Feat ring flash attn #1305
Draft
Feat ring flash attn #1305
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
S1ro1
force-pushed
the
feat-ring-flash-attn
branch
from
6c64e0b to
c749b20
Compare
samsja
reviewed
Nov 19, 2025
Comment on lines
+178
to
+242
| def substitute_prime_rl_flash_attn(process_group: torch.distributed.ProcessGroup, heads_k_stride: int) -> None: | ||
| from ring_flash_attn import llama3_flash_attn_varlen_func | ||
|
|
||
| global ATTN_IMPL2CLASS | ||
|
|
||
| class RingFlashAttention(FlashAttention): | ||
| def forward( | ||
| self, | ||
| hidden_states: torch.Tensor, | ||
| position_embeddings: tuple[torch.Tensor, torch.Tensor], | ||
| cu_seqlens: torch.LongTensor | None = None, | ||
| max_seqlen: int | None = None, | ||
| ) -> tuple[torch.Tensor, torch.Tensor | None]: | ||
| input_shape = hidden_states.shape[:-1] | ||
| hidden_shape = (*input_shape, -1, self.head_dim) | ||
|
|
||
| query_states = self.q_proj(hidden_states).view(hidden_shape) | ||
| key_states = self.k_proj(hidden_states).view(hidden_shape) | ||
| value_states = self.v_proj(hidden_states).view(hidden_shape) | ||
|
|
||
| if self.use_qk_norm: # main diff from Llama | ||
| query_states = self.q_norm(query_states) | ||
| key_states = self.k_norm(key_states) | ||
|
|
||
| query_states = query_states.transpose(1, 2) | ||
| key_states = key_states.transpose(1, 2) | ||
| value_states = value_states.transpose(1, 2) | ||
|
|
||
| cos, sin = position_embeddings | ||
| query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin) | ||
|
|
||
| from ring_flash_attn.adapters.hf_adapter import DATA_PARAMS | ||
|
|
||
| cu_seqlens_q = DATA_PARAMS["cu_seqlens_q"] | ||
| cu_seqlens_k = DATA_PARAMS["cu_seqlens_k"] | ||
| max_seqlen_q = DATA_PARAMS["max_seqlen_q"] | ||
| max_seqlen_k = DATA_PARAMS["max_seqlen_k"] | ||
| local_k_slice = DATA_PARAMS["local_k_slice"] | ||
|
|
||
| # TODO: Can we optimize the rotary applicaiton instead of double transpose? | ||
| query_states = query_states.transpose(1, 2) | ||
| key_states = key_states.transpose(1, 2) | ||
| value_states = value_states.transpose(1, 2) | ||
| out = llama3_flash_attn_varlen_func( | ||
| query_states[0], | ||
| key_states[0], | ||
| value_states[0], | ||
| cu_seqlens_q=cu_seqlens_q, | ||
| cu_seqlens_k=cu_seqlens_k, | ||
| max_seqlen_q=max_seqlen_q, | ||
| max_seqlen_k=max_seqlen_k, | ||
| local_k_slice=local_k_slice, | ||
| causal=True, | ||
| group=process_group, | ||
| heads_k_stride=heads_k_stride, | ||
| ) | ||
| out = out.contiguous() | ||
| attn_output = out.view(1, out.shape[0], -1) | ||
| attn_weights = None | ||
|
|
||
| # attn_output = attn_output.reshape(*input_shape, -1).contiguous() | ||
| attn_output = self.o_proj(attn_output) | ||
| return attn_output, attn_weights | ||
|
|
||
| ATTN_IMPL2CLASS["flash_attention_2"].forward = RingFlashAttention.forward |
Member
There was a problem hiding this comment.
not the biggest fan of this pattern, smell like transformer magic.
what about just having flash_attenion_2_cp in attn.impl ? and doing parralel.cp 2 would automatically change the attention to it
samsja
reviewed
Nov 19, 2025
Comment on lines
+244
to
+255
| input_ids = maybe_shard_for_cp( | ||
| micro_batch["input_ids"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | ||
| ) | ||
| advantages = maybe_shard_for_cp( | ||
| micro_batch["advantages"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | ||
| ) | ||
| loss_mask = maybe_shard_for_cp( | ||
| micro_batch["loss_mask"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | ||
| ) | ||
| inference_logprobs = maybe_shard_for_cp( | ||
| micro_batch["inference_logprobs"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | ||
| ) |
Member
There was a problem hiding this comment.
I would rather do
Suggested change
| input_ids = maybe_shard_for_cp( | |
| micro_batch["input_ids"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | |
| ) | |
| advantages = maybe_shard_for_cp( | |
| micro_batch["advantages"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | |
| ) | |
| loss_mask = maybe_shard_for_cp( | |
| micro_batch["loss_mask"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | |
| ) | |
| inference_logprobs = maybe_shard_for_cp( | |
| micro_batch["inference_logprobs"].to("cuda"), cp_rank=cp_rank, cp_world_size=config.model.cp | |
| ) | |
| if config.model.cp > 1: | |
| shard_for_cp(...) |
not fan of the maybe pattern
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I was too lazy to explain my changes, decreasing my chances of this ever getting merged.
GitHub Issue: [Issue ID]
Linear Issue: Resolves [Issue ID]