From c26e3c06098146e858b7319844816de42cbe404d Mon Sep 17 00:00:00 2001 From: IlyasMoutawwakil Date: Mon, 5 Feb 2024 15:01:34 +0100 Subject: [PATCH] update examples --- examples/neural_compressor_ptq_bert.yaml | 5 +++-- examples/onnxruntime_static_quant_vit.yaml | 4 ++-- examples/openvino_diffusion.yaml | 6 +++--- examples/openvino_static_quant_bert.yaml | 4 ++-- examples/pytorch_bert.yaml | 7 +++---- examples/pytorch_llama.yaml | 6 ++++-- examples/pytorch_timm.yaml | 6 ++++-- examples/tgi_llama.yaml | 7 +++---- examples/trt_llama.yaml | 6 ++++-- 9 files changed, 28 insertions(+), 23 deletions(-) diff --git a/examples/neural_compressor_ptq_bert.yaml b/examples/neural_compressor_ptq_bert.yaml index 6268fbb0..c19a5723 100644 --- a/examples/neural_compressor_ptq_bert.yaml +++ b/examples/neural_compressor_ptq_bert.yaml @@ -8,13 +8,14 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging experiment_name: openvino_static_quant_bert -model: bert-base-uncased -device: cpu backend: + model: bert-base-uncased + no_weights: true ptq_quantization: true calibration: true + device: cpu benchmark: input_shapes: diff --git a/examples/onnxruntime_static_quant_vit.yaml b/examples/onnxruntime_static_quant_vit.yaml index 478cb1f7..0b06bc0e 100644 --- a/examples/onnxruntime_static_quant_vit.yaml +++ b/examples/onnxruntime_static_quant_vit.yaml @@ -8,14 +8,14 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging experiment_name: onnxruntime_static_quant_vit -model: google/vit-base-patch16-224 -device: cpu backend: + model: google/vit-base-patch16-224 quantization: true quantization_config: is_static: true per_channel: false + device: cpu calibration: true diff --git a/examples/openvino_diffusion.yaml b/examples/openvino_diffusion.yaml index 0c42eb0c..3591ecd7 100644 --- a/examples/openvino_diffusion.yaml +++ b/examples/openvino_diffusion.yaml @@ -8,13 +8,13 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging model: stabilityai/stable-diffusion-2-1 -experiment_name: openvino_diffusion -device: cpu backend: - reshape: true + experiment_name: openvino_diffusion export: true + reshape: true half: true + device: cpu benchmark: input_shapes: diff --git a/examples/openvino_static_quant_bert.yaml b/examples/openvino_static_quant_bert.yaml index ba5f2051..c349f3ea 100644 --- a/examples/openvino_static_quant_bert.yaml +++ b/examples/openvino_static_quant_bert.yaml @@ -8,15 +8,15 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging experiment_name: openvino_static_quant_bert -model: bert-base-uncased -device: cpu backend: + model: bert-base-uncased export: true no_weights: true quantization: true calibration: true reshape: true + device: cpu benchmark: input_shapes: diff --git a/examples/pytorch_bert.yaml b/examples/pytorch_bert.yaml index 41dc9d8c..71a087f0 100644 --- a/examples/pytorch_bert.yaml +++ b/examples/pytorch_bert.yaml @@ -8,11 +8,10 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging experiment_name: pytorch_bert -model: bert-base-uncased -device: cuda -launcher: - device_isolation: true +backend: + model: bert-base-uncased + device: cpu hydra: run: diff --git a/examples/pytorch_llama.yaml b/examples/pytorch_llama.yaml index e5630dee..f6b29792 100644 --- a/examples/pytorch_llama.yaml +++ b/examples/pytorch_llama.yaml @@ -8,8 +8,10 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging experiment_name: pytorch_llama -model: TheBloke/Llama-2-70B-AWQ -device: cuda + +backend: + model: TheBloke/Llama-2-70B-AWQ + device: cuda launcher: device_isolation: true diff --git a/examples/pytorch_timm.yaml b/examples/pytorch_timm.yaml index 154eba5e..03125599 100644 --- a/examples/pytorch_timm.yaml +++ b/examples/pytorch_timm.yaml @@ -7,9 +7,11 @@ defaults: - override hydra/job_logging: colorlog # colorful logging - override hydra/hydra_logging: colorlog # colorful logging -model: timm/mobilenetv3_large_100.ra_in1k experiment_name: pytorch_timm -device: cuda + +backend: + model: timm/mobilenetv3_large_100.ra_in1k + device: cuda launcher: device_isolation: true diff --git a/examples/tgi_llama.yaml b/examples/tgi_llama.yaml index 60204a81..9bf8b4d1 100644 --- a/examples/tgi_llama.yaml +++ b/examples/tgi_llama.yaml @@ -8,13 +8,12 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging experiment_name: tgi_llama -model: TheBloke/Llama-2-70B-AWQ -device: cuda backend: - sharded: false + model: TheBloke/Llama-2-7B-AWQ quantization_scheme: awq - # no_weights: true # wok in progress + sharded: false + device: cuda benchmark: input_shapes: diff --git a/examples/trt_llama.yaml b/examples/trt_llama.yaml index e0f168d0..e3f8844d 100644 --- a/examples/trt_llama.yaml +++ b/examples/trt_llama.yaml @@ -8,8 +8,10 @@ defaults: - override hydra/hydra_logging: colorlog # colorful logging experiment_name: trt_llama -model: NousResearch/Llama-2-7b-hf -device: cuda + +backend: + model: NousResearch/Llama-2-7b-hf + device: cuda benchmark: input_shapes: