[Docs] Fix warnings in mkdocs build (continued) (#23743)
Signed-off-by: Zerohertz <ohg3417@gmail.com> Signed-off-by: Hyogeun Oh (오효근) <ohg3417@gmail.com>
This commit is contained in:
@@ -525,9 +525,6 @@ class InputBatch:
|
||||
Any consecutive empty indices at the very end of the list are not
|
||||
filled.
|
||||
|
||||
Args:
|
||||
empty_req_indices: empty indices which may be filled.
|
||||
|
||||
Returns:
|
||||
swaps: list of (from,to) swap tuples for moved requests
|
||||
empty_req_indices: indices not filled by condensation
|
||||
|
||||
@@ -2955,7 +2955,7 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
|
||||
Args:
|
||||
kv_cache_config: The KV cache config
|
||||
kv_cache_raw_tensors: The KV cache buffer of each layer, with
|
||||
correct size but uninitialized shape.
|
||||
correct size but uninitialized shape.
|
||||
Returns:
|
||||
Dict[str, torch.Tensor]: A map between layer names to their
|
||||
corresponding memory buffer for KV cache.
|
||||
|
||||
@@ -552,7 +552,7 @@ class TPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
|
||||
return kv_cache_spec
|
||||
|
||||
def _get_slot_mapping_metadata(self, num_reqs,
|
||||
num_scheduled_tokens_per_req):
|
||||
num_scheduled_tokens_per_req) -> np.ndarray:
|
||||
"""
|
||||
Computes metadata for mapping slots to blocks in the key-value (KV)
|
||||
cache for a batch of requests.
|
||||
@@ -565,15 +565,15 @@ class TPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
|
||||
Args:
|
||||
num_reqs (int): Number of requests in the current batch.
|
||||
num_scheduled_tokens_per_req (int or np.ndarray): Number of tokens
|
||||
to be scheduled for each request.
|
||||
to be scheduled for each request.
|
||||
|
||||
Returns:
|
||||
np.ndarray: A 2D array of shape (total_block_len, 3), where each row
|
||||
contains:
|
||||
contains:
|
||||
- kv_cache_start_index (int): The starting index in the KV cache
|
||||
for the corresponding slice.
|
||||
for the corresponding slice.
|
||||
- new_kv_start_index (int): The starting index in the new KV
|
||||
cache for the corresponding slice.
|
||||
cache for the corresponding slice.
|
||||
- slice_len (int): The length of the slice.
|
||||
"""
|
||||
slices_start = self.input_batch.num_computed_tokens_cpu[:num_reqs]
|
||||
|
||||
@@ -172,10 +172,10 @@ def scatter_mm_placeholders(
|
||||
|
||||
Args:
|
||||
embeds: The multimodal embeddings.
|
||||
Shape: `(num_embeds, embed_dim)`
|
||||
Shape: `(num_embeds, embed_dim)`
|
||||
is_embed: A boolean mask indicating which positions in the placeholder
|
||||
tokens need to be filled with multimodal embeddings.
|
||||
Shape: `(num_placeholders, num_embeds)`
|
||||
tokens need to be filled with multimodal embeddings.
|
||||
Shape: `(num_placeholders, num_embeds)`
|
||||
"""
|
||||
if is_embed is None:
|
||||
return embeds
|
||||
@@ -278,7 +278,7 @@ def bind_kv_cache(
|
||||
Args:
|
||||
kv_caches: The allocated kv_caches with layer names as keys.
|
||||
forward_context: The global forward context containing all Attention
|
||||
layers with layer names as keys.
|
||||
layers with layer names as keys.
|
||||
runner_kv_caches: The kv_cache declared by ModelRunner.
|
||||
"""
|
||||
# Bind kv_caches to ModelRunner
|
||||
|
||||
@@ -36,8 +36,8 @@ class WorkerBase(WorkerBaseV0):
|
||||
local_rank: Local device index
|
||||
rank: Global rank in distributed setup
|
||||
distributed_init_method: Distributed initialization method
|
||||
is_driver_worker: Whether this worker handles driver
|
||||
responsibilities
|
||||
is_driver_worker: Whether this worker handles driver
|
||||
responsibilities
|
||||
"""
|
||||
# Configuration storage
|
||||
super().__init__(vllm_config=vllm_config)
|
||||
|
||||
Reference in New Issue
Block a user