feat: update skills.

feat: add unidrive_vla_trt_edge_llm fp16 test, but failed

fix: update tencent dev machine's ip

feature: claude skills

fix: onnx export, consolidate the model onto one device.

fix: unidrive vla quantization, oom failure during calib. add dynamic loading of calibration data, add max_pixel to limit mass loading.

feat: run.sh

feat: unstrict evaluation

feat: sync for offline environment.

update: add UniDriveVLA deps into prepare_env.sh

fix: load whole model on single gpu to avoid quant error.

fix: remove ONNX_NVFP4 model for UniDriveVLA_Nusc_Base_Stage1 model.

fix: GPT-score can't work in China.

add: unidrive_vla_nusc_base_evaluation

fix: finetune output folder.

feat: add qat functions.

update: log output.

feat: add qat for hgnetv2

fix: exclude QKV, attention, MatMuls, LayerNorm, MatMuls layers for int8 quant.

add: mode selector for eval_quantized script.

add: vit_large_patch16_224 quant & eval feature

feat: add prepare cloud env script. optimize script for limit network env.

feat: finished with regnet_x_800mf PTQ scripts.

feat: finished with hgnetv2 PTQ scripts.

feat: save engine and calibration function, model quantize function.

changed the evaluation datasets.

fix: gradio lite index.html

add: model infer, model eval, within HuggingFace APIs, hello

initial commit

Duplicate from gradio/gradio-lite-template