Adding regression benchmark for the transformers SHA 5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,39455.686656,0.0328,30.5,6.53,30.6

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/0/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:27:19,729][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:27:19,730][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:27:20,016][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:27:20,017][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:27:20,017][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:27:20,152][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:27:20,166][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:27:20,167][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 15:28:26,614][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:28:26,615][inference][INFO] - Running inference benchmark
+[2023-08-11 15:28:34,701][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:28:36,016][memory_tracker][INFO] - Peak memory usage: 39455.686656 MB
+[2023-08-11 15:28:36,017][inference][INFO] - 	+ Forward pass peak memory: 39455.686656 (MB)
+[2023-08-11 15:28:36,017][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:28:36,336][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:28:56,621][inference][INFO] - 	+ Forward pass latency: 3.28e-02 (s)
+[2023-08-11 15:28:56,622][inference][INFO] - 	+ Forward pass throughput: 30.50 (samples/s)
+[2023-08-11 15:28:56,623][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:29:03,962][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:29:30,082][inference][INFO] - 	+ Generation pass latency: 6.53e+00 (s)
+[2023-08-11 15:29:30,086][inference][INFO] - 	+ Generation pass throughput: 30.60 (tokens/s)
+[2023-08-11 15:29:30,086][inference][INFO] - Saving inference results
+[2023-08-11 15:29:30,095][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,54148.333567999995,0.0643,15.6,5.71,35.0

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/1/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:29:30,829][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:29:30,830][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:29:31,018][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:29:31,018][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:29:31,019][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:29:31,239][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:29:31,265][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:29:31,266][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 15:29:48,052][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:29:48,053][inference][INFO] - Running inference benchmark
+[2023-08-11 15:29:55,792][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:29:55,888][memory_tracker][INFO] - Peak memory usage: 54148.333567999995 MB
+[2023-08-11 15:29:55,888][inference][INFO] - 	+ Forward pass peak memory: 54148.333567999995 (MB)
+[2023-08-11 15:29:55,888][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:29:58,007][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:31:04,195][inference][INFO] - 	+ Forward pass latency: 6.43e-02 (s)
+[2023-08-11 15:31:04,196][inference][INFO] - 	+ Forward pass throughput: 15.60 (samples/s)
+[2023-08-11 15:31:04,196][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:31:09,917][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:31:32,752][inference][INFO] - 	+ Generation pass latency: 5.71e+00 (s)
+[2023-08-11 15:31:32,755][inference][INFO] - 	+ Generation pass throughput: 35.00 (tokens/s)
+[2023-08-11 15:31:32,756][inference][INFO] - Saving inference results
+[2023-08-11 15:31:32,762][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,40212.758528,0.0324,61.7,6.11,65.5

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/2/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:31:33,257][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:31:33,258][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:31:33,446][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:31:33,446][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:31:33,446][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:31:33,666][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:31:33,689][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:31:33,690][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 15:31:44,143][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:31:44,144][inference][INFO] - Running inference benchmark
+[2023-08-11 15:31:51,874][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:31:51,914][memory_tracker][INFO] - Peak memory usage: 40212.758528 MB
+[2023-08-11 15:31:51,915][inference][INFO] - 	+ Forward pass peak memory: 40212.758528 (MB)
+[2023-08-11 15:31:51,915][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:31:52,395][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:32:22,140][inference][INFO] - 	+ Forward pass latency: 3.24e-02 (s)
+[2023-08-11 15:32:22,141][inference][INFO] - 	+ Forward pass throughput: 61.70 (samples/s)
+[2023-08-11 15:32:22,141][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:32:29,336][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:32:53,786][inference][INFO] - 	+ Generation pass latency: 6.11e+00 (s)
+[2023-08-11 15:32:53,790][inference][INFO] - 	+ Generation pass throughput: 65.50 (tokens/s)
+[2023-08-11 15:32:53,790][inference][INFO] - Saving inference results
+[2023-08-11 15:32:53,796][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,30780.817408,0.115,17.4,8.88,45.0

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/3/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:32:54,284][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:32:54,285][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:32:54,493][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:32:54,493][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:32:54,493][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:32:54,594][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:32:54,617][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:32:54,618][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 15:33:11,911][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:33:11,913][inference][INFO] - Running inference benchmark
+[2023-08-11 15:33:19,823][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:33:19,949][memory_tracker][INFO] - Peak memory usage: 30780.817408 MB
+[2023-08-11 15:33:19,949][inference][INFO] - 	+ Forward pass peak memory: 30780.817408 (MB)
+[2023-08-11 15:33:19,950][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:33:23,777][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:34:34,698][inference][INFO] - 	+ Forward pass latency: 1.15e-01 (s)
+[2023-08-11 15:34:34,698][inference][INFO] - 	+ Forward pass throughput: 17.40 (samples/s)
+[2023-08-11 15:34:34,699][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:34:41,894][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:35:08,550][inference][INFO] - 	+ Generation pass latency: 8.88e+00 (s)
+[2023-08-11 15:35:08,554][inference][INFO] - 	+ Generation pass throughput: 45.00 (tokens/s)
+[2023-08-11 15:35:08,554][inference][INFO] - Saving inference results
+[2023-08-11 15:35:08,560][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,40833.51552,0.0311,129.0,6.12,131.0

