| { | |
| "config_groups": { | |
| "group_0": { | |
| "targets": [ | |
| "Linear" | |
| ], | |
| "weights": { | |
| "num_bits": 4, | |
| "type": "float", | |
| "symmetric": true, | |
| "group_size": 16, | |
| "strategy": "tensor_group", | |
| "block_structure": null, | |
| "dynamic": false, | |
| "actorder": null, | |
| "observer": "minmax", | |
| "observer_kwargs": {} | |
| }, | |
| "input_activations": { | |
| "num_bits": 4, | |
| "type": "float", | |
| "symmetric": true, | |
| "group_size": 16, | |
| "strategy": "tensor_group", | |
| "block_structure": null, | |
| "dynamic": "local", | |
| "actorder": null, | |
| "observer": "minmax", | |
| "observer_kwargs": {} | |
| }, | |
| "output_activations": null | |
| } | |
| }, | |
| "quant_method": "compressed-tensors", | |
| "kv_cache_scheme": null, | |
| "format": "nvfp4-pack-quantized", | |
| "quantization_status": "compressed", | |
| "global_compression_ratio": null, | |
| "ignore": [ | |
| "lm_head" | |
| ] | |
| } |