[Mini-VLLM] Remove LoRA, plugins, speech-to-text and scale-out surfaces
Production runs one static model behind the OpenAI-compatible API. Every adapter path and non-OpenAI HTTP surface is unreachable there. - Delete the LoRA stack: vllm/lora, LoRARequest on the engine wire, the scheduler, worker, cudagraph and MoE plumbing, LoRA metrics, the lora CLI group, and the LoRA tests. - Drop the plugin loader in vllm/plugins. Platform resolution is now CUDA-only and the general/stat-logger plugin hooks are gone. - Delete the speech-to-text and realtime entrypoints, the scale-out render routes, and vllm/config/speech_to_text.py. GenerationTask is generate-only. - Cut DecodeBenchConnector and the disagg-prefill doc and image. The Multi, Offloading and Example connectors stay; production uses the CPU KV offload tier. - Delete the ROCm-only fused KDA chunk kernel (2346 lines) with its CMake arch gate, declarations and op registrations. The CUDA build never defined the guard and no Python caller existed. - Delete the orphan Mistral and Pixtral chat fixtures and examples/. - Fix _check_model: removing the LoRA early return made every request fail with "The model ... does not exist. Valid aliases: ...". Verified on an RTX 3090: import sweep reports 1242 modules with 7 known optional-dependency failures; a live server answers chat, completions, responses, tokenize and detokenize with zero log errors and 404 for the removed surfaces; collection yields 8558 tests with the 3 pre-existing errors; ruff adds no new finding; the incremental _C build succeeds. Co-authored-by: Hinata <hinata@idihore.id> Signed-off-by: MOVZX <movzx@yahoo.com>
This commit is contained in:
@@ -9,7 +9,6 @@ import numpy as np
|
||||
import torch
|
||||
|
||||
from vllm.config.reasoning import ReasoningConfig
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.multimodal.inputs import MultiModalFeatureSpec
|
||||
from vllm.pooling_params import PoolingParams
|
||||
from vllm.sampling_params import SamplingParams, SamplingType
|
||||
@@ -46,7 +45,6 @@ class CachedRequestState:
|
||||
mrope_positions: torch.Tensor | None = None
|
||||
mrope_position_delta: int | None = None
|
||||
|
||||
lora_request: LoRARequest | None = None
|
||||
prompt_embeds: torch.Tensor | None = None
|
||||
# To accumulate prompt logprobs tensor chunks across prefill steps.
|
||||
in_progress_prompt_logprobs_cpu: LogprobsTensors | None = None
|
||||
@@ -252,11 +250,6 @@ class InputBatch:
|
||||
)
|
||||
self.num_accepted_tokens_cpu = self.num_accepted_tokens_cpu_tensor.numpy()
|
||||
|
||||
# lora related
|
||||
self.request_lora_mapping = np.zeros((self.max_num_reqs,), dtype=np.int64)
|
||||
self.lora_id_to_request_ids: dict[int, set[str]] = {}
|
||||
self.lora_id_to_lora_request: dict[int, LoRARequest] = {}
|
||||
|
||||
# req_index -> generator
|
||||
# NOTE(woosuk): The indices of the requests that do not have their own
|
||||
# generator should not be included in the dictionary.
|
||||
@@ -478,19 +471,6 @@ class InputBatch:
