| {"sae": {"expansion_factor": 32, "normalize_decoder": true, "num_latents": 131072, "k": 256, "multi_topk": false}, "batch_size": 1, "grad_acc_steps": 4, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["model.layers.24"], "layers": [24], "layer_stride": 1, "distribute_modules": false, "save_every": 1000, "log_to_wandb": true, "run_name": "llava-hf_llama3-llava-next-8b-hf-lmms-lab_LLaVA-NeXT-Data-sae", "wandb_log_frequency": 1, "mm_data": true, "model": "llava-hf/llama3-llava-next-8b-hf", "dataset": "lmms-lab/LLaVA-NeXT-Data", "split": "train", "ctx_len": 2048, "hf_token": null, "load_in_8bit": false, "max_examples": null, "resume": false, "seed": 42, "data_preprocessing_num_proc": 32} |