Update dataset

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/.hydra/config.yaml

@@ -0,0 +1,77 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase-1b
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase-1b
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase-1b
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+bfloat16
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- model=bigcode/starcoderbase-1b

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/cli.log

@@ -0,0 +1,73 @@
+[2024-02-14 19:26:44,460][launcher][INFO] - Configuring process launcher
+[2024-02-14 19:26:44,460][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 19:26:44,462][process][INFO] - 	+ Launched worker process with PID 1953909.
+[2024-02-14 19:26:44,463][isolation][INFO] - 	+ Launched device(s) isolation process 1953910.
+[2024-02-14 19:26:50,764][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2024-02-14 19:26:50,764][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2024-02-14 19:26:50,765][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2024-02-14 19:26:50,875][datasets][INFO] - PyTorch version 2.1.2+cu118 available.
+[2024-02-14 19:26:51,986][backend][INFO] - Configuring pytorch backend
+[2024-02-14 19:26:51,986][pytorch][INFO] - 	+ Inferred class AutoModelForCausalLM for task text-generation and model_type gpt_bigcode
+[2024-02-14 19:26:51,986][pytorch][INFO] - 	+ Disabling gradients
+[2024-02-14 19:26:51,987][pytorch][INFO] - 	+ Loading model with no weights
+[2024-02-14 19:26:51,987][pytorch][INFO] - 	+ Creating no weights model directory
+[2024-02-14 19:26:51,987][pytorch][INFO] - 	+ Saving pretrained config
+[2024-02-14 19:26:51,988][pytorch][INFO] - 	+ Creating no weights model
+[2024-02-14 19:26:51,992][pytorch][INFO] - 	+ Saving no weights model
+[2024-02-14 19:26:51,993][pytorch][INFO] - 	+ Loading no weights model
+[2024-02-14 19:26:51,993][pytorch][INFO] - 	+ Loading model directly on device: cuda
+[2024-02-14 19:26:53,977][pytorch][INFO] - 	+ Randomizing model weights
+[2024-02-14 19:26:53,981][pytorch][INFO] - 	+ Tying model weights after randomization
+[2024-02-14 19:26:53,982][pytorch][INFO] - 	+ Turning on model's eval mode
+[2024-02-14 19:26:54,091][benchmark][INFO] - Configuring inference benchmark
+[2024-02-14 19:26:54,091][inference][INFO] - Running inference benchmark
+[2024-02-14 19:26:54,091][inference][INFO] - 	+ Updating input shapes with model shapes
+[2024-02-14 19:26:54,091][inference][INFO] - 	+ Preparing backend for inference
+[2024-02-14 19:26:54,091][inference][INFO] - 	+ Creating input generator
+[2024-02-14 19:26:54,091][input-generator][INFO] - Using text-generation task generator
+[2024-02-14 19:26:54,092][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 19:26:54,092][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:26:54,092][inference][INFO] - 	+ Tracking forward pass peak memory
+[2024-02-14 19:26:54,092][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 19:26:54,092][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 19:26:55,968][inference][INFO] - 	+ Forward pass max memory used: 3930 (MB)
+[2024-02-14 19:26:55,968][inference][INFO] - 	+ Forward pass max memory reserved: 2472 (MB)
+[2024-02-14 19:26:55,969][inference][INFO] - 	+ Forward pass max memory allocated: 2414 (MB)
+[2024-02-14 19:26:55,969][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 19:26:55,969][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:26:55,969][inference][INFO] - 	+ Tracking generation pass peak memory
+[2024-02-14 19:26:55,969][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 19:26:55,969][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 19:27:01,723][inference][INFO] - 	+ Generation pass max memory used: 3938 (MB)
+[2024-02-14 19:27:01,723][inference][INFO] - 	+ Generation pass max memory reserved: 2478 (MB)
+[2024-02-14 19:27:01,723][inference][INFO] - 	+ Generation pass max memory allocated: 2417 (MB)
+[2024-02-14 19:27:01,723][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 19:27:01,723][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:27:01,724][inference][INFO] - 	+ Warming up the forward pass
+[2024-02-14 19:27:01,830][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2024-02-14 19:27:11,860][inference][INFO] - 	+ Forward pass latency: 1.09e-02 (s)
+[2024-02-14 19:27:11,861][inference][INFO] - 	+ Forward pass throughput: 91.70 (samples/s)
+[2024-02-14 19:27:11,861][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 19:27:11,861][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:27:11,861][inference][INFO] - 	+ Warming up the generation pass
+[2024-02-14 19:27:14,794][inference][INFO] - 	+ Tracking generation latency and throughput
+[2024-02-14 19:27:26,489][inference][INFO] - 	+ Generation pass latency: 2.92e+00 (s)
+[2024-02-14 19:27:26,490][inference][INFO] - 	+ Generation pass throughput: 87.70 (tokens/s)
+[2024-02-14 19:27:26,490][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 19:27:26,490][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:27:26,490][inference][INFO] - 	+ Tracking forward pass energy consumption
+[2024-02-14 19:27:42,565][inference][INFO] - 	+ Forward pass energy consumption: 1.14e-06 (kWh/sample)
+[2024-02-14 19:27:42,565][inference][INFO] - 	+ Forward pass carbon emissions: 7.72e-08 (kgCO2eq/sample)
+[2024-02-14 19:27:42,565][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/forward_codecarbon.csv
+[2024-02-14 19:27:42,566][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 19:27:42,566][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:27:42,566][inference][INFO] - 	+ Tracking generation pass energy consumption
+[2024-02-14 19:27:59,895][inference][INFO] - 	+ Generation pass energy consumption: 9.15e-07 (kWh/token)
+[2024-02-14 19:27:59,896][inference][INFO] - 	+ Generation pass carbon emissions: 6.16e-08 (kgCO2eq/token)
+[2024-02-14 19:27:59,896][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/generate_codecarbon.csv
+[2024-02-14 19:27:59,896][inference][INFO] - Saving results
+[2024-02-14 19:27:59,899][backend][INFO] - Cleaning pytorch backend
+[2024-02-14 19:27:59,899][backend][INFO] - 	+ Deleting pretrained model
+[2024-02-14 19:28:00,128][pytorch][INFO] - 	+ Cleaning temporary directory
+[2024-02-14 19:28:00,128][pytorch][INFO] - 	+ Emptying CUDA cache
+[2024-02-14 19:28:02,208][isolation][INFO] - 	+ Closing device(s) isolation process...

