[[Misc]Upgrade bitsandbytes to the latest version 0.44.0 (#8768)

2024-09-25 08:08:55 +08:00
parent 1e7d5c01f5
commit 13f9f7a3d0
7 changed files with 44 additions and 34 deletions
--- a/tests/quantization/test_bitsandbytes.py
+++ b/tests/quantization/test_bitsandbytes.py
@@ -107,7 +107,7 @@ def validate_generated_texts(hf_runner,
                     quantization='bitsandbytes',
                     load_format='bitsandbytes',
                     tensor_parallel_size=vllm_tp_size,
-                     enforce_eager=True,
+                     enforce_eager=False,
                     gpu_memory_utilization=0.8) as llm:
        vllm_outputs = llm.generate_greedy(prompts, 8)
        vllm_logs = log_generated_texts(prompts, vllm_outputs, "VllmRunner")