Adding regression benchmark for the transformers SHA 0acf56224bd98cbf0413218f362a3a61805a7844

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/0/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float16
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 1
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/0/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,58202.128383999996,0.0319,31.3,7.12,28.1

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/0/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 14:49:41,960][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 14:49:41,960][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 14:49:42,262][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 14:49:42,262][backend][INFO] - Configuring pytorch backend
+[2023-08-11 14:49:42,262][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 14:49:42,525][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 14:49:42,535][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 14:49:42,536][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 14:50:51,092][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 14:50:51,094][inference][INFO] - Running inference benchmark
+[2023-08-11 14:50:59,079][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 14:51:01,978][memory_tracker][INFO] - Peak memory usage: 58202.128383999996 MB
+[2023-08-11 14:51:01,978][inference][INFO] - 	+ Forward pass peak memory: 58202.128383999996 (MB)
+[2023-08-11 14:51:01,978][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 14:51:02,633][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 14:51:45,045][inference][INFO] - 	+ Forward pass latency: 3.19e-02 (s)
+[2023-08-11 14:51:45,046][inference][INFO] - 	+ Forward pass throughput: 31.30 (samples/s)
+[2023-08-11 14:51:45,047][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 14:51:52,813][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 14:52:14,166][inference][INFO] - 	+ Generation pass latency: 7.12e+00 (s)
+[2023-08-11 14:52:14,169][inference][INFO] - 	+ Generation pass throughput: 28.10 (tokens/s)
+[2023-08-11 14:52:14,169][inference][INFO] - Saving inference results
+[2023-08-11 14:52:14,178][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/1/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float32
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 1
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/1/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,72324.34995199999,0.144,6.94,10.6,18.9

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/1/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 14:52:14,640][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 14:52:14,641][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 14:52:14,843][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 14:52:14,843][backend][INFO] - Configuring pytorch backend
+[2023-08-11 14:52:14,843][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 14:52:15,066][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 14:52:15,089][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 14:52:15,089][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 14:52:31,638][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 14:52:31,640][inference][INFO] - Running inference benchmark
+[2023-08-11 14:52:39,294][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 14:52:39,442][memory_tracker][INFO] - Peak memory usage: 72324.34995199999 MB
+[2023-08-11 14:52:39,442][inference][INFO] - 	+ Forward pass peak memory: 72324.34995199999 (MB)
+[2023-08-11 14:52:39,443][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 14:52:44,241][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 14:53:51,968][inference][INFO] - 	+ Forward pass latency: 1.44e-01 (s)
+[2023-08-11 14:53:51,969][inference][INFO] - 	+ Forward pass throughput: 6.94 (samples/s)
+[2023-08-11 14:53:51,969][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 14:54:02,843][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 14:54:24,101][inference][INFO] - 	+ Generation pass latency: 1.06e+01 (s)
+[2023-08-11 14:54:24,104][inference][INFO] - 	+ Generation pass throughput: 18.90 (tokens/s)
+[2023-08-11 14:54:24,104][inference][INFO] - Saving inference results
+[2023-08-11 14:54:24,110][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/2/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float16
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 2
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/2/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,59911.307263999995,0.0412,48.5,7.58,52.8

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/2/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 14:54:24,613][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 14:54:24,613][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 14:54:24,802][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 14:54:24,802][backend][INFO] - Configuring pytorch backend
+[2023-08-11 14:54:24,802][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 14:54:25,025][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 14:54:25,046][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 14:54:25,047][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 14:54:35,688][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 14:54:35,690][inference][INFO] - Running inference benchmark
+[2023-08-11 14:54:43,378][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 14:54:43,428][memory_tracker][INFO] - Peak memory usage: 59911.307263999995 MB
+[2023-08-11 14:54:43,428][inference][INFO] - 	+ Forward pass peak memory: 59911.307263999995 (MB)
+[2023-08-11 14:54:43,428][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 14:54:44,488][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 14:55:34,919][inference][INFO] - 	+ Forward pass latency: 4.12e-02 (s)
+[2023-08-11 14:55:34,920][inference][INFO] - 	+ Forward pass throughput: 48.50 (samples/s)
+[2023-08-11 14:55:34,921][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 14:55:43,264][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 14:56:06,008][inference][INFO] - 	+ Generation pass latency: 7.58e+00 (s)
+[2023-08-11 14:56:06,011][inference][INFO] - 	+ Generation pass throughput: 52.80 (tokens/s)
+[2023-08-11 14:56:06,012][inference][INFO] - Saving inference results
+[2023-08-11 14:56:06,017][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/3/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float32
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 2
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/3/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,30778.720255999997,0.153,13.1,15.3,26.1

