{ "model": "Dexy2/Piko-9b", "label": "piko-9b", "environment": { "timestamp": "2026-07-29T14:39:26-0400", "python": "3.12.3", "platform": "Linux-6.6.87.2-microsoft-standard-WSL2-x86_64-with-glibc2.39", "torch": "2.10.0+cu128", "transformers": "5.5.0", "gpu": "NVIDIA GeForce RTX 5070 Ti", "vram_total_gb": 15.92, "dtype": "bfloat16", "quantization": "4bit" }, "load": { "cold_load_seconds": 101.4, "weights_vram_gb": 7.342, "peak_during_load_gb": 7.371, "host_rss_before_gb": 0.81, "host_rss_after_gb": 1.283, "reported_parameters": 5724972272, "reported_vision_parameters": 230421232, "note": "quantized parameters report packed element counts, not logical parameters" }, "contexts": [ { "context_tokens": 512, "prefill_seconds": 0.805, "baseline_vram_gb": 7.342, "peak_vram_gb": 7.631, "activation_and_cache_gb": 0.289 }, { "context_tokens": 2048, "prefill_seconds": 0.402, "baseline_vram_gb": 7.35, "peak_vram_gb": 8.4, "activation_and_cache_gb": 1.05 }, { "context_tokens": 8192, "prefill_seconds": 1.59, "baseline_vram_gb": 7.35, "peak_vram_gb": 11.476, "activation_and_cache_gb": 4.126 } ], "failures": [ { "context_tokens": 32768, "error": "RuntimeError: CUDA driver error: device not ready" } ], "max_context_measured": 8192, "model_source_note": "run from a local copy of this checkpoint on internal NVMe; the path has been normalised to the canonical repository id" }