Skip to content

Commit

Permalink
[torch.compile] hide slicing under custom op for inductor (vllm-proje…
Browse files Browse the repository at this point in the history
  • Loading branch information
youkaichao authored Sep 12, 2024
1 parent 42ffba1 commit 7de49aa
Show file tree
Hide file tree
Showing 2 changed files with 74 additions and 35 deletions.
4 changes: 3 additions & 1 deletion tests/compile/test_full_graph.py
Original file line number Diff line number Diff line change
Expand Up @@ -16,5 +16,7 @@ def test_full_graph(model):
"The future of AI is",
]
sampling_params = SamplingParams(temperature=0)
llm = LLM(model="meta-llama/Meta-Llama-3-8B")
llm = LLM(model="meta-llama/Meta-Llama-3-8B",
enforce_eager=True,
load_format="dummy")
llm.generate(prompts, sampling_params)
105 changes: 71 additions & 34 deletions vllm/attention/backends/flash_attn.py
Original file line number Diff line number Diff line change
Expand Up @@ -122,6 +122,40 @@ def _(
return torch.empty_like(decode_query)


@torch.library.custom_op("vllm::reshape_and_cache_flash",
mutates_args=["kv_cache"])
def reshape_and_cache_flash(
key: torch.Tensor,
value: torch.Tensor,
kv_cache: torch.Tensor,
slot_mapping: torch.Tensor,
kv_cache_dtype: str,
k_scale: float,
v_scale: float,
) -> None:
"""Inductor cannot deal with inplace operations on views.
See https://github.com/pytorch/pytorch/issues/131192
and https://github.com/pytorch/pytorch/issues/130174
This is a workaround to hide the view operation from the inductor.
"""
return torch.ops._C_cache_ops.reshape_and_cache_flash(
key, value, kv_cache[0], kv_cache[1], slot_mapping, kv_cache_dtype,
k_scale, v_scale)


@reshape_and_cache_flash.register_fake # type: ignore
def _(
key: torch.Tensor,
value: torch.Tensor,
kv_cache: torch.Tensor,
slot_mapping: torch.Tensor,
kv_cache_dtype: str,
k_scale: float,
v_scale: float,
) -> None:
pass


class FlashAttentionBackend(AttentionBackend):

@staticmethod
Expand Down Expand Up @@ -653,11 +687,10 @@ def forward(
# Reshape the input keys and values and store them in the cache.
# If kv_cache is not provided, the new key and value tensors are
# not cached. This happens during the initial memory profiling run.
ops.reshape_and_cache_flash(
torch.ops.vllm.reshape_and_cache_flash(
key,
value,
key_cache,
value_cache,
kv_cache,
attn_metadata.slot_mapping.flatten(),
self.kv_cache_dtype,
k_scale,
Expand All @@ -669,7 +702,6 @@ def forward(
assert key.shape[0] == num_prefill_tokens + num_decode_tokens
assert value.shape[0] == num_prefill_tokens + num_decode_tokens

output = torch.empty_like(query)
# Query for decode. KV is not needed because it is already cached.
decode_query = query[num_prefill_tokens:]
# QKV for prefill.
Expand All @@ -680,14 +712,17 @@ def forward(
assert query.shape[0] == num_prefill_tokens
assert decode_query.shape[0] == num_decode_tokens

prefill_output: Optional[torch.Tensor] = None
decode_output: Optional[torch.Tensor] = None

if prefill_meta := attn_metadata.prefill_metadata:
# Prompt run.
if (kv_cache is None or prefill_meta.block_tables is None
or prefill_meta.block_tables.numel() == 0):
# normal attention
# When block_tables are not filled, it means q and k are the
# prompt, and they have the same length.
out = torch.ops.vllm.flash_attn_varlen_func(
prefill_output = torch.ops.vllm.flash_attn_varlen_func(
q=query,
k=key,
v=value,
Expand All @@ -701,42 +736,44 @@ def forward(
alibi_slopes=self.alibi_slopes,
softcap=self.logits_soft_cap,
)
assert output[:num_prefill_tokens].shape == out.shape
output[:num_prefill_tokens] = out
else:
# prefix-enabled attention
assert prefill_meta.seq_lens is not None
max_seq_len = max(prefill_meta.seq_lens)
output[:
num_prefill_tokens] = torch.ops.vllm.flash_attn_varlen_func( # noqa
q=query,
k=key_cache,
v=value_cache,
cu_seqlens_q=prefill_meta.query_start_loc,
max_seqlen_q=prefill_meta.max_query_len,
cu_seqlens_k=prefill_meta.seq_start_loc,
max_seqlen_k=max_seq_len,
softmax_scale=self.scale,
causal=True,
alibi_slopes=self.alibi_slopes,
block_table=prefill_meta.block_tables,
softcap=self.logits_soft_cap,
)

if decode_meta := attn_metadata.decode_metadata:
# Decoding run.
output[
num_prefill_tokens:] = torch.ops.vllm.flash_attn_with_kvcache(
decode_query.unsqueeze(1),
key_cache,
value_cache,
block_table=decode_meta.block_tables,
cache_seqlens=decode_meta.seq_lens_tensor,
prefill_output = torch.ops.vllm.flash_attn_varlen_func( # noqa
q=query,
k=key_cache,
v=value_cache,
cu_seqlens_q=prefill_meta.query_start_loc,
max_seqlen_q=prefill_meta.max_query_len,
cu_seqlens_k=prefill_meta.seq_start_loc,
max_seqlen_k=max_seq_len,
softmax_scale=self.scale,
causal=True,
alibi_slopes=self.alibi_slopes,
block_table=prefill_meta.block_tables,
softcap=self.logits_soft_cap,
).squeeze(1)
)

# Reshape the output tensor.
if decode_meta := attn_metadata.decode_metadata:
# Decoding run.
decode_output = torch.ops.vllm.flash_attn_with_kvcache(
decode_query.unsqueeze(1),
key_cache,
value_cache,
block_table=decode_meta.block_tables,
cache_seqlens=decode_meta.seq_lens_tensor,
softmax_scale=self.scale,
causal=True,
alibi_slopes=self.alibi_slopes,
softcap=self.logits_soft_cap,
).squeeze(1)

if prefill_output is None:
assert decode_output is not None
return decode_output.view(num_decode_tokens, hidden_size)
if decode_output is None:
assert prefill_output is not None
return prefill_output.view(num_prefill_tokens, hidden_size)
output = torch.cat([prefill_output, decode_output], dim=0)
return output.view(num_tokens, hidden_size)

0 comments on commit 7de49aa

Please sign in to comment.