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/forward_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T19:27:42,codecarbon,4ac85be7-6246-4fd4-8388-098bcecf1b49,10.569979667663574,7.113921164897584e-05,6.7303073313007325e-06,112.5,615.1367847265042,0.34110403060913086,0.00033028850704431537,0.0007245664129857943,7.796199457894204e-07,0.001055634539975899,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/generate_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T19:27:59,codecarbon,943bf2df-34c1-40c0-8c00-50a5b50c477d,12.401583433151245,6.31205777395431e-05,5.089719234627134e-06,112.5,208.46505591460803,0.3414473533630371,0.0003875252157449722,0.0005481851607687105,9.356798677255066e-07,0.0009366460563814082,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/hydra_config.yaml

@@ -0,0 +1,85 @@
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+backend:
+  name: pytorch
+  version: 2.1.2+cu118
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: true
+  disable_grad: true
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: false
+  can_generate: true
+  forward_kwargs: {}
+  generate_kwargs:
+    num_return_sequences: 1
+    max_new_tokens: 256
+    min_new_tokens: 256
+    do_sample: false
+    use_cache: true
+    pad_token_id: 0
+    temperature: 1.0
+    num_beams: 1
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase-1b
+task: text-generation
+library: transformers
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-1b/inference_results.csv

@@ -0,0 +1,2 @@
+forward.latency(s),forward.throughput(samples/s),forward.peak_memory(MB),forward.max_memory_used(MB),forward.max_memory_allocated(MB),forward.max_memory_reserved(MB),forward.energy_consumption(kWh/sample),forward.carbon_emissions(kgCO2eq/sample),generate.latency(s),generate.throughput(tokens/s),decode.latency(s),decode.throughput(tokens/s),generate.peak_memory(MB),generate.max_memory_used(MB),generate.max_memory_allocated(MB),generate.max_memory_reserved(MB),generate.energy_consumption(kWh/token),generate.carbon_emissions(kgCO2eq/token)
+0.0109,91.7,3930,3930,2414,2472,1.14e-06,7.72e-08,2.92,87.7,2.91,87.6,3938,3938,2417,2478,9.15e-07,6.16e-08

