A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 11:00 UTC
[Misc] Remove unused numpy tensor (#18084)
Signed-off-by: Roger Wang <hey@rogerw.me>
Roger Wang committed
May 13, 2025 at 19:33 UTC
6e27c6d86b6d9d7fcc8cb067ca2aa42d2f7f1682
2 files changed
-2
vllm/v1/worker/gpu_model_runner.py
-1
@@ -287,7 +287,6 @@ class GPUModelRunner(LoRAModelRunnerMixin):
287
dtype=torch.int32,
288
device="cpu",
289
pin_memory=self.pin_memory)
290
- self.input_ids_np = self.input_ids_cpu.numpy()
290
self.positions_cpu = torch.zeros(self.max_num_tokens,
291
dtype=torch.int64,
292
device="cpu",
vllm/v1/worker/tpu_model_runner.py
-1
@@ -191,7 +191,6 @@ class TPUModelRunner(LoRAModelRunnerMixin):
191
self.input_ids_cpu = torch.zeros(self.max_num_tokens,
192
dtype=torch.int32,
193
device="cpu")
194
- self.input_ids_np = self.input_ids_cpu.numpy()
194
195
self.positions_cpu = torch.zeros(self.max_num_tokens,
196
dtype=torch.int32,