[Bugfix][Hybrid][NemotronH] Fix mamba_cache_mode=all + speculative decoding crash (#41233)

Signed-off-by: Roi Koren <roik@nvidia.com>
This commit is contained in:
roikoren755
2026-05-18 14:54:00 +03:00
committed by GitHub
parent e414e1f1c0
commit 737bfa3a43
10 changed files with 568 additions and 117 deletions
+41 -11
View File
@@ -888,6 +888,11 @@ class GPUModelRunner(
self.kv_connector_output: KVConnectorOutput | None = None
self.mamba_state_idx: dict[str, int] = {}
self._mamba_copy_bufs: mamba_utils.MambaCopyBuffers | None = None
self.mamba_prev_last_scheduled_idx: CpuGpuBuffer | None = None
if self.cache_config.mamba_cache_mode == "all" and self.num_spec_tokens > 0:
self.mamba_prev_last_scheduled_idx = self._make_buffer(
self.max_num_reqs, dtype=torch.int32
)
self.layerwise_nvtx_hooks_registered = False
def update_max_model_len(self, max_model_len: int) -> None:
@@ -1493,21 +1498,12 @@ class GPUModelRunner(
num_reqs = output_token_ids.size(0)
self.num_accepted_tokens.gpu[:num_reqs] = (output_token_ids != -1).sum(dim=1)
if self.cache_config.mamba_cache_mode == "align":
is_align = self.cache_config.mamba_cache_mode == "align"
if is_align:
for i, num_tokens in enumerate(
self.num_accepted_tokens.gpu[:num_reqs].cpu().numpy()
):
self.input_batch.num_accepted_tokens_cpu[i] = num_tokens
mamba_utils.postprocess_mamba(
scheduler_output,
self.kv_cache_config,
self.input_batch,
self.requests,
self.mamba_state_idx,
self.compilation_config.static_forward_context,
self.model.get_mamba_state_copy_func(),
self._get_mamba_copy_bufs(),
)
else:
self.input_batch.num_accepted_tokens_cpu_tensor[:num_reqs].copy_(
self.num_accepted_tokens.gpu[:num_reqs], non_blocking=True
@@ -1515,6 +1511,24 @@ class GPUModelRunner(
assert self.num_accepted_tokens_event is not None
self.num_accepted_tokens_event.record()
mamba_utils.postprocess_mamba(
scheduler_output,
self.kv_cache_config,
self.cache_config,
self.input_batch,
self.requests,
self.mamba_state_idx,
self.num_spec_tokens,
num_reqs,
forward_context=(
self.compilation_config.static_forward_context if is_align else None
),
mamba_state_copy_funcs=(
self.model.get_mamba_state_copy_func() if is_align else None
),
copy_bufs=self._get_mamba_copy_bufs() if is_align else None,
)
def _update_streaming_request(
self, req_id: str, new_req_data: NewRequestData
) -> CachedRequestState:
@@ -2012,6 +2026,15 @@ class GPUModelRunner(
self.num_accepted_tokens.np.fill(1)
self.num_accepted_tokens.gpu.fill_(1)
if self.mamba_prev_last_scheduled_idx is not None:
mamba_utils.preprocess_mamba_all_specdec(
scheduler_output,
self.input_batch,
self.mamba_state_idx,
num_reqs,
self.mamba_prev_last_scheduled_idx,
)
# Update num_computed_tokens on GPU. In async spec decode,
# CPU values are optimistic (all drafts accepted). The kernel
# corrects on GPU using the previous step's
@@ -2319,6 +2342,13 @@ class GPUModelRunner(
:num_reqs_padded
],
)
if (
isinstance(builder, Mamba2AttentionMetadataBuilder)
and self.mamba_prev_last_scheduled_idx is not None
):
extra_attn_metadata_args["prev_last_scheduled_idx"] = (
self.mamba_prev_last_scheduled_idx.gpu[:num_reqs_padded]
)
if for_cudagraph_capture:
attn_metadata_i = builder.build_for_cudagraph_capture(
+108 -51
View File
@@ -144,6 +144,24 @@ def do_mamba_copy_block(copy_bufs: MambaCopyBuffers):
)
def cleanup_mamba_state_idx(
scheduler_output: SchedulerOutput,
mamba_state_idx: dict[str, int],
) -> None:
"""Pop stale `mamba_state_idx` entries for finished/preempted/resumed reqs.
