| { |
| "_external_rope_config_kwargs": {}, |
| "add_cross_attention": false, |
| "architectures": [ |
| "Gemma4ForConditionalGeneration" |
| ], |
| "attn_mechanism": "vanilla", |
| "audio_config": { |
| "_name_or_path": "", |
| "architectures": null, |
| "attention_chunk_size": 12, |
| "attention_context_left": 13, |
| "attention_context_right": 0, |
| "attention_invalid_logits_value": -1000000000.0, |
| "attention_logit_cap": 50.0, |
| "chunk_size_feed_forward": 0, |
| "conv_kernel_size": 5, |
| "dtype": "bfloat16", |
| "gradient_clipping": 10000000000.0, |
| "hidden_act": "silu", |
| "hidden_size": 1024, |
| "id2label": { |
| "0": "LABEL_0", |
| "1": "LABEL_1" |
| }, |
| "initializer_range": 0.02, |
| "is_encoder_decoder": false, |
| "label2id": { |
| "LABEL_0": 0, |
| "LABEL_1": 1 |
| }, |
| "model_type": "gemma4_audio", |
| "num_attention_heads": 8, |
| "num_hidden_layers": 12, |
| "output_attentions": false, |
| "output_hidden_states": false, |
| "output_proj_dims": 1536, |
| "problem_type": null, |
| "residual_weight": 0.5, |
| "return_dict": true, |
| "rms_norm_eps": 1e-06, |
| "subsampling_conv_channels": [ |
| 128, |
| 32 |
| ], |
| "use_clipped_linears": true |
| }, |
| "audio_token_id": 258881, |
| "backend": null, |
| "bits": null, |
| "blocksize_b": 1, |
| "blocksize_k": 512, |
| "blocksize_q": 512, |
| "boa_token_id": 256000, |
| "boi_token_id": 255999, |
| "bos_token_id": null, |
| "cross_attention_hidden_size": null, |
| "decode_attn_mechanism": null, |
| "decoder_start_token_id": null, |
| "dtype": "bfloat16", |
| "easy_method": "train", |
| "eoa_token_id": 258883, |
| "eoa_token_index": 258883, |
| "eoi_token_id": 258882, |
| "eos_token_id": null, |
| "fcm_max_ratio": 0.0, |
| "fcm_min_ratio": 0.0, |
| "flash_attention_backward_pass_impl": "triton", |
| "fsdp_is_ep_bound": true, |
| "gradient_checkpointing": "", |
| "gradient_checkpointing_targets": null, |
| "hardware_abstraction": false, |
| "image_token_id": 258880, |
| "initializer_range": 0.02, |
| "is_decoder": false, |
| "kv_cache_quantization_config": null, |
| "kv_cache_sharding_sequence_axis_name": "sp", |
| "kvdtype": "bfloat16", |
| "lmhead_chunksize": null, |
| "max_position_embeddings": null, |
| "mla_attn_dtype": "bfloat16", |
| "mla_attn_mechanism": "auto", |
| "mla_attn_softmax_dtype": "float32", |
| "model_type": "gemma4", |
| "moe_force_xla_gmm": false, |
| "moe_method": "fused_moe", |
| "moe_tiling_size_batch": 4, |
| "moe_tiling_size_dim": 128, |
| "moe_tiling_size_seqlen": 128, |
| "operation_configs": null, |
| "pad_token_id": null, |
| "pallas_k_block_size": 128, |
| "pallas_m_block_size": 128, |
| "pallas_n_block_size": 128, |
| "partition_axis": { |
| "attention_dim_axis": null, |
| "attention_kv_dim_axis": null, |
| "batch_axis": [ |
| "fsdp", |
| "dp" |
| ], |
| "bias_head_sequence_axis": null, |
| "bias_key_sequence_axis": null, |
| "data_parallel_axis": "dp", |
| "decode_attention_dim_axis": null, |
| "decode_attention_kv_dim_axis": null, |
| "decode_batch_axis": [ |
| "fsdp", |
| "dp" |
| ], |
| "decode_head_axis": "tp", |
| "decode_key_sequence_axis": "sp", |
| "decode_kv_head_axis": "tp", |
| "decode_query_sequence_axis": null, |
| "expert_axis": "ep", |
| "expert_gate_axis": null, |
