rudrashah/RLM_hingu
319
1{2 "module": "keras_hub.src.models.gemma3.gemma3_causal_lm",3 "class_name": "Gemma3CausalLM",4 "config": {5 "backbone": {6 "module": "keras_hub.src.models.gemma3.gemma3_backbone",7 "class_name": "Gemma3Backbone",8 "config": {9 "name": "gemma3_backbone",10 "trainable": true,11 "vocabulary_size": 262144,12 "image_size": null,13 "num_layers": 26,14 "num_query_heads": 4,15 "num_key_value_heads": 1,16 "hidden_dim": 1152,17 "intermediate_dim": 6912,18 "head_dim": 256,19 "query_head_dim_normalize": true,20 "use_query_key_norm": true,21 "use_post_ffw_norm": true,22 "use_post_attention_norm": true,23 "attention_logit_soft_cap": null,24 "final_logit_soft_cap": null,25 "use_sliding_window_attention": true,26 "sliding_window_size": 512,27 "local_rope_scaling_factor": 1.0,28 "global_rope_scaling_factor": 1.0,29 "vision_encoder": null,30 "layer_norm_epsilon": 1e-06,31 "dropout": 032 },33 "registered_name": "keras_hub>Gemma3Backbone"34 },35 "preprocessor": {36 "module": "keras_hub.src.models.gemma3.gemma3_causal_lm_preprocessor",37 "class_name": "Gemma3CausalLMPreprocessor",38 "config": {39 "name": "gemma3_causal_lm_preprocessor",40 "trainable": true,41 "dtype": {42 "module": "keras",43 "class_name": "DTypePolicy",44 "config": {45 "name": "float32"46 },47 "registered_name": null48 },49 "tokenizer": {50 "module": "keras_hub.src.models.gemma3.gemma3_tokenizer",51 "class_name": "Gemma3Tokenizer",52 "config": {53 "name": "gemma3_tokenizer",54 "trainable": true,55 "dtype": {56 "module": "keras",57 "class_name": "DTypePolicy",58 "config": {59 "name": "int32"60 },61 "registered_name": null62 },63 "config_file": "tokenizer.json",64 "proto": null,65 "sequence_length": null,66 "add_bos": false,67 "add_eos": false68 },69 "registered_name": "keras_hub>Gemma3Tokenizer"70 },71 "config_file": "preprocessor.json",72 "sequence_length": 256,73 "add_start_token": true,74 "add_end_token": true,75 "num_vision_tokens_per_image": 0,76 "max_images_per_prompt": 077 },78 "registered_name": "keras_hub>Gemma3CausalLMPreprocessor"79 },80 "name": "gemma3_causal_lm"81 },82 "registered_name": "keras_hub>Gemma3CausalLM"83}