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/.hydra/config.yaml

@@ -0,0 +1,77 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase-3b
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase-3b
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase-3b
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+bfloat16
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- model=bigcode/starcoderbase-3b

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/cli.log

@@ -0,0 +1,80 @@
+[2024-02-14 19:28:08,098][launcher][INFO] - Configuring process launcher
+[2024-02-14 19:28:08,098][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 19:28:08,100][process][INFO] - 	+ Launched worker process with PID 1955027.
+[2024-02-14 19:28:08,101][isolation][INFO] - 	+ Launched device(s) isolation process 1955028.
+[2024-02-14 19:28:14,735][isolation][INFO] - 	+ Closing device(s) isolation process...
+[2024-02-14 19:28:15,121][process][ERROR] - Worker process exited with code 1, forwarding...
+[2024-02-14 20:26:50,732][launcher][INFO] - Configuring process launcher
+[2024-02-14 20:26:50,732][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 20:26:50,735][process][INFO] - 	+ Launched worker process with PID 2018531.
+[2024-02-14 20:26:50,735][isolation][INFO] - 	+ Launched device(s) isolation process 2018532.
+[2024-02-14 20:26:56,346][isolation][ERROR] - Found non-permitted process(es) running on system device(s): {1988336, 2017825}
+[2024-02-14 20:26:56,346][isolation][ERROR] - Terminating benchmark process 2018318
+[2024-02-14 20:26:57,030][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2024-02-14 20:26:57,030][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2024-02-14 20:26:57,030][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2024-02-14 20:26:57,144][datasets][INFO] - PyTorch version 2.1.2+cu118 available.
+[2024-02-14 20:26:58,313][backend][INFO] - Configuring pytorch backend
+[2024-02-14 20:26:58,313][pytorch][INFO] - 	+ Inferred class AutoModelForCausalLM for task text-generation and model_type gpt_bigcode
+[2024-02-14 20:26:58,313][pytorch][INFO] - 	+ Disabling gradients
+[2024-02-14 20:26:58,313][pytorch][INFO] - 	+ Loading model with no weights
+[2024-02-14 20:26:58,314][pytorch][INFO] - 	+ Creating no weights model directory
+[2024-02-14 20:26:58,314][pytorch][INFO] - 	+ Saving pretrained config
+[2024-02-14 20:26:58,315][pytorch][INFO] - 	+ Creating no weights model
+[2024-02-14 20:26:58,319][pytorch][INFO] - 	+ Saving no weights model
+[2024-02-14 20:26:58,320][pytorch][INFO] - 	+ Loading no weights model
+[2024-02-14 20:26:58,320][pytorch][INFO] - 	+ Loading model directly on device: cuda
+[2024-02-14 20:27:00,044][pytorch][INFO] - 	+ Randomizing model weights
+[2024-02-14 20:27:00,050][pytorch][INFO] - 	+ Tying model weights after randomization
+[2024-02-14 20:27:00,052][pytorch][INFO] - 	+ Turning on model's eval mode
+[2024-02-14 20:27:00,163][benchmark][INFO] - Configuring inference benchmark
+[2024-02-14 20:27:00,163][inference][INFO] - Running inference benchmark
+[2024-02-14 20:27:00,163][inference][INFO] - 	+ Updating input shapes with model shapes
+[2024-02-14 20:27:00,163][inference][INFO] - 	+ Preparing backend for inference
+[2024-02-14 20:27:00,163][inference][INFO] - 	+ Creating input generator
+[2024-02-14 20:27:00,163][input-generator][INFO] - Using text-generation task generator
+[2024-02-14 20:27:00,164][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 20:27:00,164][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 20:27:00,174][inference][INFO] - 	+ Tracking forward pass peak memory
+[2024-02-14 20:27:00,174][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 20:27:00,174][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 20:27:03,328][inference][INFO] - 	+ Forward pass max memory used: 20568 (MB)
+[2024-02-14 20:27:03,328][inference][INFO] - 	+ Forward pass max memory reserved: 6404 (MB)
+[2024-02-14 20:27:03,328][inference][INFO] - 	+ Forward pass max memory allocated: 6278 (MB)
+[2024-02-14 20:27:03,328][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 20:27:03,328][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 20:27:03,355][inference][INFO] - 	+ Tracking generation pass peak memory
+[2024-02-14 20:27:03,355][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 20:27:03,355][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 20:27:13,280][inference][INFO] - 	+ Generation pass max memory used: 52928 (MB)
+[2024-02-14 20:27:13,280][inference][INFO] - 	+ Generation pass max memory reserved: 6425 (MB)
+[2024-02-14 20:27:13,280][inference][INFO] - 	+ Generation pass max memory allocated: 6282 (MB)
+[2024-02-14 20:27:13,280][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 20:27:13,281][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 20:27:13,292][inference][INFO] - 	+ Warming up the forward pass
+[2024-02-14 20:27:13,704][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2024-02-14 20:27:23,774][inference][INFO] - 	+ Forward pass latency: 4.71e-02 (s)
+[2024-02-14 20:27:23,774][inference][INFO] - 	+ Forward pass throughput: 21.20 (samples/s)
+[2024-02-14 20:27:23,775][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 20:27:23,775][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 20:27:23,778][inference][INFO] - 	+ Warming up the generation pass
+[2024-02-14 20:27:29,395][inference][INFO] - 	+ Tracking generation latency and throughput
+[2024-02-14 20:27:41,732][inference][INFO] - 	+ Generation pass latency: 6.17e+00 (s)
+[2024-02-14 20:27:41,733][inference][INFO] - 	+ Generation pass throughput: 41.50 (tokens/s)
+[2024-02-14 20:27:41,733][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 20:27:41,733][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 20:27:41,734][inference][INFO] - 	+ Tracking forward pass energy consumption
+[2024-02-14 20:27:57,881][inference][INFO] - 	+ Forward pass energy consumption: 4.71e-06 (kWh/sample)
+[2024-02-14 20:27:57,881][inference][INFO] - 	+ Forward pass carbon emissions: 3.18e-07 (kgCO2eq/sample)
+[2024-02-14 20:27:57,881][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/forward_codecarbon.csv
+[2024-02-14 20:27:57,881][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 20:27:57,882][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 20:27:57,885][inference][INFO] - 	+ Tracking generation pass energy consumption
+[2024-02-14 20:28:17,303][inference][INFO] - 	+ Generation pass energy consumption: 2.73e-06 (kWh/token)
+[2024-02-14 20:28:17,303][inference][INFO] - 	+ Generation pass carbon emissions: 1.84e-07 (kgCO2eq/token)
+[2024-02-14 20:28:17,303][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/generate_codecarbon.csv
+[2024-02-14 20:28:17,303][inference][INFO] - Saving results
+[2024-02-14 20:28:17,305][backend][INFO] - Cleaning pytorch backend
+[2024-02-14 20:28:17,305][backend][INFO] - 	+ Deleting pretrained model
+[2024-02-14 20:28:17,540][pytorch][INFO] - 	+ Cleaning temporary directory
+[2024-02-14 20:28:17,541][pytorch][INFO] - 	+ Emptying CUDA cache

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/forward_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T20:27:57,codecarbon,034cbdf7-bfd1-4f08-92b6-5837690b2e0c,10.64318060874939,7.335821861880229e-05,6.892509045510046e-06,112.5,769.2913669213469,0.34687042236328125,0.00033257462829351436,0.0007552114375037888,7.763116890048612e-07,0.0010885623774863078,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/generate_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T20:28:17,codecarbon,0fc3d538-cc37-4a10-9cbb-ce4841c1e09b,13.590092897415161,9.40505366891559e-05,6.920521986059737e-06,112.5,264.50751714613153,0.34786319732666016,0.00042466054111719136,0.0009699432759529003,1.0118037810126645e-06,0.0013956156208511038,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/hydra_config.yaml

@@ -0,0 +1,85 @@
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+backend:
+  name: pytorch
+  version: 2.1.2+cu118
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: true
+  disable_grad: true
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: false
+  can_generate: true
+  forward_kwargs: {}
+  generate_kwargs:
+    num_return_sequences: 1
+    max_new_tokens: 256
+    min_new_tokens: 256
+    do_sample: false
+    use_cache: true
+    pad_token_id: 0
+    temperature: 1.0
+    num_beams: 1
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase-3b
+task: text-generation
+library: transformers
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-3b/inference_results.csv

