metrollm / model.py

Commit History

sync: HF Space ↔ /simulate prompt-construction parity
bb7b69a

Remco Hendriks Claude Opus 4.7 (1M context) commited on

probe runtime attn_implementation post-PEFT merge
8003669

Remco Hendriks Claude Opus 4.7 (1M context) commited on

flash-attn for ZeroGPU + regional scenario flavor + subtitle
32606be

Remco Hendriks Claude Opus 4.7 (1M context) commited on

deploy: 9B + ZeroGPU (low_cpu_mem_usage for module-load)
b08e50f

Remco Hendriks commited on

init: MetroLLM-Bench kiosk demo (2B + LoRA, free CPU)
cbfaae5

Remco Hendriks commited on