Force-preempted requests (e.g., during reset_prefix_cache / KV cache
flush) appear in resumed_req_ids without a corresponding entry in
preempted_req_ids, leaving stale entries that can point to block
indices beyond the new (smaller) block allocation.
"""
finished_req_ids = scheduler_output.finished_req_ids
preempted_req_ids = scheduler_output.preempted_req_ids or set()
resumed_req_ids = scheduler_output.scheduled_cached_reqs.resumed_req_ids
for req_id in itertools.chain(finished_req_ids, preempted_req_ids, resumed_req_ids):
mamba_state_idx.pop(req_id, None)
def preprocess_mamba(
scheduler_output: SchedulerOutput,
kv_cache_config: KVCacheConfig,
@@ -165,16 +183,7 @@ def preprocess_mamba(
# TODO(Chen): we need to optimize this function a lot
assert cache_config.enable_prefix_caching
block_size = mamba_spec.block_size
finished_req_ids = scheduler_output.finished_req_ids
preempted_req_ids = scheduler_output.preempted_req_ids or set()
# We need to clear mamba_state_idx for resumed requests. When requests are
# force-preempted (e.g., during reset_prefix_cache / KV cache flush),
# they appear in resumed_req_ids without a corresponding entry in
# preempted_req_ids, leaving stale mamba_state_idx entries that can
# point to block indices beyond the new (smaller) block allocation.
resumed_req_ids = scheduler_output.scheduled_cached_reqs.resumed_req_ids
for req_id in itertools.chain(finished_req_ids, preempted_req_ids, resumed_req_ids):
mamba_state_idx.pop(req_id, None)
cleanup_mamba_state_idx(scheduler_output, mamba_state_idx)
copy_bufs.offset = 0
for i, req_id in enumerate(input_batch.req_ids):
@@ -222,52 +231,100 @@ def preprocess_mamba(
def postprocess_mamba(
scheduler_output: SchedulerOutput,
kv_cache_config: KVCacheConfig,
cache_config: CacheConfig,
input_batch: GPUInputBatch,
requests: dict[str, CachedRequestState],
mamba_state_idx: dict[str, int],
forward_context: dict[str, Any],
mamba_state_copy_funcs: tuple[MambaStateCopyFunc, ...],
copy_bufs: MambaCopyBuffers,
num_spec_tokens: int,
num_reqs: int,
*,
forward_context: dict[str, Any] | None = None,
mamba_state_copy_funcs: tuple[MambaStateCopyFunc, ...] | None = None,
copy_bufs: MambaCopyBuffers | None = None,
):
"""
If a blocks is converted from partial block to full block in this step, copy the
state from the block for running state to the new full block.
Post-model-execute mamba prefix-caching bookkeeping. Dispatched by
cache_config.mamba_cache_mode:
- "align": if a block is converted from partial to full this step,
copy the running state into the new full block.
- "all" + num_spec_tokens > 0: record per-request the block index of
the last token scheduled this step, so the next step can anchor
its in-place writes when accepted drafts leave the sequence at a
non-block-aligned position.
"""
num_scheduled_tokens_dict = scheduler_output.num_scheduled_tokens
scheduled_spec_decode_tokens_dict = scheduler_output.scheduled_spec_decode_tokens
num_accepted_tokens_cpu = input_batch.num_accepted_tokens_cpu
mamba_group_ids = copy_bufs.mamba_group_ids
mamba_spec = copy_bufs.mamba_spec
copy_bufs.offset = 0
for i, req_id in enumerate(input_batch.req_ids):
req_state = requests[req_id]
num_computed_tokens = req_state.num_computed_tokens
num_draft_tokens = len(scheduled_spec_decode_tokens_dict.get(req_id, []))
num_scheduled_tokens = num_scheduled_tokens_dict[req_id]
num_accepted_tokens = num_accepted_tokens_cpu[i]
num_tokens_running_state = (
num_computed_tokens + num_scheduled_tokens - num_draft_tokens
if cache_config.mamba_cache_mode == "align":
assert forward_context is not None
assert mamba_state_copy_funcs is not None
assert copy_bufs is not None
num_scheduled_tokens_dict = scheduler_output.num_scheduled_tokens
scheduled_spec_decode_tokens_dict = (
scheduler_output.scheduled_spec_decode_tokens
)
new_num_computed_tokens = num_tokens_running_state + num_accepted_tokens - 1
aligned_new_computed_tokens = (
new_num_computed_tokens // mamba_spec.block_size * mamba_spec.block_size
)