| "expert_parallel_axis": "ep", |
| "fully_sharded_data_parallel_axis": "fsdp", |
| "head_axis": "tp", |
| "hidden_state_axis": "tp", |
| "key_sequence_axis": "sp", |
| "kv_head_axis": "tp", |
| "mlp_intermediate_axis": "tp", |
| "query_sequence_axis": "sp", |
| "sequence_axis": "sp", |
| "sequence_parallel_axis": "sp", |
| "tensor_parallel_axis": "tp", |
| "vocab_axis": "tp" |
| }, |
| "platform": null, |
| "precompute_masks": true, |
| "pretraining_tp": 1, |
| "qmm_platform_override": null, |
| "qmm_tpu_path_override": null, |
| "quantization_config": null, |
| "scan_attention_layers": false, |
| "scan_mlp_chunk_size": 1024, |
| "scan_ring_attention": true, |
| "sep_token_id": null, |
| "sequence_axis_name": "sp", |
| "sharding_axis_dims": [ |
| 1, |
| -1, |
| 1, |
| 1, |
| 1 |
| ], |
| "sharding_axis_names": [ |
| "dp", |
| "fsdp", |
| "ep", |
| "tp", |
| "sp" |
| ], |
| "sharding_dcn_axis_dims": null, |
| "sp_is_ep_bound": true, |
| "text_config": { |
| "_external_rope_config_kwargs": {}, |
| "add_cross_attention": false, |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "attention_k_eq_v": false, |
| "attn_dtype": "bfloat16", |
| "attn_mechanism": "vanilla", |
| "attn_softmax_dtype": "float32", |
| "backend": null, |
| "bits": null, |
| "blocksize_b": 1, |
| "blocksize_k": 512, |
| "blocksize_q": 512, |
| "bos_token_id": 2, |
| "cross_attention_hidden_size": null, |
| "decode_attn_mechanism": null, |
| "decoder_start_token_id": null, |
| "dtype": "bfloat16", |
| "easy_method": "train", |
| "enable_moe_block": false, |
| "eos_token_id": 1, |
| "expert_intermediate_size": null, |
| "fcm_max_ratio": 0.0, |
| "fcm_min_ratio": 0.0, |
| "final_logit_softcapping": 30.0, |
| "flash_attention_backward_pass_impl": "triton", |
| "fsdp_is_ep_bound": true, |
| "global_head_dim": 512, |
| "gradient_checkpointing": "", |
| "gradient_checkpointing_targets": null, |
| "hardware_abstraction": false, |
| "head_dim": 256, |
| "hidden_activation": "gelu_pytorch_tanh", |
| "hidden_size": 1536, |
| "hidden_size_per_layer_input": 256, |
| "initializer_range": 0.02, |
| "intermediate_size": 6144, |
| "is_decoder": false, |
| "kv_cache_quantization_config": null, |
| "kv_cache_sharding_sequence_axis_name": "sp", |
| "kvdtype": "bfloat16", |
| "layer_types": [ |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention" |
| ], |
| "lmhead_chunksize": null, |
| "max_position_embeddings": 131072, |
| "mla_attn_dtype": "bfloat16", |
| "mla_attn_mechanism": "auto", |
| "mla_attn_softmax_dtype": "float32", |
| "model_type": "gemma4_text", |
| "moe_force_xla_gmm": false, |
| "moe_intermediate_size": null, |
| "moe_method": "fused_moe", |
| "moe_tiling_size_batch": 4, |
| "moe_tiling_size_dim": 128, |
| "moe_tiling_size_seqlen": 128, |
| "num_attention_heads": 8, |
| "num_experts": null, |
| "num_global_key_value_heads": null, |
| "num_hidden_layers": 35, |
| "num_key_value_heads": 1, |
| "num_kv_shared_layers": 20, |
| "num_local_experts": null, |
| "operation_configs": null, |
| "pad_token_id": 0, |
| "pallas_k_block_size": 128, |
| "pallas_m_block_size": 128, |
| "pallas_n_block_size": 128, |
| "partition_axis": { |
| "attention_dim_axis": null, |
| "attention_kv_dim_axis": null, |
| "batch_axis": [ |
| "fsdp", |
| "dp" |
| ], |