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/3/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 14:56:06,501][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 14:56:06,502][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 14:56:06,700][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 14:56:06,700][backend][INFO] - Configuring pytorch backend
+[2023-08-11 14:56:06,700][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 14:56:06,931][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 14:56:06,953][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 14:56:06,953][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 14:56:23,698][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 14:56:23,700][inference][INFO] - Running inference benchmark
+[2023-08-11 14:56:31,594][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 14:56:31,741][memory_tracker][INFO] - Peak memory usage: 30778.720255999997 MB
+[2023-08-11 14:56:31,742][inference][INFO] - 	+ Forward pass peak memory: 30778.720255999997 (MB)
+[2023-08-11 14:56:31,742][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 14:56:35,563][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 14:57:47,315][inference][INFO] - 	+ Forward pass latency: 1.53e-01 (s)
+[2023-08-11 14:57:47,315][inference][INFO] - 	+ Forward pass throughput: 13.10 (samples/s)
+[2023-08-11 14:57:47,316][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 14:58:02,645][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 14:58:33,331][inference][INFO] - 	+ Generation pass latency: 1.53e+01 (s)
+[2023-08-11 14:58:33,334][inference][INFO] - 	+ Generation pass throughput: 26.10 (tokens/s)
+[2023-08-11 14:58:33,334][inference][INFO] - Saving inference results
+[2023-08-11 14:58:33,343][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/4/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float16
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 4
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/4/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,59007.434751999994,0.0574,69.7,8.3,96.4

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/4/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 14:58:33,877][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 14:58:33,878][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 14:58:34,075][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 14:58:34,076][backend][INFO] - Configuring pytorch backend
+[2023-08-11 14:58:34,076][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 14:58:34,310][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 14:58:34,332][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 14:58:34,333][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 14:58:45,266][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 14:58:45,267][inference][INFO] - Running inference benchmark
+[2023-08-11 14:58:52,992][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 14:58:53,052][memory_tracker][INFO] - Peak memory usage: 59007.434751999994 MB
+[2023-08-11 14:58:53,052][inference][INFO] - 	+ Forward pass peak memory: 59007.434751999994 (MB)
+[2023-08-11 14:58:53,053][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 14:58:54,625][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 14:59:51,099][inference][INFO] - 	+ Forward pass latency: 5.74e-02 (s)
+[2023-08-11 14:59:51,100][inference][INFO] - 	+ Forward pass throughput: 69.70 (samples/s)
+[2023-08-11 14:59:51,101][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:00:00,339][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:00:25,244][inference][INFO] - 	+ Generation pass latency: 8.30e+00 (s)
+[2023-08-11 15:00:25,248][inference][INFO] - 	+ Generation pass throughput: 96.40 (tokens/s)
+[2023-08-11 15:00:25,248][inference][INFO] - Saving inference results
+[2023-08-11 15:00:25,253][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/5/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float32
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 4
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/5/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,75262.459904,0.402,9.95,16.6,48.2

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/5/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:00:25,760][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:00:25,761][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:00:25,950][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:00:25,950][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:00:25,950][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:00:26,184][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:00:26,206][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:00:26,207][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 15:00:42,893][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:00:42,895][inference][INFO] - Running inference benchmark
+[2023-08-11 15:00:50,667][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:00:51,091][memory_tracker][INFO] - Peak memory usage: 75262.459904 MB
+[2023-08-11 15:00:51,091][inference][INFO] - 	+ Forward pass peak memory: 75262.459904 (MB)
+[2023-08-11 15:00:51,100][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:01:06,031][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:02:21,973][inference][INFO] - 	+ Forward pass latency: 4.02e-01 (s)
+[2023-08-11 15:02:21,974][inference][INFO] - 	+ Forward pass throughput: 9.95 (samples/s)
+[2023-08-11 15:02:21,975][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:02:38,596][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:03:11,866][inference][INFO] - 	+ Generation pass latency: 1.66e+01 (s)
+[2023-08-11 15:03:11,870][inference][INFO] - 	+ Generation pass throughput: 48.20 (tokens/s)
+[2023-08-11 15:03:11,870][inference][INFO] - Saving inference results
+[2023-08-11 15:03:11,876][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/6/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float16
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 16
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/6/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,63279.333375999995,0.12,133.0,6.44,497.0