@@ -0,0 +1,2 @@
+forward.latency(s),forward.throughput(samples/s),forward.peak_memory(MB),forward.max_memory_used(MB),forward.max_memory_allocated(MB),forward.max_memory_reserved(MB),forward.energy_consumption(kWh/sample),forward.carbon_emissions(kgCO2eq/sample),generate.latency(s),generate.throughput(tokens/s),decode.latency(s),decode.throughput(tokens/s),generate.peak_memory(MB),generate.max_memory_used(MB),generate.max_memory_allocated(MB),generate.max_memory_reserved(MB),generate.energy_consumption(kWh/token),generate.carbon_emissions(kgCO2eq/token)
+0.0471,21.2,20568,20568,6278,6404,4.71e-06,3.18e-07,6.17,41.5,6.12,41.7,52928,52928,6282,6425,2.73e-06,1.84e-07

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/.hydra/config.yaml

@@ -0,0 +1,77 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase-7b
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase-7b
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase-7b
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+bfloat16
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- model=bigcode/starcoderbase-7b

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/cli.log

@@ -0,0 +1,73 @@
+[2024-02-14 19:28:19,825][launcher][INFO] - Configuring process launcher
+[2024-02-14 19:28:19,825][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 19:28:19,828][process][INFO] - 	+ Launched worker process with PID 1955507.
+[2024-02-14 19:28:19,828][isolation][INFO] - 	+ Launched device(s) isolation process 1955508.
+[2024-02-14 19:28:26,394][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2024-02-14 19:28:26,394][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2024-02-14 19:28:26,394][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2024-02-14 19:28:26,504][datasets][INFO] - PyTorch version 2.1.2+cu118 available.
+[2024-02-14 19:28:29,160][backend][INFO] - Configuring pytorch backend
+[2024-02-14 19:28:29,160][pytorch][INFO] - 	+ Inferred class AutoModelForCausalLM for task text-generation and model_type gpt_bigcode
+[2024-02-14 19:28:29,160][pytorch][INFO] - 	+ Disabling gradients
+[2024-02-14 19:28:29,161][pytorch][INFO] - 	+ Loading model with no weights
+[2024-02-14 19:28:29,161][pytorch][INFO] - 	+ Creating no weights model directory
+[2024-02-14 19:28:29,161][pytorch][INFO] - 	+ Saving pretrained config
+[2024-02-14 19:28:29,162][pytorch][INFO] - 	+ Creating no weights model
+[2024-02-14 19:28:29,166][pytorch][INFO] - 	+ Saving no weights model
+[2024-02-14 19:28:29,166][pytorch][INFO] - 	+ Loading no weights model
+[2024-02-14 19:28:29,167][pytorch][INFO] - 	+ Loading model directly on device: cuda
+[2024-02-14 19:28:31,185][pytorch][INFO] - 	+ Randomizing model weights
+[2024-02-14 19:28:31,192][pytorch][INFO] - 	+ Tying model weights after randomization
+[2024-02-14 19:28:31,193][pytorch][INFO] - 	+ Turning on model's eval mode
+[2024-02-14 19:28:31,303][benchmark][INFO] - Configuring inference benchmark
+[2024-02-14 19:28:31,303][inference][INFO] - Running inference benchmark
+[2024-02-14 19:28:31,303][inference][INFO] - 	+ Updating input shapes with model shapes
+[2024-02-14 19:28:31,303][inference][INFO] - 	+ Preparing backend for inference
+[2024-02-14 19:28:31,303][inference][INFO] - 	+ Creating input generator
+[2024-02-14 19:28:31,303][input-generator][INFO] - Using text-generation task generator
+[2024-02-14 19:28:31,304][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 19:28:31,304][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:28:31,304][inference][INFO] - 	+ Tracking forward pass peak memory
