[None][fix] Display the GPU memory information in GiB unit. (NVIDIA#9070)

nv-guomingz · suyoggupta · commit 4c0837b58d3e · 2025-11-11T22:17:23.000-08:00
Signed-off-by: nv-guomingz &lt;137257613+nv-guomingz@users.noreply.github.com&gt;
diff --git a/tensorrt_llm/_torch/pyexecutor/resource_manager.py b/tensorrt_llm/_torch/pyexecutor/resource_manager.py
@@ -718,7 +718,7 @@ def calculate_max_num_blocks(self,
             if kv_cache_config.free_gpu_memory_fraction is not None:
                 max_tokens = min(kv_cache_config.max_tokens, max_tokens)
                 logger.warning(
-                    f'Both free_gpu_memory_fraction and max_tokens are set (to {free_mem_fraction} and {max_tokens} with free memory {free_mem / (1 << 32)} of total memory {total_mem / (1<<32)}, respectively). The smaller value will be used.'
+                    f'Both free_gpu_memory_fraction and max_tokens are set (to {free_mem_fraction} and {max_tokens} with free memory {free_mem / (1 << 30)}GiB of total memory {total_mem / (1<<30)}GiB, respectively). The smaller value will be used.'
                 )
             else:
                 max_tokens = kv_cache_config.max_tokens

Original file line number	Diff line number	Diff line change
`@@ -718,7 +718,7 @@ def calculate_max_num_blocks(self,`
`718`	`718`	`if kv_cache_config.free_gpu_memory_fraction is not None:`
`719`	`719`	`max_tokens = min(kv_cache_config.max_tokens, max_tokens)`
`720`	`720`	`logger.warning(`
`721`		`- f'Both free_gpu_memory_fraction and max_tokens are set (to {free_mem_fraction} and {max_tokens} with free memory {free_mem / (1 << 32)} of total memory {total_mem / (1<<32)}, respectively). The smaller value will be used.'`
	`721`	`+ f'Both free_gpu_memory_fraction and max_tokens are set (to {free_mem_fraction} and {max_tokens} with free memory {free_mem / (1 << 30)}GiB of total memory {total_mem / (1<<30)}GiB, respectively). The smaller value will be used.'`
`722`	`722`	`)`
`723`	`723`	`else:`
`724`	`724`	`max_tokens = kv_cache_config.max_tokens`