Adding regression benchmark for the transformers SHA 41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,16195.125247999998,0.0328,30.5,6.49,30.8

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/0/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:07:20,622][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:07:20,623][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:07:20,908][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:07:20,908][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:07:20,908][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:07:21,161][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:07:21,175][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:07:21,176][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 11:08:27,069][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:08:27,071][inference][INFO] - Running inference benchmark
+[2023-08-11 11:08:35,134][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:08:36,418][memory_tracker][INFO] - Peak memory usage: 16195.125247999998 MB
+[2023-08-11 11:08:36,418][inference][INFO] - 	+ Forward pass peak memory: 16195.125247999998 (MB)
+[2023-08-11 11:08:36,419][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:08:36,745][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:08:57,022][inference][INFO] - 	+ Forward pass latency: 3.28e-02 (s)
+[2023-08-11 11:08:57,023][inference][INFO] - 	+ Forward pass throughput: 30.50 (samples/s)
+[2023-08-11 11:08:57,023][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:09:04,430][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:09:30,383][inference][INFO] - 	+ Generation pass latency: 6.49e+00 (s)
+[2023-08-11 11:09:30,387][inference][INFO] - 	+ Generation pass throughput: 30.80 (tokens/s)
+[2023-08-11 11:09:30,387][inference][INFO] - Saving inference results
+[2023-08-11 11:09:30,396][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,52225.245184,0.0655,15.3,5.72,35.0

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/1/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:09:30,898][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:09:30,898][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:09:31,079][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:09:31,079][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:09:31,079][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:09:31,310][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:09:31,345][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:09:31,346][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 11:09:48,320][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:09:48,321][inference][INFO] - Running inference benchmark
+[2023-08-11 11:09:56,313][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:09:56,390][memory_tracker][INFO] - Peak memory usage: 52225.245184 MB
+[2023-08-11 11:09:56,390][inference][INFO] - 	+ Forward pass peak memory: 52225.245184 (MB)
+[2023-08-11 11:09:56,390][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:09:58,507][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:11:04,810][inference][INFO] - 	+ Forward pass latency: 6.55e-02 (s)
+[2023-08-11 11:11:04,811][inference][INFO] - 	+ Forward pass throughput: 15.30 (samples/s)
+[2023-08-11 11:11:04,812][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:11:10,577][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:11:33,455][inference][INFO] - 	+ Generation pass latency: 5.72e+00 (s)
+[2023-08-11 11:11:33,457][inference][INFO] - 	+ Generation pass throughput: 35.00 (tokens/s)
+[2023-08-11 11:11:33,457][inference][INFO] - Saving inference results
+[2023-08-11 11:11:33,464][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,46114.144256,0.0315,63.5,6.18,64.7

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/2/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:11:33,947][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:11:33,948][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:11:34,136][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:11:34,137][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:11:34,137][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:11:34,362][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:11:34,395][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:11:34,396][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 11:11:45,029][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:11:45,030][inference][INFO] - Running inference benchmark
+[2023-08-11 11:11:52,988][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:11:53,027][memory_tracker][INFO] - Peak memory usage: 46114.144256 MB
+[2023-08-11 11:11:53,028][inference][INFO] - 	+ Forward pass peak memory: 46114.144256 (MB)
+[2023-08-11 11:11:53,028][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:11:53,512][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:12:24,017][inference][INFO] - 	+ Forward pass latency: 3.15e-02 (s)
+[2023-08-11 11:12:24,018][inference][INFO] - 	+ Forward pass throughput: 63.50 (samples/s)
+[2023-08-11 11:12:24,019][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:12:31,237][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:12:55,978][inference][INFO] - 	+ Generation pass latency: 6.18e+00 (s)
+[2023-08-11 11:12:55,981][inference][INFO] - 	+ Generation pass throughput: 64.70 (tokens/s)
+[2023-08-11 11:12:55,981][inference][INFO] - Saving inference results
+[2023-08-11 11:12:55,987][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,60511.092736,0.109,18.3,7.04,56.8