|
||||
# Speculative decoding: by default 1 token is generated.
|
||||
self.num_accepted_tokens_cpu[req_index] = 1
|
||||
|
||||
# Add request lora ID
|
||||
if request.lora_request:
|
||||
lora_id = request.lora_request.lora_int_id
|
||||
if lora_id not in self.lora_id_to_request_ids:
|
||||
self.lora_id_to_request_ids[lora_id] = set()
|
||||
|
||||
self.request_lora_mapping[req_index] = lora_id
|
||||
self.lora_id_to_request_ids[lora_id].add(request.req_id)
|
||||
self.lora_id_to_lora_request[lora_id] = request.lora_request
|
||||
else:
|
||||
# No LoRA
|
||||
self.request_lora_mapping[req_index] = 0
|
||||
|
||||
return req_index
|
||||
|
||||
def update_req_spec_token_ids(
|
||||
@@ -541,16 +521,6 @@ class InputBatch:
|
||||
self.block_table.clear_row(req_index)
|
||||
self.req_prompt_embeds.pop(req_index, None)
|
||||
|
||||
# LoRA
|
||||
lora_id = self.request_lora_mapping[req_index]
|
||||
if lora_id != 0:
|
||||
lora_req_ids = self.lora_id_to_request_ids[lora_id]
|
||||
lora_req_ids.discard(req_id)
|
||||
if not lora_req_ids:
|
||||
del self.lora_id_to_request_ids[lora_id]
|
||||
del self.lora_id_to_lora_request[lora_id]
|
||||
self.request_lora_mapping[req_index] = 0
|
||||
|
||||
if self.is_pooling_model:
|
||||
self.pooling_params.pop(req_id, None)
|
||||
self.pooling_states.pop(req_id, None)
|
||||
@@ -646,11 +616,6 @@ class InputBatch:
|
||||
|
||||
self.block_table.swap_row(i1, i2)
|
||||
|
||||
self.request_lora_mapping[i1], self.request_lora_mapping[i2] = (
|
||||
self.request_lora_mapping[i2],
|
||||
self.request_lora_mapping[i1],
|
||||
)
|
||||
|
||||
if self.is_pooling_model:
|
||||
# Sampling and logits parameters don't apply to pooling models.
|
||||
return
|
||||
@@ -776,10 +741,6 @@ class InputBatch:
|
||||
]
|
||||
self.block_table.move_row(last_req_index, empty_index)
|
||||
|
||||
self.request_lora_mapping[empty_index] = self.request_lora_mapping[
|
||||
last_req_index
|
||||
]
|
||||
|
||||
if self.is_pooling_model:
|
||||
last_req_index -= 1
|
||||
# Sampling state not used by pooling models.
|
||||
@@ -993,32 +954,6 @@ class InputBatch:
|
||||
prompt_token_ids[i, self.num_prompt_tokens[i] :] = self.vocab_size
|
||||
return prompt_token_ids_cpu_tensor
|
||||
|
||||
def make_lora_inputs(
|
||||
self, num_scheduled_tokens: np.ndarray, num_sampled_tokens: np.ndarray
|
||||
) -> tuple[tuple[int, ...], tuple[int, ...], set[LoRARequest]]:
|
||||
"""Given the num_scheduled_tokens for each request in the batch, return
|
||||
datastructures used to activate the current LoRAs.
|
||||
|
||||
Returns:
|
||||
1. prompt_lora_mapping: A tuple of size np.sum(num_sampled_tokens)
|
||||
where, prompt_lora_mapping[i] is the LoRA id to use for the ith
|
||||
sampled token.
|
||||
2. token_lora_mapping: A tuple of size np.sum(num_scheduled_tokens)
|
||||
where, token_lora_mapping[i] is the LoRA id to use for ith
|
||||
token.
|
||||
3. lora_requests: Set of relevant LoRA requests.
|
||||
|
||||
"""
|
||||
req_lora_mapping = self.request_lora_mapping[: self.num_reqs]
|
||||
prompt_lora_mapping = tuple(req_lora_mapping.repeat(num_sampled_tokens))
|
||||
token_lora_mapping = tuple(req_lora_mapping.repeat(num_scheduled_tokens))
|
||||
|
||||
active_lora_requests: set[LoRARequest] = set(
|
||||
self.lora_id_to_lora_request.values()
|
||||
)
|
||||
|
||||
return prompt_lora_mapping, token_lora_mapping, active_lora_requests
|
||||
|
||||
def set_async_sampled_token_ids(
|
||||
self,
|
||||
sampled_token_ids_cpu: torch.Tensor,
|
||||
|
||||
Reference in New Issue
Block a user