企业官网建设流程全解析

Transformer类接口

【免费下载链接】ops-transformer本项目是CANN提供的transformer类大模型算子库，实现网络在NPU上加速计算。项目地址: https://gitcode.com/cann/ops-transformer

• Transformer类torch接口列表
• causal_conv1d_fn
• causal_conv1d_update
• compressor
• elastic_buffer
• flash_attn
• get_low_latency_ccl_buffer_size
• grouped_matmul_activation_quant
• indexer_quant_cache
• inplace_partial_rotary_mul
• inplace_partial_rotary_mul_backward
• kv_compress_epilog
• lightning_indexer
• low_latency_combine
• low_latency_dispatch
• mega_moe
• mhc_post
• mhc_pre_sinkhorn
• mixed_quant_sparse_flash_mla
• moe_token_permute
• qkv_rms_norm_rope_cache_with_k_scale
• quant_lightning_indexer
• scatter_pa_kv_cache_with_k_scale
• sparse_flash_mla
• sparse_flash_mla_grad
• sparse_lightning_indexer_kl_loss_grad

【免费下载链接】ops-transformer本项目是CANN提供的transformer类大模型算子库，实现网络在NPU上加速计算。项目地址: https://gitcode.com/cann/ops-transformer