biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

54
tests/v1/worker/test_utils.py
View File

@@ -10,32 +10,28 @@ def test_bind_kv_cache():
from vllm.attention import Attention
ctx = {
'layers.0.self_attn': Attention(32, 128, 0.1),
'layers.1.self_attn': Attention(32, 128, 0.1),
'layers.2.self_attn': Attention(32, 128, 0.1),
'layers.3.self_attn': Attention(32, 128, 0.1),
"layers.0.self_attn": Attention(32, 128, 0.1),
"layers.1.self_attn": Attention(32, 128, 0.1),
"layers.2.self_attn": Attention(32, 128, 0.1),
"layers.3.self_attn": Attention(32, 128, 0.1),
}
kv_cache = {
'layers.0.self_attn': torch.zeros((1, )),
'layers.1.self_attn': torch.zeros((1, )),
'layers.2.self_attn': torch.zeros((1, )),
'layers.3.self_attn': torch.zeros((1, )),
"layers.0.self_attn": torch.zeros((1,)),
"layers.1.self_attn": torch.zeros((1,)),
"layers.2.self_attn": torch.zeros((1,)),
"layers.3.self_attn": torch.zeros((1,)),
}
runner_kv_caches: list[torch.Tensor] = []
bind_kv_cache(kv_cache, ctx, runner_kv_caches)
assert ctx['layers.0.self_attn'].kv_cache[0] is kv_cache[
'layers.0.self_attn']
assert ctx['layers.1.self_attn'].kv_cache[0] is kv_cache[
'layers.1.self_attn']
assert ctx['layers.2.self_attn'].kv_cache[0] is kv_cache[
'layers.2.self_attn']
assert ctx['layers.3.self_attn'].kv_cache[0] is kv_cache[
'layers.3.self_attn']
assert ctx["layers.0.self_attn"].kv_cache[0] is kv_cache["layers.0.self_attn"]
assert ctx["layers.1.self_attn"].kv_cache[0] is kv_cache["layers.1.self_attn"]
assert ctx["layers.2.self_attn"].kv_cache[0] is kv_cache["layers.2.self_attn"]
assert ctx["layers.3.self_attn"].kv_cache[0] is kv_cache["layers.3.self_attn"]
assert runner_kv_caches[0] is kv_cache['layers.0.self_attn']
assert runner_kv_caches[1] is kv_cache['layers.1.self_attn']
assert runner_kv_caches[2] is kv_cache['layers.2.self_attn']
assert runner_kv_caches[3] is kv_cache['layers.3.self_attn']
assert runner_kv_caches[0] is kv_cache["layers.0.self_attn"]
assert runner_kv_caches[1] is kv_cache["layers.1.self_attn"]
assert runner_kv_caches[2] is kv_cache["layers.2.self_attn"]
assert runner_kv_caches[3] is kv_cache["layers.3.self_attn"]
def test_bind_kv_cache_non_attention():
@@ -43,21 +39,19 @@ def test_bind_kv_cache_non_attention():
# example from Jamba PP=2
ctx = {
'model.layers.20.attn': Attention(32, 128, 0.1),
'model.layers.28.attn': Attention(32, 128, 0.1),
"model.layers.20.attn": Attention(32, 128, 0.1),
"model.layers.28.attn": Attention(32, 128, 0.1),
}
kv_cache = {
'model.layers.20.attn': torch.zeros((1, )),
'model.layers.28.attn': torch.zeros((1, )),
"model.layers.20.attn": torch.zeros((1,)),
"model.layers.28.attn": torch.zeros((1,)),
}
runner_kv_caches: list[torch.Tensor] = []
bind_kv_cache(kv_cache, ctx, runner_kv_caches)
assert ctx['model.layers.20.attn'].kv_cache[0] is kv_cache[
'model.layers.20.attn']
assert ctx['model.layers.28.attn'].kv_cache[0] is kv_cache[
'model.layers.28.attn']
assert ctx["model.layers.20.attn"].kv_cache[0] is kv_cache["model.layers.20.attn"]
assert ctx["model.layers.28.attn"].kv_cache[0] is kv_cache["model.layers.28.attn"]
assert runner_kv_caches[0] is kv_cache['model.layers.20.attn']
assert runner_kv_caches[1] is kv_cache['model.layers.28.attn']
assert runner_kv_caches[0] is kv_cache["model.layers.20.attn"]
assert runner_kv_caches[1] is kv_cache["model.layers.28.attn"]