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/6/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:03:12,405][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:03:12,405][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:03:12,613][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:03:12,613][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:03:12,613][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:03:12,841][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:03:12,862][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:03:12,863][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 15:03:23,477][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:03:23,479][inference][INFO] - Running inference benchmark
+[2023-08-11 15:03:31,158][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:03:31,382][memory_tracker][INFO] - Peak memory usage: 63279.333375999995 MB
+[2023-08-11 15:03:31,382][inference][INFO] - 	+ Forward pass peak memory: 63279.333375999995 (MB)
+[2023-08-11 15:03:31,382][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:03:37,045][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:04:31,170][inference][INFO] - 	+ Forward pass latency: 1.20e-01 (s)
+[2023-08-11 15:04:31,171][inference][INFO] - 	+ Forward pass throughput: 133.00 (samples/s)
+[2023-08-11 15:04:31,172][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:04:37,857][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:05:03,638][inference][INFO] - 	+ Generation pass latency: 6.44e+00 (s)
+[2023-08-11 15:05:03,641][inference][INFO] - 	+ Generation pass throughput: 497.00 (tokens/s)
+[2023-08-11 15:05:03,642][inference][INFO] - Saving inference results
+[2023-08-11 15:05:03,647][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/7/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: float32
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 20
+  input_shapes:
+    batch_size: 16
+    sequence_length: 200
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 200
+experiment_name: llama_1gpu_inference
+model: togethercomputer/LLaMA-2-7B-32K
+device: cuda
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/7/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,35824.467968,0.684,23.4,13.0,246.0

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/llama_1gpu_inference/7/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:05:04,239][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:05:04,240][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:05:04,590][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:05:04,590][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:05:04,590][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:05:04,697][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:05:04,719][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:05:04,720][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 15:05:21,726][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:05:21,728][inference][INFO] - Running inference benchmark
+[2023-08-11 15:05:29,485][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:05:30,186][memory_tracker][INFO] - Peak memory usage: 35824.467968 MB
+[2023-08-11 15:05:30,186][inference][INFO] - 	+ Forward pass peak memory: 35824.467968 (MB)
+[2023-08-11 15:05:30,202][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:05:55,550][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:07:13,498][inference][INFO] - 	+ Forward pass latency: 6.84e-01 (s)
+[2023-08-11 15:07:13,499][inference][INFO] - 	+ Forward pass throughput: 23.40 (samples/s)
+[2023-08-11 15:07:13,500][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:07:27,210][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:07:53,172][inference][INFO] - 	+ Generation pass latency: 1.30e+01 (s)
+[2023-08-11 15:07:53,175][inference][INFO] - 	+ Generation pass throughput: 246.00 (tokens/s)
+[2023-08-11 15:07:53,175][inference][INFO] - Saving inference results
+[2023-08-11 15:07:53,181][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/pytorch_bert_inference/0/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: null
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 10
+  input_shapes:
+    batch_size: 1
+    sequence_length: 16
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 100
+experiment_name: pytorch_bert_inference
+model: hf-internal-testing/tiny-random-bert
+device: cpu
+task: text-classification
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/pytorch_bert_inference/0/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s)
+0,459.95622399999996,0.00326,307.0

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/pytorch_bert_inference/0/main.log

