add more pytorch related tests for torch nightly (vllm-project#17422)

Signed-off-by: Yang Wang <[email protected]>

Author: yangw-dev

.buildkite/test-pipeline.yaml

@@ -293,6 +293,7 @@ steps:
   parallelism: 4
 
 - label: PyTorch Compilation Unit Tests
+  torch_nightly: true
   source_file_dependencies:
     - vllm/
     - tests/compile
@@ -302,6 +303,7 @@ steps:
     - pytest -v -s compile/test_sequence_parallelism.py
 
 - label: PyTorch Fullgraph Smoke Test # 9min
+  torch_nightly: true
   source_file_dependencies:
   - vllm/
   - tests/compile
@@ -312,6 +314,7 @@ steps:
   - pytest -v -s compile/piecewise/test_toy_llama.py
 
 - label: PyTorch Fullgraph Test # 18min
+  torch_nightly: true
   source_file_dependencies:
   - vllm/
   - tests/compile
@@ -436,6 +439,7 @@ steps:
 #####  models test  #####
 
 - label: Basic Models Test # 24min
+  torch_nightly: true
   source_file_dependencies:
   - vllm/
   - tests/models

requirements/nightly_torch_test.txt

@@ -23,5 +23,11 @@ runai-model-streamer-s3==0.11.0
 tensorizer>=2.9.0
 lm-eval==0.4.8
 buildkite-test-collector==0.1.9
-
 lm-eval[api]==0.4.8 # required for model evaluation test
+
+# required for quantization test
+bitsandbytes>=0.45.3
+
+# required for minicpmo_26 test
+vector_quantize_pytorch
+vocos

vllm/sampling_params.py

@@ -186,9 +186,9 @@ class SamplingParams(
         logits_processors: list of functions that modify logits based on
             previously generated tokens, and optionally prompt tokens as
             a first argument.
-        truncate_prompt_tokens: If set to -1, will use the truncation size 
-            supported by the model. If set to an integer k, will use only 
-            the last k tokens from the prompt (i.e., left truncation). 
+        truncate_prompt_tokens: If set to -1, will use the truncation size
+            supported by the model. If set to an integer k, will use only
+            the last k tokens from the prompt (i.e., left truncation).
             Defaults to None (i.e., no truncation).
         guided_decoding: If provided, the engine will construct a guided
             decoding logits processor from these parameters. Defaults to None.