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/4/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:35:09,107][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:35:09,107][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:35:09,292][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:35:09,292][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:35:09,292][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:35:09,535][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:35:09,559][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:35:09,560][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 15:35:19,934][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:35:19,936][inference][INFO] - Running inference benchmark
+[2023-08-11 15:35:27,605][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:35:27,650][memory_tracker][INFO] - Peak memory usage: 40833.51552 MB
+[2023-08-11 15:35:27,650][inference][INFO] - 	+ Forward pass peak memory: 40833.51552 (MB)
+[2023-08-11 15:35:27,651][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:35:28,415][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:36:16,906][inference][INFO] - 	+ Forward pass latency: 3.11e-02 (s)
+[2023-08-11 15:36:16,907][inference][INFO] - 	+ Forward pass throughput: 129.00 (samples/s)
+[2023-08-11 15:36:16,907][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:36:23,989][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:36:48,487][inference][INFO] - 	+ Generation pass latency: 6.12e+00 (s)
+[2023-08-11 15:36:48,490][inference][INFO] - 	+ Generation pass throughput: 131.00 (tokens/s)
+[2023-08-11 15:36:48,490][inference][INFO] - Saving inference results
+[2023-08-11 15:36:48,497][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,31481.266175999997,0.215,18.6,7.71,104.0

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/5/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:36:48,997][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:36:48,998][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:36:49,182][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:36:49,182][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:36:49,182][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:36:49,286][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:36:49,312][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:36:49,313][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 15:37:06,400][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:37:06,401][inference][INFO] - Running inference benchmark
+[2023-08-11 15:37:14,292][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:37:14,505][memory_tracker][INFO] - Peak memory usage: 31481.266175999997 MB
+[2023-08-11 15:37:14,505][inference][INFO] - 	+ Forward pass peak memory: 31481.266175999997 (MB)
+[2023-08-11 15:37:14,510][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:37:21,438][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:38:35,969][inference][INFO] - 	+ Forward pass latency: 2.15e-01 (s)
+[2023-08-11 15:38:35,971][inference][INFO] - 	+ Forward pass throughput: 18.60 (samples/s)
+[2023-08-11 15:38:35,972][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:38:43,813][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:39:06,951][inference][INFO] - 	+ Generation pass latency: 7.71e+00 (s)
+[2023-08-11 15:39:06,955][inference][INFO] - 	+ Generation pass throughput: 104.00 (tokens/s)
+[2023-08-11 15:39:06,955][inference][INFO] - Saving inference results
+[2023-08-11 15:39:06,961][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,43849.220096,0.0974,164.0,6.59,486.0

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/6/main.log

@@ -0,0 +1,23 @@
+[2023-08-11 15:39:07,533][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:39:07,534][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:39:07,717][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:39:07,718][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:39:07,718][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:39:07,961][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:39:07,985][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:39:07,986][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float16 on device: cuda
+[2023-08-11 15:39:18,666][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:39:18,668][inference][INFO] - Running inference benchmark
+[2023-08-11 15:39:26,443][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:39:26,554][memory_tracker][INFO] - Peak memory usage: 43849.220096 MB
+[2023-08-11 15:39:26,554][inference][INFO] - 	+ Forward pass peak memory: 43849.220096 (MB)
+[2023-08-11 15:39:26,554][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:39:29,188][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:40:23,336][inference][INFO] - 	+ Forward pass latency: 9.74e-02 (s)
+[2023-08-11 15:40:23,336][inference][INFO] - 	+ Forward pass throughput: 164.00 (samples/s)
+[2023-08-11 15:40:23,337][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:40:30,015][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:40:56,380][inference][INFO] - 	+ Generation pass latency: 6.59e+00 (s)
+[2023-08-11 15:40:56,381][inference][INFO] - 	+ Generation pass throughput: 486.00 (tokens/s)
+[2023-08-11 15:40:56,381][inference][INFO] - Saving inference results
+[2023-08-11 15:40:56,386][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/llama_1gpu_inference/7/main.log