+[2024-02-14 19:28:31,304][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 19:28:31,304][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 19:28:33,213][inference][INFO] - 	+ Forward pass max memory used: 16374 (MB)
+[2024-02-14 19:28:33,214][inference][INFO] - 	+ Forward pass max memory reserved: 14917 (MB)
+[2024-02-14 19:28:33,214][inference][INFO] - 	+ Forward pass max memory allocated: 14853 (MB)
+[2024-02-14 19:28:33,214][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 19:28:33,214][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:28:33,219][inference][INFO] - 	+ Tracking generation pass peak memory
+[2024-02-14 19:28:33,219][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 19:28:33,219][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 19:28:41,827][inference][INFO] - 	+ Generation pass max memory used: 16399 (MB)
+[2024-02-14 19:28:41,827][inference][INFO] - 	+ Generation pass max memory reserved: 14940 (MB)
+[2024-02-14 19:28:41,827][inference][INFO] - 	+ Generation pass max memory allocated: 14855 (MB)
+[2024-02-14 19:28:41,827][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 19:28:41,827][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:28:41,827][inference][INFO] - 	+ Warming up the forward pass
+[2024-02-14 19:28:42,106][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2024-02-14 19:28:52,132][inference][INFO] - 	+ Forward pass latency: 3.15e-02 (s)
+[2024-02-14 19:28:52,133][inference][INFO] - 	+ Forward pass throughput: 31.70 (samples/s)
+[2024-02-14 19:28:52,133][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 19:28:52,133][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:28:52,133][inference][INFO] - 	+ Warming up the generation pass
+[2024-02-14 19:28:57,019][inference][INFO] - 	+ Tracking generation latency and throughput
+[2024-02-14 19:29:11,714][inference][INFO] - 	+ Generation pass latency: 4.90e+00 (s)
+[2024-02-14 19:29:11,714][inference][INFO] - 	+ Generation pass throughput: 52.20 (tokens/s)
+[2024-02-14 19:29:11,714][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 19:29:11,715][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:29:11,715][inference][INFO] - 	+ Tracking forward pass energy consumption
+[2024-02-14 19:29:28,481][inference][INFO] - 	+ Forward pass energy consumption: 3.27e-06 (kWh/sample)
+[2024-02-14 19:29:28,481][inference][INFO] - 	+ Forward pass carbon emissions: 2.21e-07 (kgCO2eq/sample)
+[2024-02-14 19:29:28,481][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/forward_codecarbon.csv
+[2024-02-14 19:29:28,481][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 19:29:28,481][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 19:29:28,482][inference][INFO] - 	+ Tracking generation pass energy consumption
+[2024-02-14 19:29:48,958][inference][INFO] - 	+ Generation pass energy consumption: 2.01e-06 (kWh/token)
+[2024-02-14 19:29:48,958][inference][INFO] - 	+ Generation pass carbon emissions: 1.35e-07 (kgCO2eq/token)
+[2024-02-14 19:29:48,958][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/generate_codecarbon.csv
+[2024-02-14 19:29:48,958][inference][INFO] - Saving results
+[2024-02-14 19:29:48,960][backend][INFO] - Cleaning pytorch backend
+[2024-02-14 19:29:48,960][backend][INFO] - 	+ Deleting pretrained model
+[2024-02-14 19:29:49,186][pytorch][INFO] - 	+ Cleaning temporary directory
+[2024-02-14 19:29:49,187][pytorch][INFO] - 	+ Emptying CUDA cache
+[2024-02-14 19:29:52,609][isolation][INFO] - 	+ Closing device(s) isolation process...