# TODO: how to ensure all blocks that cache_blocks called are cached here?
if aligned_new_computed_tokens >= num_tokens_running_state:
accept_token_bias = aligned_new_computed_tokens - num_tokens_running_state
src_block_idx = mamba_state_idx[req_id]
dest_block_idx = aligned_new_computed_tokens // mamba_spec.block_size - 1
collect_mamba_copy_meta(
copy_bufs,
kv_cache_config,
mamba_state_copy_funcs,
mamba_group_ids,
src_block_idx,
dest_block_idx,
accept_token_bias,
req_state,
forward_context,
num_accepted_tokens_cpu = input_batch.num_accepted_tokens_cpu
mamba_group_ids = copy_bufs.mamba_group_ids
mamba_spec = copy_bufs.mamba_spec
copy_bufs.offset = 0
for i, req_id in enumerate(input_batch.req_ids):
req_state = requests[req_id]
num_computed_tokens = req_state.num_computed_tokens
num_draft_tokens = len(scheduled_spec_decode_tokens_dict.get(req_id, []))
num_scheduled_tokens = num_scheduled_tokens_dict[req_id]
num_accepted_tokens = num_accepted_tokens_cpu[i]
num_tokens_running_state = (
num_computed_tokens + num_scheduled_tokens - num_draft_tokens
)
if src_block_idx == dest_block_idx:
num_accepted_tokens_cpu[i] = 1
do_mamba_copy_block(copy_bufs)
new_num_computed_tokens = num_tokens_running_state + num_accepted_tokens - 1
aligned_new_computed_tokens = (
new_num_computed_tokens // mamba_spec.block_size * mamba_spec.block_size
)
# TODO: how to ensure all blocks that cache_blocks called are cached here?
if aligned_new_computed_tokens >= num_tokens_running_state:
accept_token_bias = (
aligned_new_computed_tokens - num_tokens_running_state
)
src_block_idx = mamba_state_idx[req_id]
dest_block_idx = (
aligned_new_computed_tokens // mamba_spec.block_size - 1
)
collect_mamba_copy_meta(
copy_bufs,
kv_cache_config,
mamba_state_copy_funcs,
mamba_group_ids,
src_block_idx,
dest_block_idx,
accept_token_bias,
req_state,
forward_context,
)
if src_block_idx == dest_block_idx:
num_accepted_tokens_cpu[i] = 1
do_mamba_copy_block(copy_bufs)
elif cache_config.mamba_cache_mode == "all" and num_spec_tokens > 0:
_, mamba_spec = get_mamba_groups(kv_cache_config)
block_size = mamba_spec.block_size
full_decode_len = 1 + num_spec_tokens
scheduled = scheduler_output.num_scheduled_tokens
for req_id in input_batch.req_ids[:num_reqs]:
num_query = scheduled.get(req_id, 0)
if num_query == full_decode_len:
req = requests[req_id]
seq_len = req.num_computed_tokens + num_query
mamba_state_idx[req_id] = max(0, (seq_len - 1) // block_size)
else:
mamba_state_idx.pop(req_id, None)
def preprocess_mamba_all_specdec(
scheduler_output: SchedulerOutput,
input_batch: GPUInputBatch,
mamba_state_idx: dict[str, int],
num_reqs: int,
prev_last_scheduled_idx_buf: CpuGpuBuffer,
) -> None:
cleanup_mamba_state_idx(scheduler_output, mamba_state_idx)
np_view = prev_last_scheduled_idx_buf.np
for i, req_id in enumerate(input_batch.req_ids[:num_reqs]):
np_view[i] = mamba_state_idx.get(req_id, -1)
np_view[num_reqs:].fill(-1)
prev_last_scheduled_idx_buf.copy_to_gpu()