Prefix caching

backends/v3/src/queue.rs

@@ -316,10 +316,15 @@ impl State {
                         + self.speculate
                         - 1;
 
-                    match block_allocator
-                        .allocate(tokens, entry.request.input_ids.clone())
-                        .await
-                    {
+                    // If users wants the prefill logprobs, we cannot reuse the cache.
+                    // So no input_ids for the radix tree.
+                    let input_ids = if entry.request.decoder_input_details {
+                        None
+                    } else {
+                        entry.request.input_ids.clone()
+                    };
+
+                    match block_allocator.allocate(tokens, input_ids).await {
                         None => {
                             // Entry is over budget
                             // Add it back to the front

backends/v3/src/radix.rs

@@ -205,6 +205,7 @@ pub struct RadixTrie {
     /// call that a real time lookup would require.
     time: u64,
 }
+
 impl Default for RadixTrie {
     fn default() -> Self {
         Self::new()

flake.nix

@@ -81,6 +81,7 @@
                 grpcio-status
                 grpcio-tools
                 hf-transfer
+                ipdb
                 loguru
                 mamba-ssm
                 marlin-kernels

server/text_generation_server/layers/attention/__init__.py

@@ -6,7 +6,12 @@
 if os.getenv("USE_FLASH_ATTENTION", "").lower() == "false":
     raise ImportError("`USE_FLASH_ATTENTION` is false.")
 if SYSTEM == "cuda":
-    from .cuda import attention, paged_attention, reshape_and_cache, SUPPORTS_WINDOWING
+    from .cuda import (
+        attention,
+        paged_attention,
+        reshape_and_cache,
+        SUPPORTS_WINDOWING,
+    )
 elif SYSTEM == "rocm":
     from .rocm import attention, paged_attention, reshape_and_cache, SUPPORTS_WINDOWING
 elif SYSTEM == "ipex":

server/text_generation_server/layers/attention/cuda.py

@@ -76,7 +76,7 @@ def paged_attention(
     # sequences or heads is large, we use V1 since there is enough work
     # to parallelize.
     if ATTENTION == "flashinfer":
-        from text_generation_server.layers.attention.flash_infer import decode_state
+        from text_generation_server.layers.attention.flashinfer import decode_state
 
         return decode_state.get().forward(
             query.contiguous(),
@@ -221,24 +221,27 @@ def paged_attention(
 if ATTENTION == "flashinfer":
 
     def attention(
-        q,
-        k,
-        v,
+        q: torch.Tensor,
+        k: torch.Tensor,
+        v: torch.Tensor,
+        key_cache: torch.Tensor,
+        value_cache: torch.Tensor,
         cu_seqlens,
         max_s,
         softmax_scale,
         window_size_left=-1,
         causal=True,
         softcap=0.0,
     ):
-        from text_generation_server.layers.attention.flash_infer import prefill_state
+        assert window_size_left == -1, "Windowing is not supported with flash infer"
+        from text_generation_server.layers.attention.flashinfer import (
+            prefill_with_paged_kv_state,
+        )
 
-        return prefill_state.get().forward(
-            q,
-            k,
-            v,
+        return prefill_with_paged_kv_state.get().forward(
+            q.contiguous(),
             causal=causal,
-            window_left=window_size_left,
+            paged_kv_cache=(key_cache, value_cache),
             logits_soft_cap=softcap,
             sm_scale=softmax_scale,
         )
@@ -249,6 +252,8 @@ def attention(
         q,
         k,
         v,
+        key_cache: torch.Tensor,
+        value_cache: torch.Tensor,
         cu_seqlens,
         max_s,
         softmax_scale,
@@ -289,6 +294,8 @@ def attention(
         q,
         k,
         v,
+        key_cache: torch.Tensor,
+        value_cache: torch.Tensor,
         cu_seqlens,
         max_s,
         softmax_scale,

server/text_generation_server/layers/attention/flash_infer.py → server/text_generation_server/layers/attention/flashinfer.py

@@ -9,6 +9,10 @@
     "prefill_state"
 )
 
+prefill_with_paged_kv_state: ContextVar[
+    flashinfer.BatchPrefillWithPagedKVCacheWrapper
+] = ContextVar("prefill_with_paged_kv_state")
+
 decode_state: ContextVar[flashinfer.BatchDecodeWithPagedKVCacheWrapper] = ContextVar(
     "decode_state"
 )
@@ -24,6 +28,78 @@ def get_workspace(device):
     return workspace
 