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/3/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:12:56,461][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:12:56,462][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:12:56,653][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:12:56,653][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:12:56,653][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:12:56,874][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:12:56,906][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:12:56,907][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 11:13:13,487][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:13:13,489][inference][INFO] - Running inference benchmark
+[2023-08-11 11:13:21,326][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:13:21,451][memory_tracker][INFO] - Peak memory usage: 60511.092736 MB
+[2023-08-11 11:13:21,451][inference][INFO] - 	+ Forward pass peak memory: 60511.092736 (MB)
+[2023-08-11 11:13:21,451][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:13:25,276][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:14:35,896][inference][INFO] - 	+ Forward pass latency: 1.09e-01 (s)
+[2023-08-11 11:14:35,897][inference][INFO] - 	+ Forward pass throughput: 18.30 (samples/s)
+[2023-08-11 11:14:35,898][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:14:42,951][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:15:04,060][inference][INFO] - 	+ Generation pass latency: 7.04e+00 (s)
+[2023-08-11 11:15:04,063][inference][INFO] - 	+ Generation pass throughput: 56.80 (tokens/s)
+[2023-08-11 11:15:04,063][inference][INFO] - Saving inference results
+[2023-08-11 11:15:04,070][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,46732.804096,0.032,125.0,6.26,128.0

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/4/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:15:04,574][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:15:04,575][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:15:04,764][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:15:04,764][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:15:04,764][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:15:04,985][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:15:05,017][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:15:05,018][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 11:15:15,635][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:15:15,637][inference][INFO] - Running inference benchmark
+[2023-08-11 11:15:23,443][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:15:23,491][memory_tracker][INFO] - Peak memory usage: 46732.804096 MB
+[2023-08-11 11:15:23,491][inference][INFO] - 	+ Forward pass peak memory: 46732.804096 (MB)
+[2023-08-11 11:15:23,491][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:15:24,237][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:16:11,292][inference][INFO] - 	+ Forward pass latency: 3.20e-02 (s)
+[2023-08-11 11:16:11,294][inference][INFO] - 	+ Forward pass throughput: 125.00 (samples/s)
+[2023-08-11 11:16:11,294][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:16:18,531][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:16:43,569][inference][INFO] - 	+ Generation pass latency: 6.26e+00 (s)
+[2023-08-11 11:16:43,572][inference][INFO] - 	+ Generation pass throughput: 128.00 (tokens/s)
+[2023-08-11 11:16:43,572][inference][INFO] - Saving inference results
+[2023-08-11 11:16:43,588][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,61213.638655999996,0.187,21.4,7.67,104.0

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/5/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:16:44,166][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:16:44,167][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:16:44,362][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:16:44,362][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:16:44,362][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:16:44,583][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:16:44,616][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:16:44,617][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 11:17:01,200][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:17:01,201][inference][INFO] - Running inference benchmark
+[2023-08-11 11:17:08,942][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:17:09,150][memory_tracker][INFO] - Peak memory usage: 61213.638655999996 MB
+[2023-08-11 11:17:09,150][inference][INFO] - 	+ Forward pass peak memory: 61213.638655999996 (MB)
+[2023-08-11 11:17:09,154][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:17:16,083][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:18:31,399][inference][INFO] - 	+ Forward pass latency: 1.87e-01 (s)
+[2023-08-11 11:18:31,401][inference][INFO] - 	+ Forward pass throughput: 21.40 (samples/s)
+[2023-08-11 11:18:31,402][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:18:39,157][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:19:02,161][inference][INFO] - 	+ Generation pass latency: 7.67e+00 (s)
+[2023-08-11 11:19:02,163][inference][INFO] - 	+ Generation pass throughput: 104.00 (tokens/s)
+[2023-08-11 11:19:02,163][inference][INFO] - Saving inference results
+[2023-08-11 11:19:02,187][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,49230.512127999995,0.0974,164.0,7.21,444.0

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/6/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:19:02,801][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:19:02,802][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:19:03,173][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:19:03,174][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:19:03,174][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:19:03,399][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:19:03,431][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:19:03,432][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 11:19:14,046][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:19:14,048][inference][INFO] - Running inference benchmark
+[2023-08-11 11:19:21,757][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:19:21,867][memory_tracker][INFO] - Peak memory usage: 49230.512127999995 MB
+[2023-08-11 11:19:21,867][inference][INFO] - 	+ Forward pass peak memory: 49230.512127999995 (MB)
+[2023-08-11 11:19:21,868][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:19:24,473][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:20:18,593][inference][INFO] - 	+ Forward pass latency: 9.74e-02 (s)
+[2023-08-11 11:20:18,593][inference][INFO] - 	+ Forward pass throughput: 164.00 (samples/s)
+[2023-08-11 11:20:18,594][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:20:25,647][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:20:47,285][inference][INFO] - 	+ Generation pass latency: 7.21e+00 (s)
+[2023-08-11 11:20:47,289][inference][INFO] - 	+ Generation pass throughput: 444.00 (tokens/s)
+[2023-08-11 11:20:47,289][inference][INFO] - Saving inference results
+[2023-08-11 11:20:47,294][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,65556.840448,0.685,23.4,15.3,209.0

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/llama_1gpu_inference/7/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 11:20:47,963][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:20:47,964][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:20:48,155][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 11:20:48,156][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:20:48,156][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:20:48,379][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:20:48,416][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:20:48,417][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 11:21:05,840][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:21:05,842][inference][INFO] - Running inference benchmark
+[2023-08-11 11:21:13,793][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:21:14,497][memory_tracker][INFO] - Peak memory usage: 65556.840448 MB
+[2023-08-11 11:21:14,498][inference][INFO] - 	+ Forward pass peak memory: 65556.840448 (MB)
+[2023-08-11 11:21:14,514][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:21:39,905][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:22:57,998][inference][INFO] - 	+ Forward pass latency: 6.85e-01 (s)
+[2023-08-11 11:22:58,000][inference][INFO] - 	+ Forward pass throughput: 23.40 (samples/s)
+[2023-08-11 11:22:58,000][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:23:11,732][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:23:42,342][inference][INFO] - 	+ Generation pass latency: 1.53e+01 (s)
+[2023-08-11 11:23:42,346][inference][INFO] - 	+ Generation pass throughput: 209.00 (tokens/s)
+[2023-08-11 11:23:42,346][inference][INFO] - Saving inference results
+[2023-08-11 11:23:42,353][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/pytorch_bert_inference/0/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s)
+0,460.079104,0.00352,284.0

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/pytorch_bert_inference/0/main.log

@@ -0,0 +1,20 @@
+[2023-08-11 11:23:46,958][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:23:46,959][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:23:47,143][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForSequenceClassification for task text-classification and model_type bert
+[2023-08-11 11:23:47,143][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:23:47,143][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:23:47,143][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:23:47,145][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:23:47,145][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: None on device: cpu
+[2023-08-11 11:23:47,741][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:23:47,742][inference][INFO] - Running inference benchmark
+[2023-08-11 11:23:47,868][dummy_input][INFO] - Generating dummy input for: ['input_ids', 'attention_mask', 'token_type_ids']
+[2023-08-11 11:23:47,870][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:23:47,919][inference][INFO] - 	+ Forward pass peak memory: 460.079104 (MB)
+[2023-08-11 11:23:47,920][dummy_input][INFO] - Generating dummy input for: ['input_ids', 'attention_mask', 'token_type_ids']
+[2023-08-11 11:23:47,922][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:23:47,958][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:23:58,072][inference][INFO] - 	+ Forward pass latency: 3.52e-03 (s)
+[2023-08-11 11:23:58,075][inference][INFO] - 	+ Forward pass throughput: 284.00 (samples/s)
+[2023-08-11 11:23:58,075][inference][INFO] - Saving inference results
+[2023-08-11 11:23:58,090][backend][INFO] - Cleaning backend

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/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)
+0,463.736832,0.00398,251.0,0.505,198.0

raw_results/2023-08-11_10:30:18_41d56ea6dd972a6a7e12c2a2228e491a5ec18a5f/pytorch_gpt2_inference/0/main.log

@@ -0,0 +1,22 @@
+[2023-08-11 11:24:01,957][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 11:24:01,958][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 11:24:02,141][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type gpt2
+[2023-08-11 11:24:02,141][backend][INFO] - Configuring pytorch backend
+[2023-08-11 11:24:02,141][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 11:24:02,141][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 11:24:02,142][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 11:24:02,143][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: None on device: cpu
+[2023-08-11 11:24:02,774][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 11:24:02,774][inference][INFO] - Running inference benchmark
+[2023-08-11 11:24:02,971][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 11:24:03,014][inference][INFO] - 	+ Forward pass peak memory: 463.736832 (MB)
+[2023-08-11 11:24:03,016][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 11:24:03,051][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 11:24:13,146][inference][INFO] - 	+ Forward pass latency: 3.98e-03 (s)
+[2023-08-11 11:24:13,148][inference][INFO] - 	+ Forward pass throughput: 251.00 (samples/s)
+[2023-08-11 11:24:13,149][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 11:24:13,723][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 11:24:23,819][inference][INFO] - 	+ Generation pass latency: 5.05e-01 (s)
+[2023-08-11 11:24:23,820][inference][INFO] - 	+ Generation pass throughput: 198.00 (tokens/s)
+[2023-08-11 11:24:23,820][inference][INFO] - Saving inference results
+[2023-08-11 11:24:23,834][backend][INFO] - Cleaning backend