Skip to content
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

FlashAttention V2/V3 metadata with support for device location #1148

Merged
merged 2 commits into from
Feb 16, 2025
Merged

FlashAttention V2/V3 metadata with support for device location #1148

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d0af975
Flash attn metadata with support for device location
EricLBuehler Feb 16, 2025
e2316e6
Actually cast
EricLBuehler Feb 16, 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
6 changes: 6 additions & 0 deletions mistralrs-core/src/attention.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,9 @@ fn flash_attn(
let window_size_left = sdpa_params.sliding_window;
let window_size_right = if causal { Some(0) } else { None };

let cumulative_seqlens_q = &cumulative_seqlens_q[&q.device().location()];
let cumulative_seqlens_k = &cumulative_seqlens_k[&q.device().location()];

candle_flash_attn::flash_attn_varlen_windowed_softcap(
&q,
&k,
Expand Down Expand Up @@ -94,6 +97,9 @@ fn flash_attn(
let window_size_left = sdpa_params.sliding_window;
let window_size_right = if causal { Some(0) } else { None };

let cumulative_seqlens_q = &cumulative_seqlens_q[&q.device().location()];
let cumulative_seqlens_k = &cumulative_seqlens_k[&q.device().location()];

candle_flash_attn_v3::flash_attn_varlen_windowed(
&q,
&k,
Expand Down
41 changes: 24 additions & 17 deletions mistralrs-core/src/pipeline/inputs_processor.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,19 +120,8 @@ pub mod text_models_inputs_processor {
pub struct FlashParams {
pub max_q: u32,
pub max_k: u32,
pub cumulative_seqlens_q: Tensor,
pub cumulative_seqlens_k: Tensor,
}

impl FlashParams {
pub fn to_device(&self, device: &Device) -> candle_core::Result<Self> {
Ok(Self {
max_k: self.max_k,
max_q: self.max_q,
cumulative_seqlens_k: self.cumulative_seqlens_k.to_device(device)?,
cumulative_seqlens_q: self.cumulative_seqlens_q.to_device(device)?,
})
}
pub cumulative_seqlens_q: HashMap<DeviceLocation, Tensor>,
pub cumulative_seqlens_k: HashMap<DeviceLocation, Tensor>,
}

pub struct InputMetadata {
Expand Down Expand Up @@ -268,6 +257,15 @@ pub mod text_models_inputs_processor {
.cumsum(0)?
.to_dtype(DType::U32)?;

let mut seqlens_q_map = HashMap::new();
let mut seqlens_k_map = HashMap::new();

let devices = mapper.unwrap().get_unique_devices();
for device in devices {
seqlens_q_map.insert(device.location(), seqlens_q.to_device(&device)?);
seqlens_k_map.insert(device.location(), seqlens_k.to_device(&device)?);
}

let input = Tensor::cat(&seqs_tensors, 0).unwrap();
// Only use matmul via f16 if prompt and seqlen > 512
if input.dim(1)? > VIA_F16_TOK_THRESHOLD {
Expand Down Expand Up @@ -345,8 +343,8 @@ pub mod text_models_inputs_processor {
flash_meta: FlashParams {
max_k,
max_q,
cumulative_seqlens_k: seqlens_k,
cumulative_seqlens_q: seqlens_q,
cumulative_seqlens_k: seqlens_k_map,
cumulative_seqlens_q: seqlens_q_map,
},
})
}
Expand Down Expand Up @@ -438,6 +436,15 @@ pub mod text_models_inputs_processor {
.cumsum(0)?
.to_dtype(DType::U32)?;

let mut seqlens_q_map = HashMap::new();
let mut seqlens_k_map = HashMap::new();

let devices = mapper.unwrap().get_unique_devices();
for device in devices {
seqlens_q_map.insert(device.location(), seqlens_q.to_device(&device)?);
seqlens_k_map.insert(device.location(), seqlens_k.to_device(&device)?);
}

set_use_matmul_via_f16(false);

let paged_attn_meta = if paged_attn_metadata.is_some() {
Expand Down Expand Up @@ -502,8 +509,8 @@ pub mod text_models_inputs_processor {
flash_meta: FlashParams {
max_k,
max_q,
cumulative_seqlens_k: seqlens_k,
cumulative_seqlens_q: seqlens_q,
cumulative_seqlens_k: seqlens_k_map,
cumulative_seqlens_q: seqlens_q_map,
},
})
}
Expand Down
4 changes: 2 additions & 2 deletions mistralrs-core/src/pipeline/normal.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -860,7 +860,7 @@ impl Loader for NormalLoader {
inputs.context_lens.clone(),
inputs.position_ids.clone(),
None,
&inputs.flash_meta.to_device(model.device())?,
&inputs.flash_meta.clone(),
)
})
.collect::<candle_core::Result<Vec<_>>>()?;
Expand Down Expand Up @@ -1314,7 +1314,7 @@ impl Pipeline for NormalPipeline {
let seqlen_offsets = seqlen_offsets.clone();
let context_lens = context_lens.clone();
let position_ids = position_ids.clone();
let flash_meta = flash_meta.to_device(model.device())?;
let flash_meta = flash_meta.clone();

handles.push(std::thread::spawn(move || {
#[cfg(feature = "cuda")]
Expand Down
Loading