 
+def create_prefill_with_paged_kv_state(
+    *,
+    device: torch.device,
+):
+    """Create a prefill state that uses the KV cache."""
+    workspace_buffer = get_workspace(device)
+    return flashinfer.BatchPrefillWithPagedKVCacheWrapper(
+        workspace_buffer, kv_layout="NHD", use_cuda_graph=False
+    )
+
+
+@contextmanager
+def use_prefill_with_paged_kv_state(
+    *,
+    state: flashinfer.BatchPrefillWithPagedKVCacheWrapper,
+    block_tables: torch.Tensor,
+    cu_seqlens: torch.Tensor,
+    input_lengths: torch.Tensor,
+    num_heads: int,
+    num_kv_heads: int,
+    head_size: int,
+    page_size: int,
+    query_dtype: str = "float16",
+):
+    """
+    Context manager to set the active flashinfer prefill state to the given
+    `state` and parameters. This state will be used by all calls to the
+    `attention` function while the context manager is active.
+    """
+
+    indptr = torch.zeros(
+        input_lengths.shape[0] + 1, device=input_lengths.device, dtype=torch.int32
+    )
+    # Round up to page size and then calculate the cumulative sum to get
+    # the indices into the block table.
+    torch.add(input_lengths, page_size - 1, out=indptr[1:])
+    indptr[1:].div_(page_size, rounding_mode="floor")
+    indptr[1:].cumsum_(-1)
+
+    # Get the lengths of the last page in a block.
+    if page_size == 1:
+        last_page_len = torch.ones(
+            input_lengths.shape[0], dtype=torch.int32, device=input_lengths.device
+        )
+    else:
+        last_page_len = torch.empty(
+            input_lengths.shape[0], dtype=torch.int32, device=input_lengths.device
+        )
+        torch.sub(input_lengths, 1, out=last_page_len)
+        last_page_len.remainder_(page_size)
+        last_page_len += 1
+
+    token = prefill_with_paged_kv_state.set(state)
+    try:
+        state.begin_forward(
+            qo_indptr=cu_seqlens,
+            paged_kv_indptr=indptr,
+            paged_kv_indices=block_tables,
+            paged_kv_last_page_len=last_page_len,
+            num_qo_heads=num_heads,
+            num_kv_heads=num_kv_heads,
+            head_dim=head_size,
+            q_data_type=query_dtype,
+            page_size=page_size,
+        )
+        yield
+    finally:
+        state.end_forward()
+        if token is not None:
+            prefill_with_paged_kv_state.reset(token)
+
+
 def create_prefill_state(
     *,
     device: torch.device,

server/text_generation_server/layers/medusa.py

@@ -32,6 +32,8 @@ def __init__(self, config, medusa_config, weights):
         )
 
     def forward(self, x):
+        if not self.heads:
+            return None
         speculative_logits = torch.stack([head(x) for head in self.heads], dim=1)
         return speculative_logits
 

server/text_generation_server/models/custom_modeling/flash_cohere_modeling.py

@@ -298,6 +298,8 @@ def forward(
                 query,
                 key,
                 value,
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_dbrx_modeling.py

@@ -337,6 +337,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_deepseek_v2_modeling.py

@@ -365,6 +365,8 @@ def forward(
                 query,
                 key,
                 value,
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_gemma2_modeling.py

@@ -238,6 +238,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_gemma_modeling.py

@@ -232,6 +232,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_gpt2_modeling.py

@@ -232,6 +232,8 @@ def forward(
                 query,
                 key,
                 value,
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_llama_modeling.py

@@ -220,6 +220,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_mistral_modeling.py

@@ -219,6 +219,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_mixtral_modeling.py

@@ -276,6 +276,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_neox_modeling.py

@@ -173,6 +173,8 @@ def forward(
                 qkv[:, 0],
                 qkv[:, 1],
                 qkv[:, 2],
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_phi_modeling.py

@@ -194,6 +194,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_qwen2_modeling.py

@@ -137,6 +137,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_rw_modeling.py

@@ -208,6 +208,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,
@@ -326,6 +328,8 @@ def forward(
                 query,
                 torch.select(kv, dim=2, index=0),
                 torch.select(kv, dim=2, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_santacoder_modeling.py

@@ -293,6 +293,8 @@ def forward(
                 query,
                 torch.select(key_value, dim=1, index=0),
                 torch.select(key_value, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,

server/text_generation_server/models/custom_modeling/flash_starcoder2_modeling.py

@@ -242,6 +242,8 @@ def forward(
                 query,
                 torch.select(kv, dim=1, index=0),
                 torch.select(kv, dim=1, index=1),
+                kv_cache[0],
+                kv_cache[1],
                 cu_seqlen_prefill,
                 max_s,
                 self.softmax_scale,