xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-10 05:47:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

separate ubatch and normal runs

Browse Source 
Signed-off-by: Sage Moore <sage@neuralmagic.com>
This commit is contained in:
Sage Moore 2025-07-03 17:07:58 +00:00
parent 510e839429
commit bb0645c644
1 changed files with 21 additions and 38 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
vllm/v1/worker/gpu_model_runner.py
View File

@ -1588,38 +1588,20 @@ class GPUModelRunner(LoRAModelRunnerMixin):
return ubatch_metadata return ubatch_metadata
def _run(context,
input_ids,
positions,
inputs_embeds,
intermediate_tensors):
with context:
model_output = self.model(
input_ids=input_ids,
positions=positions,
intermediate_tensors=intermediate_tensors,
inputs_embeds=inputs_embeds,
)
if isinstance(context, UBatchContext):
# Clone before we leave the ubatch context
model_output = model_output.clone()
return model_output
@torch.inference_mode() @torch.inference_mode()
def _ubatch_thread(results, ubatch_metadata): def _ubatch_thread(results, model, ubatch_metadata):
# print(f"Starting Request on ubatch: {ubatch_ctx.id}", flush=True) # print(f"Starting Request on ubatch: {ubatch_ctx.id}", flush=True)
context = ubatch_metadata.context with ubatch_metadata.context:
model_output = _run(context=context, model_output = model(
input_ids=ubatch_metadata.input_ids, input_ids=ubatch_metadata.input_ids,
positions=ubatch_metadata.positions, positions=ubatch_metadata.positions,
inputs_embeds=ubatch_metadata.inputs_embeds, intermediate_tensors=ubatch_metadata.intermediate_tensors,
intermediate_tensors=ubatch_metadata.intermediate_tensors) inputs_embeds=ubatch_metadata.inputs_embeds,
)
results.append((ubatch_metadata.context.id, model_output)) results.append((ubatch_metadata.context.id, model_output))
# print(f"Finishing Request on ubatch: {ubatch_ctx.id}", flush=True) # print(f"Finishing Request on ubatch: {ubatch_ctx.id}", flush=True)
def _run_ubatches(ubatch_metadata) -> torch.Tensor: def _run_ubatches(ubatch_metadata, model) -> torch.Tensor:
results: list[tuple[int, torch.Tensor]] = [] results: list[tuple[int, torch.Tensor]] = []
# Ubatches will manually manage the forward context, so we override # Ubatches will manually manage the forward context, so we override
@ -1630,6 +1612,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
thread = threading.Thread(target=_ubatch_thread, thread = threading.Thread(target=_ubatch_thread,
args=( args=(
results, results,
model,
metadata, metadata,
)) ))
ubatch_threads.append(thread) ubatch_threads.append(thread)
@ -1657,22 +1640,22 @@ class GPUModelRunner(LoRAModelRunnerMixin):
num_tokens_across_dp=num_tokens_across_dp, num_tokens_across_dp=num_tokens_across_dp,
skip_cuda_graphs=skip_cuda_graphs skip_cuda_graphs=skip_cuda_graphs
) )
return _run_ubatches(ubatch_metadata) return _run_ubatches(ubatch_metadata, self.model)
# run normal batch # run normal batch
else: else:
input_ids, positions, inputs_embeds, intermediate_tensors = \ input_ids, positions, inputs_embeds, intermediate_tensors = \
model_inputs(slice(0, num_scheduled_tokens), is_dummy_run) model_inputs(slice(0, num_scheduled_tokens), is_dummy_run)
logger.info(f"NORMAL RUN {num_scheduled_tokens}") logger.info(f"NORMAL RUN {num_scheduled_tokens}")
return _run( with set_forward_context(attn_metadata,
context = set_forward_context(attn_metadata, vllm_config=self.vllm_config,
vllm_config=self.vllm_config, num_tokens=num_scheduled_tokens or 1,
num_tokens=num_scheduled_tokens or 1, num_tokens_across_dp=num_tokens_across_dp,
num_tokens_across_dp=num_tokens_across_dp, skip_cuda_graphs=skip_cuda_graphs):
skip_cuda_graphs=skip_cuda_graphs), return self.model(
input_ids=input_ids, input_ids=input_ids,
positions=positions, positions=positions,
inputs_embeds=inputs_embeds, intermediate_tensors=intermediate_tensors,
intermediate_tensors=intermediate_tensors inputs_embeds=inputs_embeds,
) )
def _pool( def _pool(
self, self,