| { |
| "allowed_zero_lora_B": [], |
| "alpha": 256, |
| "base_model": "/high_perf_store3/world-model/zhuyu/unidrivevla/CKPT/UniDriveVLA_Nusc_Large_Stage1", |
| "checkpoint": "/high_perf_store3/world-model/zhuyu/unidrivevla-main_coc/nuScenes/work_dirs2/UniDriveVLA/unidrivevla_stage2_8b_nusc_large_stage1_coc_weight01_visual_wasp_beta0005_r128a256_top16_projector_lora_r128a256_llm_wasp_beta0003_r128a256_top16_2node16gpu_bs2_accum4_zero2_vgc/iter_17580/global_step4395/mp_rank_00_model_states.pt", |
| "checkpoint_state": "module (raw; not EMA)", |
| "constructed_adapter_counts": { |
| "language": 252, |
| "projector": 8, |
| "vision": 108 |
| }, |
| "formula_validation": { |
| "language": { |
| "adapter": "wasp", |
| "atol": 2e-05, |
| "bf16_quantization_close": true, |
| "exact_equal": true, |
| "max_abs_error": 0.0, |
| "max_input_aware_tolerance": null, |
| "max_rel_error": 0.0, |
| "module_before_merge": "model.language_model.layers.0.self_attn.q_proj", |
| "rtol": 0.015625, |
| "weight_after_merge": "model.language_model.layers.0.self_attn.q_proj.weight" |
| }, |
| "projector": { |
| "adapter": "lora", |
| "atol": 2e-05, |
| "bf16_quantization_close": true, |
| "exact_equal": false, |
| "max_abs_error": 0.00048828125, |
| "max_input_aware_tolerance": 0.0016679492546245456, |
| "max_rel_error": 1.0, |
| "module_before_merge": "model.visual.base_model.model.merger.linear_fc1", |
| "rtol": 0.015625, |
| "weight_after_merge": "model.visual.merger.linear_fc1.weight" |
| }, |
| "vision": { |
| "adapter": "wasp", |
| "atol": 2e-05, |
| "bf16_quantization_close": true, |
| "exact_equal": true, |
| "max_abs_error": 0.0, |
| "max_input_aware_tolerance": null, |
| "max_rel_error": 0.0, |
| "module_before_merge": "model.visual.base_model.model.blocks.0.attn.qkv", |
| "rtol": 0.015625, |
| "weight_after_merge": "model.visual.blocks.0.attn.qkv.weight" |
| } |
| }, |
| "language_top_k": 16, |
| "merged_wasp_counts": { |
| "language_wasp": 252, |
| "vision_tree_wasp": 108 |
| }, |
| "output_dir": "/high_perf_store3/world-model/zhuyu/UniDriveVLA/vqa_evaluation/LingoQA/runs/topk_four_serial_dynamic_conversion_20260814_0147/large_v16_l16_hf_merged_raw", |
| "preset": "visual_wasp_projector_lora_llm_wasp_8b", |
| "preset_config": { |
| "expected_language": 252, |
| "expected_lora": 8, |
| "expected_projector": 8, |
| "expected_vision": 108, |
| "expected_wasp": 360, |
| "language": "wasp", |
| "projector": "lora", |
| "vision": "wasp" |
| }, |
| "projector_lora_alpha": 256, |
| "projector_lora_r": 128, |
| "rank": 128, |
| "raw_adapter_validation": { |
| "allowed_zero_lora_B": [], |
| "layout": { |
| "U_top": 360, |
| "V_top": 360, |
| "lora_A_total": 368, |
| "lora_B_total": 368, |
| "peft_lora_A": 8, |
| "peft_lora_B": 8, |
| "wasp_lora_A": 360, |
| "wasp_lora_B": 360 |
| }, |
| "zero_lora_B": [] |
| }, |
| "saved_hf_validation": { |
| "adapter_residual_count": 0, |
| "mismatched_keys": [], |
| "missing_keys": [], |
| "safetensors": [ |
| "/high_perf_store3/world-model/zhuyu/UniDriveVLA/vqa_evaluation/LingoQA/runs/topk_four_serial_dynamic_conversion_20260814_0147/large_v16_l16_hf_merged_raw/model-00001-of-00004.safetensors", |
| "/high_perf_store3/world-model/zhuyu/UniDriveVLA/vqa_evaluation/LingoQA/runs/topk_four_serial_dynamic_conversion_20260814_0147/large_v16_l16_hf_merged_raw/model-00002-of-00004.safetensors", |
| "/high_perf_store3/world-model/zhuyu/UniDriveVLA/vqa_evaluation/LingoQA/runs/topk_four_serial_dynamic_conversion_20260814_0147/large_v16_l16_hf_merged_raw/model-00003-of-00004.safetensors", |
| "/high_perf_store3/world-model/zhuyu/UniDriveVLA/vqa_evaluation/LingoQA/runs/topk_four_serial_dynamic_conversion_20260814_0147/large_v16_l16_hf_merged_raw/model-00004-of-00004.safetensors" |
| ], |
| "tensor_count": 750, |
| "unexpected_keys": [], |
| "vocab_size": 151675 |
| }, |
| "top_k": 16, |
| "vision_top_k": 16, |
| "wasp_alpha": 256, |
| "wasp_rank": 128 |
| } |
|
|