| "bias_head_sequence_axis": null, |
| "bias_key_sequence_axis": null, |
| "data_parallel_axis": "dp", |
| "decode_attention_dim_axis": null, |
| "decode_attention_kv_dim_axis": null, |
| "decode_batch_axis": [ |
| "fsdp", |
| "dp" |
| ], |
| "decode_head_axis": "tp", |
| "decode_key_sequence_axis": "sp", |
| "decode_kv_head_axis": "tp", |
| "decode_query_sequence_axis": null, |
| "expert_axis": "ep", |
| "expert_gate_axis": null, |
| "expert_parallel_axis": "ep", |
| "fully_sharded_data_parallel_axis": "fsdp", |
| "head_axis": "tp", |
| "hidden_state_axis": "tp", |
| "key_sequence_axis": "sp", |
| "kv_head_axis": "tp", |
| "mlp_intermediate_axis": "tp", |
| "query_sequence_axis": "sp", |
| "sequence_axis": "sp", |
| "sequence_parallel_axis": "sp", |
| "tensor_parallel_axis": "tp", |
| "vocab_axis": "tp" |
| }, |
| "platform": null, |
| "precompute_masks": true, |
| "pretraining_tp": 1, |
| "qmm_platform_override": null, |
| "qmm_tpu_path_override": null, |
| "quantization_config": null, |
| "rms_norm_eps": 1e-06, |
| "rope_parameters": { |
| "full_attention": { |
| "partial_rotary_factor": 0.25, |
| "rope_theta": 1000000.0, |
| "rope_type": "proportional", |
| "type": "proportional" |
| }, |
| "sliding_attention": { |
| "rope_theta": 10000.0, |
| "rope_type": "default", |
| "type": "default" |
| } |
| }, |
| "scan_attention_layers": false, |
| "scan_layers": false, |
| "scan_mlp_chunk_size": 1024, |
| "scan_ring_attention": true, |
| "sep_token_id": null, |
| "sequence_axis_name": "sp", |
| "sharding_axis_dims": [ |
| 1, |
| -1, |
| 1, |
| 1, |
| 1 |
| ], |
| "sharding_axis_names": [ |
| "dp", |
| "fsdp", |
| "ep", |
| "tp", |
| "sp" |
| ], |
| "sharding_dcn_axis_dims": null, |
| "sliding_window": 512, |
| "sp_is_ep_bound": true, |
| "tie_encoder_decoder": false, |
| "tie_word_embeddings": true, |
| "top_k_experts": null, |
| "use_bidirectional_attention": null, |
| "use_cache": true, |
| "use_double_wide_mlp": true, |
| "use_expert_tensor_mode": false, |
| "use_qmm_best_config": false, |
| "use_ring_of_experts": false, |
| "use_scan_mlp": false, |
| "use_sharded_kv_caching": false, |
| "use_sharding_constraint": false, |
| "vocab_size": 262144, |
| "vocab_size_per_layer_input": 262144 |
| }, |
| "tie_encoder_decoder": false, |
| "tie_word_embeddings": true, |
| "transformers_version": "5.5.0", |
| "use_expert_tensor_mode": false, |
| "use_qmm_best_config": false, |
| "use_ring_of_experts": false, |
| "use_scan_mlp": false, |
| "use_sharded_kv_caching": false, |
| "use_sharding_constraint": false, |
| "video_token_id": 258884, |
| "vision_config": { |
| "_external_rope_config_kwargs": {}, |
| "add_cross_attention": false, |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "attn_dtype": "bfloat16", |
| "attn_mechanism": "vanilla", |
| "attn_softmax_dtype": "float32", |
| "backend": null, |
| "bits": null, |
| "blocksize_b": 1, |
| "blocksize_k": 512, |
| "blocksize_q": 512, |
| "bos_token_id": null, |
| "cross_attention_hidden_size": null, |
| "decode_attn_mechanism": null, |
| "decoder_start_token_id": null, |
| "default_output_length": 280, |
| "dtype": "bfloat16", |
| "easy_method": "train", |
| "eos_token_id": null, |
| "fcm_max_ratio": 0.0, |
| "fcm_min_ratio": 0.0, |
