{ "bits": 4, "group_size": 128, "sym": true, "data_type": "int", "enable_quanted_input": true, "enable_minmax_tuning": true, "seqlen": 512, "batch_size": 4, "scale_dtype": "torch.float16", "lr": 0.005, "minmax_lr": 0.005, "gradient_accumulate_steps": 1, "iters": 200, "amp": false, "nsamples": 128, "low_gpu_mem_usage": false, "to_quant_block_names": null, "enable_norm_bias_tuning": false, "dataset": "NeelNanda/pile-10k", "autoround_version": "0.4.2", "quant_method": "intel/auto-round", "backend": "auto_round:gptq:exllamav2" }