@@ -0,0 +1,20 @@
+[2023-08-11 15:07:58,180][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:07:58,181][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:07:58,370][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForSequenceClassification for task text-classification and model_type bert
+[2023-08-11 15:07:58,371][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:07:58,371][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:07:58,371][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:07:58,374][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:07:58,374][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: None on device: cpu
+[2023-08-11 15:07:58,989][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:07:58,989][inference][INFO] - Running inference benchmark
+[2023-08-11 15:07:59,113][dummy_input][INFO] - Generating dummy input for: ['input_ids', 'attention_mask', 'token_type_ids']
+[2023-08-11 15:07:59,114][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:07:59,167][inference][INFO] - 	+ Forward pass peak memory: 459.95622399999996 (MB)
+[2023-08-11 15:07:59,168][dummy_input][INFO] - Generating dummy input for: ['input_ids', 'attention_mask', 'token_type_ids']
+[2023-08-11 15:07:59,169][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:07:59,200][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:08:09,313][inference][INFO] - 	+ Forward pass latency: 3.26e-03 (s)
+[2023-08-11 15:08:09,316][inference][INFO] - 	+ Forward pass throughput: 307.00 (samples/s)
+[2023-08-11 15:08:09,316][inference][INFO] - Saving inference results
+[2023-08-11 15:08:09,330][backend][INFO] - Cleaning backend

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/pytorch_gpt2_inference/0/hydra_config.yaml

@@ -0,0 +1,66 @@
+backend:
+  name: pytorch
+  version: 2.0.1+cu117
+  _target_: optimum_benchmark.backends.pytorch.PyTorchBackend
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  initial_isolation_check: true
+  continous_isolation_check: true
+  delete_cache: false
+  no_weights: false
+  torch_dtype: null
+  device_map: null
+  load_in_8bit: false
+  load_in_4bit: false
+  bettertransformer: false
+  torch_compile: false
+  torch_compile_config:
+    fullgraph: false
+    dynamic: false
+    backend: inductor
+    mode: null
+    options: null
+    disable: false
+  amp_autocast: false
+  amp_dtype: null
+  disable_grad: true
+  eval_mode: true
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.InferenceBenchmark
+  seed: 42
+  memory: true
+  warmup_runs: 10
+  benchmark_duration: 10
+  input_shapes:
+    batch_size: 1
+    sequence_length: 16
+    num_choices: 4
+    width: 64
+    height: 64
+    num_channels: 3
+    point_batch_size: 3
+    nb_points_per_image: 2
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 100
+experiment_name: pytorch_gpt2_inference
+model: hf-internal-testing/tiny-random-gpt2
+device: cpu
+task: text-generation
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+environment:
+  optimum_version: 1.11.0
+  transformers_version: 4.32.0.dev0
+  accelerate_version: 0.21.0
+  diffusers_version: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz'
+  cpu_count: 96
+  cpu_ram_mb: 1204539.797504

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/pytorch_gpt2_inference/0/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s),generate.latency(s),generate.throughput(tokens/s)
+0,464.252928,0.00401,249.0,0.537,186.0

raw_results/2023-08-11_13:41:24_0acf56224bd98cbf0413218f362a3a61805a7844/pytorch_gpt2_inference/0/main.log

@@ -0,0 +1,22 @@
+[2023-08-11 15:08:13,953][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:08:13,953][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:08:14,132][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type gpt2
+[2023-08-11 15:08:14,132][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:08:14,132][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:08:14,132][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:08:14,134][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:08:14,134][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: None on device: cpu
+[2023-08-11 15:08:14,889][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:08:14,890][inference][INFO] - Running inference benchmark
+[2023-08-11 15:08:15,092][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:08:15,140][inference][INFO] - 	+ Forward pass peak memory: 464.252928 (MB)
+[2023-08-11 15:08:15,141][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:08:15,174][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:08:25,269][inference][INFO] - 	+ Forward pass latency: 4.01e-03 (s)
+[2023-08-11 15:08:25,271][inference][INFO] - 	+ Forward pass throughput: 249.00 (samples/s)
+[2023-08-11 15:08:25,272][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:08:25,871][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:08:36,073][inference][INFO] - 	+ Generation pass latency: 5.37e-01 (s)
+[2023-08-11 15:08:36,074][inference][INFO] - 	+ Generation pass throughput: 186.00 (tokens/s)
+[2023-08-11 15:08:36,074][inference][INFO] - Saving inference results
+[2023-08-11 15:08:36,088][backend][INFO] - Cleaning backend