GradientStorage Struct Reference

Collaboration diagram for GradientStorage:

Public Attributes
size_t	backprop_base
size_t	total_gradient_floats
size_t	logits_copy_offset
size_t	actual_tokens_offset
size_t	d_logits_offset
size_t	final_output_copy_offset
size_t	d_final_output_offset
size_t	d_embed_weights_offset
size_t	final_ln_input_copy_offset
size_t	final_ln_mean_copy_offset
size_t	final_ln_rstd_copy_offset
size_t	final_ln_gamma_copy_offset
size_t	final_ln_beta_copy_offset
size_t	d_final_ln_input_offset
size_t	d_final_ln_gamma_offset
size_t	d_final_ln_beta_offset
LayerGradients *	layers
size_t	d_pos_embed_offset
size_t	training_pair_tokens_offset
size_t	training_pair_tokens_count
size_t	training_pair_tokens_stride
size_t	training_pair_capacity
size_t	token_emb_m_offset
size_t	token_emb_v_offset
size_t	token_emb_ema_offset
size_t	pos_emb_m_offset
size_t	pos_emb_v_offset
size_t	pos_emb_ema_offset
size_t	final_ln_gamma_m_offset
size_t	final_ln_gamma_v_offset
size_t	final_ln_gamma_ema_offset
size_t	final_ln_beta_m_offset
size_t	final_ln_beta_v_offset
size_t	final_ln_beta_ema_offset
size_t	seq_cls_weight_m_offset
size_t	seq_cls_weight_v_offset
size_t	seq_cls_weight_ema_offset
size_t	seq_cls_bias_m_offset
size_t	seq_cls_bias_v_offset
size_t	seq_cls_bias_ema_offset
size_t	seq_cls_logits_offset
size_t	d_seq_cls_logits_offset
size_t	seq_cls_pooled_offset
size_t	d_seq_cls_pooled_offset
size_t	d_seq_cls_weight_offset
size_t	d_seq_cls_bias_offset
size_t	layer_backprop_stride

---

The documentation for this struct was generated from the following file:

• main.c