keras/smollm3_3b_en

This is a SmolLM3 model uploaded using the KerasHub library and can be used with JAX, TensorFlow, and PyTorch backends. This model is related to a CausalLM task.

Model config:

• name: smol_lm3_backbone
• trainable: True
• dtype: {'module': 'keras', 'class_name': 'DTypePolicy', 'config': {'name': 'float32'}, 'registered_name': None}
• vocabulary_size: 128256
• hidden_dim: 2048
• intermediate_dim: 11008
• num_layers: 36
• num_attention_heads: 16
• num_key_value_heads: 4
• attention_bias: False
• attention_dropout: 0.0
• rope_layer_enabled_list: [True, True, True, False, True, True, True, False, True, True, True, False, True, True, True, False, True, True, True, False, True, True, True, False, True, True, True, False, True, True, True, False, True, True, True, False]
• layer_types: ['full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention']
• mlp_bias: False
• layer_norm_epsilon: 1e-06
• max_position_embeddings: 65536
• rope_theta: 5000000.0
• partial_rotary_factor: 1.0

This model card has been generated automatically and should be completed by the model author. See Model Cards documentation for more information.