Commits

Commits on Sep 19, 2024

[Mamba2] Move dt calculations to kernel (huggingface#33520 )
vasqu
authored

Commits on Aug 29, 2024

Mamba2 conversion script for original models (huggingface#32580 )
vasqu
authored

Commits on Aug 28, 2024

Fix: Jamba batched generation (huggingface#32914 )
vasqu
authored

Commits on Aug 20, 2024

Fix: Mamba2 norm_before_gate usage (huggingface#32686 )
vasqu
authored

Commits on Aug 19, 2024

Fix: Mamba2 generation mismatch between input_ids and inputs_embeds (huggingface#32694 )
vasqu
authored

Commits on Jul 23, 2024

Fix mask creations of GPTNeoX and GPT2 (huggingface#31944 )
vasqu
authored

Commits on Jul 5, 2024

Fix galore lr display with schedulers (huggingface#31710 )
vasqu
authored

Commits on Jun 26, 2024

[GPT-NeoX] Add SDPA support (huggingface#31031 )
vasqu
authored

Commits on Jun 19, 2024

[GPT2] Add SDPA support (huggingface#31172 )
vasqu
authored

Commits on May 20, 2024

Fix a shape annotation and typos in mamba slow forward (huggingface#30691 )
vasqu
authored

Commits on May 8, 2024

Immutability for data collators (huggingface#30603 )
vasqu
authored

Commits on Apr 30, 2024

Fix seq2seq collator padding (huggingface#30556 )
vasqu
authored

Commits on Apr 24, 2024

fix uncaught init of linear layer in clip's/siglip's for image classification models (huggingface#30435 )
vasqu
authored

Commits on Apr 10, 2024

Fix accelerate kwargs for versions <0.28.0 (huggingface#30086 )
vasqu
authored

Commits on Apr 8, 2024

fix learning rate display in trainer when using galore optimizer (huggingface#30085 )
vasqu
authored

Commits on Mar 27, 2024

Mamba slow_forward gradient fix (huggingface#29563 )
vasqu
authored