namhokaist/appgen-qwen25-thinktag-lr5e-7-ep1-continue-sft-amex-mobile-use
09
namhokaist/appgen-qwen25-thinktag-lr5e-7-ep1-continue-sft-amex-mobile-use
Continue-SFT of luca0621/appgen-qwen25-thinktag-lr5e-7-ep1 (Qwen2.5-VL) on AMEX-derived GUI trajectories converted to the `mobile_use` tool-call format (absolute pixel coordinate, single mobile_use function).
- Base model:
luca0621/appgen-qwen25-thinktag-lr5e-7-ep1 - Format:
mobile_usetool call —<tool_call>{"name": "mobile_use", "arguments": {"action": "click", "coordinate": [x, y]}}</tool_call> - Training: 1 epoch, lr 5e-7, per-device bs 1 x grad-accum 8, max_len 4096
- W&B run:
qwen25-mobile-use-continue-sft-gpu01
