optimum-intel-internal-testing/tiny-random-phi-4-multimodal
014k
1{2 "_name_or_path": "/home/ea/work/my_optimum_intel/optimum-intel/tiny-random-phi-4-multimodal",3 "architectures": [4 "Phi4MMForCausalLM"5 ],6 "attention_bias": false,7 "attention_dropout": 0.0,8 "audio_processor": {9 "config": {10 "activation": "swish",11 "activation_checkpointing": {12 "interval": 1,13 "module": "transformer",14 "offload": false15 },16 "attention_dim": 16,17 "attention_heads": 2,18 "batch_norm": false,19 "bias_in_glu": true,20 "causal": true,21 "chunk_size": -1,22 "cnn_layer_norm": true,23 "conv_activation": "swish",24 "conv_glu_type": "swish",25 "depthwise_multiplier": 1,26 "depthwise_seperable_out_channel": 16,27 "dropout_rate": 0.0,28 "encoder_embedding_config": {29 "input_size": 8030 },31 "ext_pw_kernel_size": 1,32 "ext_pw_out_channel": 16,33 "input_layer": "nemo_conv",34 "input_size": 80,35 "kernel_size": 3,36 "left_chunk": 18,37 "linear_units": 24,38 "nemo_conv_settings": {39 "conv_channels": 1640 },41 "num_blocks": 2,42 "relative_attention_bias_args": {43 "t5_bias_max_distance": 500,44 "type": "t5"45 },46 "time_reduction": 847 },48 "name": "cascades"49 },50 "auto_map": {51 "AutoConfig": "configuration_phi4mm.Phi4MMConfig",52 "AutoModelForCausalLM": "modeling_phi4mm.Phi4MMForCausalLM",53 "AutoTokenizer": "Xenova/gpt-4o"54 },55 "bos_token_id": 199999,56 "embd_layer": {57 "audio_embd_layer": {58 "compression_rate": 8,59 "downsample_rate": 1,60 "embedding_cls": "audio",61 "enable_gradient_checkpointing": true,62 "projection_cls": "mlp",63 "use_conv_downsample": false,64 "use_qformer": false65 },66 "embedding_cls": "image_audio",67 "image_embd_layer": {68 "crop_size": 448,69 "embedding_cls": "tune_image",70 "enable_gradient_checkpointing": true,71 "hd_transform_order": "sub_glb",72 "image_token_compression_cls": "avg_pool_2d",73 "projection_cls": "mlp",74 "use_hd_transform": true,75 "with_learnable_separator": true76 }77 },78 "embd_pdrop": 0.0,79 "eos_token_id": 199999,80 "full_attn_mod": 1,81 "hidden_act": "silu",82 "hidden_size": 16,83 "img_processor": null,84 "initializer_range": 0.02,85 "intermediate_size": 32,86 "interpolate_factor": 1,87 "lm_head_bias": false,88 "max_position_embeddings": 131072,89 "mlp_bias": false,90 "model_type": "phi4mm",91 "num_attention_heads": 2,92 "num_hidden_layers": 2,93 "num_key_value_heads": 1,94 "original_max_position_embeddings": 4096,95 "pad_token_id": 199999,96 "partial_rotary_factor": 0.75,97 "resid_pdrop": 0.0,98 "rms_norm_eps": 1e-05,99 "rope_scaling": {100 "long_factor": [101 1.0,102 1.0,103 1.0104 ],105 "short_factor": [106 1.0,107 1.0,108 1.0109 ],110 "type": "longrope"111 },112 "rope_theta": 10000.0,113 "sliding_window": 262144,114 "speech_lora": {115 "dp": 0.01,116 "layer": "((layers.*self_attn\\.(qkv|o)_proj)|(layers.*mlp\\.(gate_up|down)_proj))",117 "lora_alpha": 16,118 "r": 8119 },120 "tie_word_embeddings": false,121 "torch_dtype": "float32",122 "transformers_version": "4.49.0",123 "use_cache": true,124 "vision_lora": {125 "dp": 0.0,126 "layer": "layers.*((self_attn\\.(qkv_proj|o_proj))|(mlp\\.(gate_up|down)_proj))",127 "lora_alpha": 16,128 "r": 8129 },130 "vocab_size": 200064131}132 