| { |
| "architectures": [ |
| "GraniteMoeHybridForCausalLM" |
| ], |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "attention_multiplier": 0.0078125, |
| "bos_token_id": 100257, |
| "dtype": "bfloat16", |
| "embedding_multiplier": 12, |
| "eos_token_id": 100257, |
| "hidden_act": "silu", |
| "hidden_size": 4096, |
| "initializer_range": 0.1, |
| "intermediate_size": 768, |
| "layer_types": [ |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "attention", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "attention", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "attention", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba", |
| "attention", |
| "mamba", |
| "mamba", |
| "mamba", |
| "mamba" |
| ], |
| "logits_scaling": 16, |
| "mamba_chunk_size": 256, |
| "mamba_conv_bias": true, |
| "mamba_d_conv": 4, |
| "mamba_d_head": 64, |
| "mamba_d_state": 128, |
| "mamba_expand": 2, |
| "mamba_n_groups": 1, |
| "mamba_n_heads": 128, |
| "mamba_proj_bias": false, |
| "max_position_embeddings": 131072, |
| "model_type": "granitemoehybrid", |
| "normalization_function": "rmsnorm", |
| "num_attention_heads": 32, |
| "num_experts_per_tok": 10, |
| "num_hidden_layers": 40, |
| "num_key_value_heads": 8, |
| "num_local_experts": 72, |
| "output_router_logits": false, |
| "pad_token_id": 100256, |
| "position_embedding_type": "nope", |
| "quantization_config": { |
| "algo_config": null, |
| "exclude": [ |
| "model.layers.0.block_sparse_moe.router.layer", |
| "model.layers.1.block_sparse_moe.router.layer", |
| "model.layers.2.block_sparse_moe.router.layer", |
| "model.layers.3.block_sparse_moe.router.layer", |
| "model.layers.4.block_sparse_moe.router.layer", |
| "model.layers.5.block_sparse_moe.router.layer", |
| "model.layers.6.block_sparse_moe.router.layer", |
| "model.layers.7.block_sparse_moe.router.layer", |
| "model.layers.8.block_sparse_moe.router.layer", |
| "model.layers.9.block_sparse_moe.router.layer", |
| "model.layers.10.block_sparse_moe.router.layer", |
| "model.layers.11.block_sparse_moe.router.layer", |
| "model.layers.12.block_sparse_moe.router.layer", |
| "model.layers.13.block_sparse_moe.router.layer", |
| "model.layers.14.block_sparse_moe.router.layer", |
| "model.layers.15.block_sparse_moe.router.layer", |
| "model.layers.16.block_sparse_moe.router.layer", |
| "model.layers.17.block_sparse_moe.router.layer", |
| "model.layers.18.block_sparse_moe.router.layer", |
| "model.layers.19.block_sparse_moe.router.layer", |
| "model.layers.20.block_sparse_moe.router.layer", |
| "model.layers.21.block_sparse_moe.router.layer", |
| "model.layers.22.block_sparse_moe.router.layer", |
| "model.layers.23.block_sparse_moe.router.layer", |
| "model.layers.24.block_sparse_moe.router.layer", |
| "model.layers.25.block_sparse_moe.router.layer", |
| "model.layers.26.block_sparse_moe.router.layer", |
| "model.layers.27.block_sparse_moe.router.layer", |
| "model.layers.28.block_sparse_moe.router.layer", |
| "model.layers.29.block_sparse_moe.router.layer", |
| "model.layers.30.block_sparse_moe.router.layer", |
| "model.layers.31.block_sparse_moe.router.layer", |
| "model.layers.32.block_sparse_moe.router.layer", |
| "model.layers.33.block_sparse_moe.router.layer", |
| "model.layers.34.block_sparse_moe.router.layer", |
| "model.layers.35.block_sparse_moe.router.layer", |
| "model.layers.36.block_sparse_moe.router.layer", |
| "model.layers.37.block_sparse_moe.router.layer", |
| "model.layers.38.block_sparse_moe.router.layer", |
| "model.layers.39.block_sparse_moe.router.layer", |
| "lm_head" |
| ], |
| "export": { |
| "kv_cache_group": [ |
| "*k_proj", |
| "*v_proj" |
| ], |
| "min_kv_scale": 0.0, |
| "pack_method": "reorder", |
| "weight_format": "real_quantized", |
| "weight_merge_groups": null |
| }, |
| "global_quant_config": { |
| "bias": null, |
| "input_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| }, |
| "output_tensors": null, |
| "target_device": null, |
| "weight": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| } |
| }, |
| "kv_cache_quant_config": { |
| "*k_proj": { |
| "bias": null, |
| "input_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| }, |
| "output_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": null, |
| "symmetric": true |
| }, |
| "target_device": null, |
| "weight": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| } |
| }, |
| "*v_proj": { |
| "bias": null, |
| "input_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| }, |
| "output_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": null, |
| "symmetric": true |
| }, |
| "target_device": null, |
| "weight": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| } |
| } |
| }, |
| "layer_quant_config": { |
| "*k_proj": { |
| "bias": null, |
| "input_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| }, |
| "output_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": null, |
| "symmetric": true |
| }, |
| "target_device": null, |
| "weight": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| } |
| }, |
| "*v_proj": { |
| "bias": null, |
| "input_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| }, |
| "output_tensors": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": null, |
| "symmetric": true |
| }, |
| "target_device": null, |
| "weight": { |
| "ch_axis": null, |
| "dtype": "fp8_e4m3", |
| "group_size": null, |
| "is_dynamic": false, |
| "is_scale_quant": false, |
| "mx_element_dtype": null, |
| "observer_cls": "PerTensorMinMaxObserver", |
| "qscheme": "per_tensor", |
| "round_method": "half_even", |
| "scale_calculation_mode": null, |
| "scale_format": null, |
| "scale_type": "float", |
| "symmetric": true |
| } |
| } |
| }, |
| "layer_type_quant_config": {}, |
| "quant_method": "quark", |
| "quant_mode": "eager_mode", |
| "softmax_quant_spec": null, |
| "version": "0.11+173148031e" |
| }, |
| "residual_multiplier": 0.22, |
| "rms_norm_eps": 1e-05, |
| "rope_scaling": null, |
| "rope_theta": 10000, |
| "router_aux_loss_coef": 0.0, |
| "shared_intermediate_size": 1536, |
| "tie_word_embeddings": true, |
| "transformers_version": "4.56.2", |
| "use_cache": true, |
| "vocab_size": 100352 |
| } |
|
|
