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 16, 03:15 UTC
[Bugfix][V1] Fix test_kv_cache_utils.py (#11738)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Jee Jee Li committed
Jan 5, 2025 at 00:07 UTC
47831430cc943cd470d38d27f8c69a5782795ec3
1 file changed
+5
-5
tests/v1/core/test_kv_cache_utils.py
+5
-5
@@ -147,12 +147,12 @@ def test_generate_block_hash_extra_keys():
147
148
# Test with no extra keys
149
extra_keys, next_mm_idx = generate_block_hash_extra_keys(request, 0, 5, 0)
150
- assert extra_keys == (("hash1", 0), )
150
+ assert extra_keys == ("hash1", )
151
assert next_mm_idx == 1
152
153
# Test with partial overlap
154
extra_keys, next_mm_idx = generate_block_hash_extra_keys(request, 3, 8, 0)
155
- assert extra_keys == (("hash1", 3), )
155
+ assert extra_keys == ("hash1", )
156
assert next_mm_idx == 1
157
158
# Test with no overlap
@@ -162,7 +162,7 @@ def test_generate_block_hash_extra_keys():
162
163
# Test with multiple extra keys
164
extra_keys, next_mm_idx = generate_block_hash_extra_keys(request, 0, 15, 0)
165
- assert extra_keys == (("hash1", 0), ("hash2", 0))
165
+ assert extra_keys == ('hash1', 'hash2')
166
assert next_mm_idx == 2
167
168
@@ -216,11 +216,11 @@ def test_hash_request_tokens():
216
217
# Check the first block
218
assert block_hashes[0].token_ids == (0, 1, 2)
219
- assert block_hashes[0].extra_keys == (("hash1", 0), )
219
+ assert block_hashes[0].extra_keys == ("hash1", )
220
221
# Check the second block
222
assert block_hashes[1].token_ids == (3, 4, 5)
223
- assert block_hashes[1].extra_keys == (("hash2", 0), )
223
+ assert block_hashes[1].extra_keys == ("hash2", )
224
225
226
def test_hash_request_tokens_no_mm_inputs():