLLM TP=4, reasoning, and tool calling
-
Updated
Jul 25, 2026 - Shell
Improve this page
Add a description, image, and links to the dspark topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the dspark topic, visit your repo's landing page and select "manage topics."