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/forward_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T19:29:28,codecarbon,ba80498a-3787-4557-b10e-7a0cb676452b,10.584910154342651,7.541600888682554e-05,7.124860559717151e-06,112.5,858.5099045401266,0.3457632064819336,0.0003307555988430977,0.0007876086856413167,7.336363766264261e-07,0.0011190979208610407,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/generate_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T19:29:48,codecarbon,ed6e1fb2-8e33-450c-a013-402d161331e3,15.274165868759155,0.00010402625471507791,6.8106013519106045e-06,112.5,514.0111410067171,0.34665727615356445,0.00047728438675403604,0.0010652127966128688,1.1480861846295432e-06,0.0015436452695515345,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/hydra_config.yaml

@@ -0,0 +1,85 @@
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+backend:
+  name: pytorch
+  version: 2.1.2+cu118
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: true
+  disable_grad: true
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: false
+  can_generate: true
+  forward_kwargs: {}
+  generate_kwargs:
+    num_return_sequences: 1
+    max_new_tokens: 256
+    min_new_tokens: 256
+    do_sample: false
+    use_cache: true
+    pad_token_id: 0
+    temperature: 1.0
+    num_beams: 1
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase-7b
+task: text-generation
+library: transformers
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase-7b/inference_results.csv

@@ -0,0 +1,2 @@
+forward.latency(s),forward.throughput(samples/s),forward.peak_memory(MB),forward.max_memory_used(MB),forward.max_memory_allocated(MB),forward.max_memory_reserved(MB),forward.energy_consumption(kWh/sample),forward.carbon_emissions(kgCO2eq/sample),generate.latency(s),generate.throughput(tokens/s),decode.latency(s),decode.throughput(tokens/s),generate.peak_memory(MB),generate.max_memory_used(MB),generate.max_memory_allocated(MB),generate.max_memory_reserved(MB),generate.energy_consumption(kWh/token),generate.carbon_emissions(kgCO2eq/token)
+0.0315,31.7,16374,16374,14853,14917,3.27e-06,2.21e-07,4.9,52.2,4.87,52.4,16399,16399,14855,14940,2.01e-06,1.35e-07

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/.hydra/config.yaml

@@ -0,0 +1,77 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+bfloat16
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- model=bigcode/starcoderbase

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/cli.log

@@ -0,0 +1,73 @@
+[2024-02-14 17:59:46,545][launcher][INFO] - Configuring process launcher
+[2024-02-14 17:59:46,546][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 17:59:46,548][process][INFO] - 	+ Launched worker process with PID 1875308.
+[2024-02-14 17:59:46,549][isolation][INFO] - 	+ Launched device(s) isolation process 1875309.
+[2024-02-14 17:59:52,797][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2024-02-14 17:59:52,797][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2024-02-14 17:59:52,797][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2024-02-14 17:59:52,905][datasets][INFO] - PyTorch version 2.1.2+cu118 available.
+[2024-02-14 17:59:54,081][backend][INFO] - Configuring pytorch backend
+[2024-02-14 17:59:54,081][pytorch][INFO] - 	+ Inferred class AutoModelForCausalLM for task text-generation and model_type gpt_bigcode
+[2024-02-14 17:59:54,081][pytorch][INFO] - 	+ Disabling gradients
+[2024-02-14 17:59:54,081][pytorch][INFO] - 	+ Loading model with no weights
+[2024-02-14 17:59:54,081][pytorch][INFO] - 	+ Creating no weights model directory
+[2024-02-14 17:59:54,082][pytorch][INFO] - 	+ Saving pretrained config
+[2024-02-14 17:59:54,082][pytorch][INFO] - 	+ Creating no weights model
+[2024-02-14 17:59:54,085][pytorch][INFO] - 	+ Saving no weights model
+[2024-02-14 17:59:54,085][pytorch][INFO] - 	+ Loading no weights model
+[2024-02-14 17:59:54,085][pytorch][INFO] - 	+ Loading model directly on device: cuda
+[2024-02-14 17:59:56,058][pytorch][INFO] - 	+ Randomizing model weights
+[2024-02-14 17:59:56,064][pytorch][INFO] - 	+ Tying model weights after randomization
+[2024-02-14 17:59:56,066][pytorch][INFO] - 	+ Turning on model's eval mode
+[2024-02-14 17:59:56,175][benchmark][INFO] - Configuring inference benchmark
+[2024-02-14 17:59:56,175][inference][INFO] - Running inference benchmark
+[2024-02-14 17:59:56,175][inference][INFO] - 	+ Updating input shapes with model shapes
+[2024-02-14 17:59:56,175][inference][INFO] - 	+ Preparing backend for inference
+[2024-02-14 17:59:56,176][inference][INFO] - 	+ Creating input generator
+[2024-02-14 17:59:56,176][input-generator][INFO] - Using text-generation task generator
+[2024-02-14 17:59:56,176][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 17:59:56,176][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 17:59:56,213][inference][INFO] - 	+ Tracking forward pass peak memory
