Popular repositories Loading

1. deepseek-v4-flash-rocm-vllm-repro deepseek-v4-flash-rocm-vllm-repro Public

   Reproducing DeepSeek-V4-Flash on AMD ROCm with vLLM: 32K correctness, 8K top-k sweep, and ModelScope DSW notebook baseline.

   Python

2. qwen36-27b-fp8-rocm-mtp-repro qwen36-27b-fp8-rocm-mtp-repro Public

   Python

3. qwen3.6-a3b-rtx4050-benchmark qwen3.6-a3b-rtx4050-benchmark Public

   Practical llama.cpp benchmark for Qwen3.6-35B-A3B GGUF on RTX 4050 Laptop 6GB

   PowerShell