@@ -0,0 +1,20 @@
+[2023-08-11 15:40:57,035][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:40:57,035][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:40:57,232][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type llama
+[2023-08-11 15:40:57,233][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:40:57,233][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:40:57,482][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:40:57,505][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:40:57,506][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: float32 on device: cuda
+[2023-08-11 15:41:15,119][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:41:15,121][inference][INFO] - Running inference benchmark
+[2023-08-11 15:41:23,009][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:41:23,733][memory_tracker][INFO] - Peak memory usage: 35824.467968 MB
+[2023-08-11 15:41:23,734][inference][INFO] - 	+ Forward pass peak memory: 35824.467968 (MB)
+[2023-08-11 15:41:23,748][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:41:54,685][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:43:11,935][inference][INFO] - 	+ Forward pass latency: 1.19e+00 (s)
+[2023-08-11 15:43:11,936][inference][INFO] - 	+ Forward pass throughput: 13.40 (samples/s)
+[2023-08-11 15:43:11,936][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:43:34,896][main][ERROR] - Error during benchmarking: CUDA out of memory. Tried to allocate 96.00 MiB (GPU 0; 79.35 GiB total capacity; 37.73 GiB already allocated; 8.12 MiB free; 39.24 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
+[2023-08-11 15:43:34,897][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/pytorch_bert_inference/0/inference_results.csv

@@ -0,0 +1,2 @@
+,forward.peak_memory(MB),forward.latency(s),forward.throughput(samples/s)
+0,460.13235199999997,0.00355,282.0

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/pytorch_bert_inference/0/main.log

@@ -0,0 +1,20 @@
+[2023-08-11 15:43:39,166][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:43:39,166][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:43:39,372][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForSequenceClassification for task text-classification and model_type bert
+[2023-08-11 15:43:39,372][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:43:39,372][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:43:39,372][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:43:39,374][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:43:39,374][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: None on device: cpu
+[2023-08-11 15:43:39,942][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:43:39,942][inference][INFO] - Running inference benchmark
+[2023-08-11 15:43:40,060][dummy_input][INFO] - Generating dummy input for: ['input_ids', 'attention_mask', 'token_type_ids']
+[2023-08-11 15:43:40,062][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:43:40,107][inference][INFO] - 	+ Forward pass peak memory: 460.13235199999997 (MB)
+[2023-08-11 15:43:40,108][dummy_input][INFO] - Generating dummy input for: ['input_ids', 'attention_mask', 'token_type_ids']
+[2023-08-11 15:43:40,110][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:43:40,147][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:43:50,256][inference][INFO] - 	+ Forward pass latency: 3.55e-03 (s)
+[2023-08-11 15:43:50,257][inference][INFO] - 	+ Forward pass throughput: 282.00 (samples/s)
+[2023-08-11 15:43:50,258][inference][INFO] - Saving inference results
+[2023-08-11 15:43:50,269][backend][INFO] - Cleaning backend

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/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)
+0,463.962112,0.00396,253.0,0.518,193.0

raw_results/2023-08-11_14:26:45_5e5fa0d88c293e6d5be2517b4f45680ba3bb5df2/pytorch_gpt2_inference/0/main.log

@@ -0,0 +1,22 @@
+[2023-08-11 15:43:54,032][benchmark][INFO] - Configuring inference benchmark
+[2023-08-11 15:43:54,033][benchmark][INFO] - 	+ Setting seed(42)
+[2023-08-11 15:43:54,212][pytorch][INFO] - 	+ Infered AutoModel class AutoModelForCausalLM for task text-generation and model_type gpt2
+[2023-08-11 15:43:54,212][backend][INFO] - Configuring pytorch backend
+[2023-08-11 15:43:54,212][backend][INFO] - 	+ Checking initial device isolation
+[2023-08-11 15:43:54,212][backend][INFO] - 	+ Checking contineous device isolation
+[2023-08-11 15:43:54,214][pytorch][INFO] - 	+ Disabling gradients
+[2023-08-11 15:43:54,214][pytorch][INFO] - 	+ Loading pretrained model weights in dtype: None on device: cpu
+[2023-08-11 15:43:54,880][pytorch][INFO] - 	+ Turning on eval mode
+[2023-08-11 15:43:54,881][inference][INFO] - Running inference benchmark
+[2023-08-11 15:43:55,081][inference][INFO] - 	+ Tracking forward pass peak memory
+[2023-08-11 15:43:55,126][inference][INFO] - 	+ Forward pass peak memory: 463.962112 (MB)
+[2023-08-11 15:43:55,127][inference][INFO] - 	+ Warming up the forward pass
+[2023-08-11 15:43:55,158][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2023-08-11 15:44:05,252][inference][INFO] - 	+ Forward pass latency: 3.96e-03 (s)
+[2023-08-11 15:44:05,255][inference][INFO] - 	+ Forward pass throughput: 253.00 (samples/s)
+[2023-08-11 15:44:05,255][inference][INFO] - 	+ Warming up the generation pass
+[2023-08-11 15:44:05,815][inference][INFO] - 	+ Tracking generation latency and throughput
+[2023-08-11 15:44:16,182][inference][INFO] - 	+ Generation pass latency: 5.18e-01 (s)
+[2023-08-11 15:44:16,183][inference][INFO] - 	+ Generation pass throughput: 193.00 (tokens/s)
+[2023-08-11 15:44:16,183][inference][INFO] - Saving inference results
+[2023-08-11 15:44:16,195][backend][INFO] - Cleaning backend