Skip to content

flash feature refactor #778

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
michaelfeil wants to merge 3 commits into
base: main
Choose a base branch
from
Open

flash feature refactor #778

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6ff8966
flash feature refactor
michaelfeil Dec 18, 2025
6679919
add varlen attention interface
michaelfeil Dec 18, 2025
6e5b163
Update lib.rs
michaelfeil Dec 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
82 changes: 27 additions & 55 deletions backends/candle/src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,6 +115,23 @@ enum Config {
XlmRoberta(BertConfig),
}

/// Helper function to check if flash attention should be used
/// require_fa2_capabilities: true for models that need flash attention v2, false for v1/v2 compatible models
fn should_use_flash_attention(require_fa2_capabilities: bool) -> bool {
let flash_attn_enabled = &std::env::var("USE_FLASH_ATTENTION").unwrap_or("true".to_string()).to_lowercase() == "true";

if cfg!(not(feature = "cuda")) {
// if not cuda support, always false for now.
return false;
};
// cuda
if require_fa2_capabilities {
cfg!(feature = "flash-attn") && flash_attn_enabled
} else {
cfg!(any(feature = "flash-attn", feature = "flash-attn-v1")) && flash_attn_enabled
}
}

pub struct CandleBackend {
device: Device,
model: Box<dyn Model + Send>,
Expand Down Expand Up @@ -307,12 +324,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::Bert(config), Device::Cuda(_)) => {
if cfg!(any(feature = "flash-attn", feature = "flash-attn-v1"))
&& dtype == DType::F16
// Allow disabling because of flash attention v1 precision problems
// See: https://github.com/huggingface/text-embeddings-inference/issues/37
&& &std::env::var("USE_FLASH_ATTENTION").unwrap_or("True".to_string()).to_lowercase() == "true"
{
if dtype == DType::F16 && should_use_flash_attention(false) {
match config {
BertConfigWrapper::JinaBert(config) => {
tracing::info!("Starting FlashJinaBert model on {:?}", device);
Expand Down Expand Up @@ -354,13 +366,8 @@ impl CandleBackend {
(
Config::Camembert(config) | Config::Roberta(config) | Config::XlmRoberta(config),
Device::Cuda(_),
) => {
if cfg!(any(feature = "flash-attn", feature = "flash-attn-v1"))
&& dtype == DType::F16
// Allow disabling because of flash attention v1 precision problems
// See: https://github.com/huggingface/text-embeddings-inference/issues/37
&& &std::env::var("USE_FLASH_ATTENTION").unwrap_or("True".to_string()).to_lowercase() == "true"
{
) => {
if dtype == DType::F16 && should_use_flash_attention(false) {
tracing::info!("Starting FlashBert model on {:?}", device);
Ok(Box::new(
FlashBertModel::load_roberta(vb, &config, model_type).s()?,
Expand All @@ -374,12 +381,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::DistilBert(config), Device::Cuda(_)) => {
if cfg!(feature = "flash-attn")
&& dtype == DType::F16
&& &std::env::var("USE_FLASH_ATTENTION")
.unwrap_or("True".to_string())
.to_lowercase()
== "true"
if dtype == DType::F16 && should_use_flash_attention(true)
{
tracing::info!("Starting FlashDistilBert model on {:?}", device);
Ok(Box::new(
Expand All @@ -405,12 +407,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::Gte(config), Device::Cuda(_)) => {
if dtype != DType::F16
|| !cfg!(any(feature = "flash-attn", feature = "flash-attn-v1"))
|| &std::env::var("USE_FLASH_ATTENTION")
.unwrap_or("True".to_string())
.to_lowercase()
!= "true"
if dtype != DType::F16 || !should_use_flash_attention(false)
{
tracing::info!("Starting GTE model on {:?}", device);
Ok(Box::new(GTEModel::load(vb, &config, model_type).s()?))
Expand All @@ -421,13 +418,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::Mistral(config), Device::Cuda(_)) => {
if dtype != DType::F16
|| !cfg!(feature = "flash-attn")
|| get_runtime_compute_cap().unwrap() < 80
|| &std::env::var("USE_FLASH_ATTENTION")
.unwrap_or("True".to_string())
.to_lowercase()
!= "true"
if dtype != DType::F16 || !should_use_flash_attention(true)
{
return Err(BackendError::Start("Mistral is only supported on Cuda devices in fp16 with flash attention v2 enabled".to_string()));
}
Expand All @@ -438,11 +429,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::ModernBert(config), Device::Cuda(_)) => {
if cfg!(feature = "flash-attn")
&& dtype == DType::F16
// Allow disabling because of flash attention v1 precision problems
// See: https://github.com/huggingface/text-embeddings-inference/issues/37
&& &std::env::var("USE_FLASH_ATTENTION").unwrap_or("True".to_string()).to_lowercase() == "true"
if dtype == DType::F16 && should_use_flash_attention(true)
{
tracing::info!("Starting FlashModernBert model on {:?}", device);
Ok(Box::new(
Expand All @@ -459,12 +446,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::NomicBert(config), Device::Cuda(_)) => {
if cfg!(feature = "flash-attn")
&& dtype == DType::F16
&& &std::env::var("USE_FLASH_ATTENTION")
.unwrap_or("True".to_string())
.to_lowercase()
== "true"
if dtype == DType::F16 && should_use_flash_attention(true)
{
tracing::info!("Starting FlashNomicBert model on {:?}", device);
Ok(Box::new(
Expand All @@ -477,12 +459,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::Qwen2(config), Device::Cuda(_)) => {
if dtype != DType::F16
|| !cfg!(any(feature = "flash-attn", feature = "flash-attn-v1"))
|| &std::env::var("USE_FLASH_ATTENTION")
.unwrap_or("True".to_string())
.to_lowercase()
!= "true"
if dtype != DType::F16 || !should_use_flash_attention(false)
{
return Err(BackendError::Start("Qwen2 is only supported on Cuda devices in fp16 with flash attention v2 enabled".to_string()));
}
Expand All @@ -493,12 +470,7 @@ impl CandleBackend {
}
#[cfg(feature = "cuda")]
(Config::Qwen3(config), Device::Cuda(_)) => {
if dtype != DType::F16
|| !cfg!(any(feature = "flash-attn", feature = "flash-attn-v1"))
|| &std::env::var("USE_FLASH_ATTENTION")
.unwrap_or("True".to_string())
.to_lowercase()
!= "true"
if dtype != DType::F16 || !should_use_flash_attention(false)
{
tracing::info!("Starting Qwen3 model on {:?}", device);
Ok(Box::new(Qwen3Model::load(vb, &config, model_type).s()?))
Expand Down