236 kB

Ctrl+K

1 contributor

v1.9: padding masks on the kernel path for NeuronXLMRobertaModel (per-image additive key bias in attention_block_l1_ln_post_kb); existing models bitwise identical to v1.8

954a874 7 days ago

__init__.py

1.09 kB
v1.8: post-LayerNorm encoders (NeuronXLMRobertaModel; post-LN epilogue outputs + no-RoPE switch) and LNC=2 fix for odd head-group counts 15 days ago
attention_block.py

52.6 kB
CHANGELOG and comment wording 20 days ago
attention_block_l1.py

21.7 kB
v1.5: two images per attention producer pass (LNC=1, <=256 tokens) + fix for <=128-token inputs 22 days ago
attention_block_l1_ln.py

32 kB
v1.9: padding masks on the kernel path for NeuronXLMRobertaModel (per-image additive key bias in attention_block_l1_ln_post_kb); existing models bitwise identical to v1.8 7 days ago
attention_inner.py

60.6 kB
v1.9: padding masks on the kernel path for NeuronXLMRobertaModel (per-image additive key bias in attention_block_l1_ln_post_kb); existing models bitwise identical to v1.8 7 days ago
mlp.py

15.1 kB
v1.1: K4 (mlp/mlp_res_ln) + K1/K3 (res_layernorm) DMA/engine-schedule rework 25 days ago
mlp_fp8.py

12.3 kB
v1.4: 5-bank QK^T PSUM rotation (<=1024 tokens) + experimental opt-in FP8 MLP 22 days ago
mlp_res_ln.py

13.7 kB
v1.8: post-LayerNorm encoders (NeuronXLMRobertaModel; post-LN epilogue outputs + no-RoPE switch) and LNC=2 fix for odd head-group counts 15 days ago
res_layernorm.py

11.8 kB
v1.7: res_layernorm_fold processes row tiles in groups when they do not fit SBUF 17 days ago
swiglu_stream.py

15.3 kB
v1.2: packed-bias one-DMA for SwiGLU (swiglu_stream/_ln) 25 days ago