| "flash_attention_backward_pass_impl": "triton", |
| "fsdp_is_ep_bound": true, |
| "global_head_dim": 64, |
| "gradient_checkpointing": "", |
| "gradient_checkpointing_targets": null, |
| "hardware_abstraction": false, |
| "head_dim": 64, |
| "hidden_activation": "gelu_pytorch_tanh", |
| "hidden_size": 768, |
| "initializer_range": 0.02, |
| "intermediate_size": 3072, |
| "is_decoder": false, |
| "kv_cache_quantization_config": null, |
| "kv_cache_sharding_sequence_axis_name": "sp", |
| "kvdtype": "bfloat16", |
| "lmhead_chunksize": null, |
| "max_position_embeddings": 131072, |
| "mla_attn_dtype": "bfloat16", |
| "mla_attn_mechanism": "auto", |
| "mla_attn_softmax_dtype": "float32", |
| "model_type": "gemma4_vision", |
| "moe_force_xla_gmm": false, |
| "moe_method": "fused_moe", |
| "moe_tiling_size_batch": 4, |
| "moe_tiling_size_dim": 128, |
| "moe_tiling_size_seqlen": 128, |
| "num_attention_heads": 12, |
| "num_hidden_layers": 16, |
| "num_key_value_heads": 12, |
| "operation_configs": null, |
| "pad_token_id": null, |
| "pallas_k_block_size": 128, |
| "pallas_m_block_size": 128, |
| "pallas_n_block_size": 128, |
| "partition_axis": { |
| "attention_dim_axis": null, |
| "attention_kv_dim_axis": null, |
| "batch_axis": [ |
| "fsdp", |
| "dp" |
| ], |
| "bias_head_sequence_axis": null, |
| "bias_key_sequence_axis": null, |
| "data_parallel_axis": "dp", |
| "decode_attention_dim_axis": null, |
| "decode_attention_kv_dim_axis": null, |
| "decode_batch_axis": [ |
| "fsdp", |
| "dp" |
| ], |
| "decode_head_axis": "tp", |
| "decode_key_sequence_axis": "sp", |
| "decode_kv_head_axis": "tp", |
| "decode_query_sequence_axis": null, |
| "expert_axis": "ep", |
| "expert_gate_axis": null, |
| "expert_parallel_axis": "ep", |
| "fully_sharded_data_parallel_axis": "fsdp", |
| "head_axis": "tp", |
| "hidden_state_axis": "tp", |
| "key_sequence_axis": "sp", |
| "kv_head_axis": "tp", |
| "mlp_intermediate_axis": "tp", |
| "query_sequence_axis": "sp", |
| "sequence_axis": "sp", |
| "sequence_parallel_axis": "sp", |
| "tensor_parallel_axis": "tp", |
| "vocab_axis": "tp" |
| }, |
| "patch_size": 16, |
| "platform": null, |
| "pooling_kernel_size": 3, |
| "position_embedding_size": 10240, |
| "precompute_masks": true, |
| "pretraining_tp": 1, |
| "qmm_platform_override": null, |
| "qmm_tpu_path_override": null, |
| "quantization_config": null, |
| "rms_norm_eps": 1e-06, |
| "rope_parameters": { |
| "rope_theta": 100.0, |
| "rope_type": "default", |
| "type": "default" |
| }, |
| "scan_attention_layers": false, |
| "scan_mlp_chunk_size": 1024, |
| "scan_ring_attention": true, |
| "sep_token_id": null, |
| "sequence_axis_name": "sp", |
| "sharding_axis_dims": [ |
| 1, |
| -1, |
| 1, |
| 1, |
| 1 |
| ], |
| "sharding_axis_names": [ |
| "dp", |
| "fsdp", |
| "ep", |
| "tp", |
| "sp" |
| ], |
| "sharding_dcn_axis_dims": null, |
| "sp_is_ep_bound": true, |
| "standardize": false, |
| "tie_encoder_decoder": false, |
| "tie_word_embeddings": true, |
| "use_clipped_linears": true, |
| "use_expert_tensor_mode": false, |
| "use_qmm_best_config": false, |
| "use_ring_of_experts": false, |
| "use_scan_mlp": false, |
| "use_sharded_kv_caching": false, |
| "use_sharding_constraint": false |
| }, |
| "vision_soft_tokens_per_image": 280 |
| } |
|
|
