Laurent Mazare b2e816752b Use the faster rms-norm kernel for llama. (#2107) ...

* Use the faster rms-norm kernel for llama.

* Use the fast variant by default.

2024-04-22 18:52:00 +02:00

..

activation.rs

Add the StarCoder2 model. (#1779)

2024-02-28 21:02:41 +01:00

batch_norm.rs

Detach the tensors on batch-norm eval. (#1702)

2024-02-13 14:26:32 +01:00

conv.rs

Bugfix for applying the bias in conv1d-transpose. (#1732)

2024-02-18 22:51:20 +01:00

embedding.rs

Add clone to various nn layers. (#910)

2023-09-20 11:33:51 +01:00

encoding.rs

Fix clippy lints for 1.76. (#1682)

2024-02-08 16:48:47 +01:00

func.rs

Allow for different behavior between training and eval (#1213)

2023-10-29 07:53:09 +01:00

group_norm.rs

Add clone to various nn layers. (#910)

2023-09-20 11:33:51 +01:00

init.rs

VarBuilder cleanup (#627)

2023-08-27 18:03:26 +01:00

layer_norm.rs

Use the faster rms-norm kernel for llama. (#2107)

2024-04-22 18:52:00 +02:00

lib.rs

Fast kernels for rotary embeddings. (#1928)

2024-03-24 22:48:52 +01:00

linear.rs

Support for attention bias in gemma + refactor things a bit. (#1744)

2024-02-22 09:35:28 +01:00

loss.rs

Add Binary Cross Entropy With Logit Loss to nn crate (#1157)

2023-10-23 17:12:44 +01:00

ops.rs

RmsNorm kernel for metal. (#1895)

2024-03-21 09:48:56 +01:00

optim.rs

Expose AdamW parameters (#1449)

2023-12-16 18:41:56 -06:00

rnn.rs

Relax the contiguous check for cuda kernels. (#2000)

2024-04-03 09:02:38 +02:00

rotary_emb.rs

Add the rope THD kernel. (#2014)

2024-04-05 08:32:58 +02:00

sequential.rs

Make func cloneable. (#1137)

2023-10-20 16:28:50 +01:00

var_builder.rs

Add a convenient way to rename tensors accessed through a varbuilder. (#2052)

2024-04-13 12:09:41 +02:00

var_map.rs

Tweak the VarMap set type. (#1758)

2024-02-25 20:50:08 +01:00