diff --git a/.github/workflows/benchmark.yml b/.github/workflows/benchmark.yml index 7469f165..ac109015 100644 --- a/.github/workflows/benchmark.yml +++ b/.github/workflows/benchmark.yml @@ -51,6 +51,7 @@ jobs: with: dotnet-version: | 8.0.x + no-cdn: true - name: Prepare models run: | diff --git a/LLama.Benchmark/LLamaExecutorBenchmark/Prefill.cs b/LLama.Benchmark/LLamaExecutorBenchmark/Prefill.cs index fb087962..2910a9ab 100644 --- a/LLama.Benchmark/LLamaExecutorBenchmark/Prefill.cs +++ b/LLama.Benchmark/LLamaExecutorBenchmark/Prefill.cs @@ -81,13 +81,14 @@ namespace LLama.Benchmark.LLamaExecutorBenchmark ContextSize = PromptAndContextLength.Item2, GpuLayerCount = ModelAndGpuLayerCount.Item2 }; + Console.WriteLine($"************ model params model path: {ModelParams.ModelPath} specified path: {ModelAndGpuLayerCount.Item1}, prefix: {Constants.ModelDir}"); Prompt = File.ReadAllText(Constants.TextCompletionPromptsFilePath).Substring(0, PromptAndContextLength.Item1); InferenceParams = new InferenceParams() { Temperature = 0.6f, MaxTokens = 1 // Only prefill, no generation here. }; - + LLamaWeights weights = LLamaWeights.LoadFromFile(ModelParams); LLamaContext context = weights.CreateContext(ModelParams); Executor = ExecutorType switch diff --git a/LLama.Benchmark/Program.cs b/LLama.Benchmark/Program.cs index 7120a1b8..9ebfb462 100644 --- a/LLama.Benchmark/Program.cs +++ b/LLama.Benchmark/Program.cs @@ -11,7 +11,7 @@ namespace LLama.Benchmark { var modelDir = args[0]; Constants.ModelDir = modelDir; - Console.WriteLine($"#################### model dir: {modelDir}"); + Console.WriteLine($"#################### model dir: {Constants.ModelDir}"); } var summary = BenchmarkRunner.Run(typeof(Program).Assembly);