+[2024-02-14 17:59:56,213][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 17:59:56,213][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 17:59:59,297][inference][INFO] - 	+ Forward pass max memory used: 32868 (MB)
+[2024-02-14 17:59:59,297][inference][INFO] - 	+ Forward pass max memory reserved: 31411 (MB)
+[2024-02-14 17:59:59,297][inference][INFO] - 	+ Forward pass max memory allocated: 31327 (MB)
+[2024-02-14 17:59:59,297][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 17:59:59,297][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 17:59:59,297][inference][INFO] - 	+ Tracking generation pass peak memory
+[2024-02-14 17:59:59,297][memory][INFO] - Tracking CUDA devices: [0]
+[2024-02-14 17:59:59,298][memory][INFO] - Tracking Pytorch CUDA devices: [0]
+[2024-02-14 18:00:08,361][inference][INFO] - 	+ Generation pass max memory used: 32893 (MB)
+[2024-02-14 18:00:08,361][inference][INFO] - 	+ Generation pass max memory reserved: 31434 (MB)
+[2024-02-14 18:00:08,361][inference][INFO] - 	+ Generation pass max memory allocated: 31327 (MB)
+[2024-02-14 18:00:08,362][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 18:00:08,362][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 18:00:08,362][inference][INFO] - 	+ Warming up the forward pass
+[2024-02-14 18:00:09,384][inference][INFO] - 	+ Tracking forward pass latency and throughput
+[2024-02-14 18:00:19,460][inference][INFO] - 	+ Forward pass latency: 5.34e-02 (s)
+[2024-02-14 18:00:19,461][inference][INFO] - 	+ Forward pass throughput: 18.70 (samples/s)
+[2024-02-14 18:00:19,461][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 18:00:19,461][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 18:00:19,461][inference][INFO] - 	+ Warming up the generation pass
+[2024-02-14 18:00:25,464][inference][INFO] - 	+ Tracking generation latency and throughput
+[2024-02-14 18:00:37,467][inference][INFO] - 	+ Generation pass latency: 6.00e+00 (s)
+[2024-02-14 18:00:37,467][inference][INFO] - 	+ Generation pass throughput: 42.70 (tokens/s)
+[2024-02-14 18:00:37,467][inference][INFO] - 	+ Preparing input for the forward pass
+[2024-02-14 18:00:37,467][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 18:00:37,467][inference][INFO] - 	+ Tracking forward pass energy consumption
+[2024-02-14 18:00:53,074][inference][INFO] - 	+ Forward pass energy consumption: 6.03e-06 (kWh/sample)
+[2024-02-14 18:00:53,074][inference][INFO] - 	+ Forward pass carbon emissions: 4.06e-07 (kgCO2eq/sample)
+[2024-02-14 18:00:53,074][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/forward_codecarbon.csv
+[2024-02-14 18:00:53,075][inference][INFO] - 	+ Preparing input for the generation pass
+[2024-02-14 18:00:53,075][pytorch][INFO] - 	+ Moving inputs tensors to device cuda
+[2024-02-14 18:00:53,075][inference][INFO] - 	+ Tracking generation pass energy consumption
+[2024-02-14 18:01:11,177][inference][INFO] - 	+ Generation pass energy consumption: 2.61e-06 (kWh/token)
+[2024-02-14 18:01:11,177][inference][INFO] - 	+ Generation pass carbon emissions: 1.76e-07 (kgCO2eq/token)
+[2024-02-14 18:01:11,177][inference][INFO] - 	+ Full details in the CodeCarbon report: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/generate_codecarbon.csv
+[2024-02-14 18:01:11,177][inference][INFO] - Saving results
+[2024-02-14 18:01:11,179][backend][INFO] - Cleaning pytorch backend
+[2024-02-14 18:01:11,179][backend][INFO] - 	+ Deleting pretrained model
+[2024-02-14 18:01:11,410][pytorch][INFO] - 	+ Cleaning temporary directory
+[2024-02-14 18:01:11,411][pytorch][INFO] - 	+ Emptying CUDA cache
+[2024-02-14 18:01:13,504][isolation][INFO] - 	+ Closing device(s) isolation process...

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/forward_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T18:00:53,codecarbon,b9b0c3b0-e693-492a-b26c-32fb4af114d8,10.493306159973145,7.430436669633096e-05,7.081120627144756e-06,112.5,866.1832244563208,0.34720659255981445,0.0003278931528329849,0.000773969785841544,7.393271858503188e-07,0.0011026022658603791,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/generate_codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,on_cloud,pue
+2024-02-14T18:01:11,codecarbon,0804188d-f1cb-48d7-bd90-aa7ff798d785,12.91486644744873,9.005441423796991e-05,6.972926480069e-06,112.5,0.0,0.34815502166748047,0.00040356197208166124,0.0009318851899529079,8.700102160601849e-07,0.0013363171722506292,France,FRA,île-de-france,,,Linux-5.4.0-166-generic-x86_64-with-glibc2.35,3.10.12,2.3.4,128,AMD EPYC 7742 64-Core Processor,1,1 x NVIDIA A100-SXM4-80GB,2.4075,48.8323,503.5396919250488,process,N,1.0

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/hydra_config.yaml

@@ -0,0 +1,85 @@
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+backend:
+  name: pytorch
+  version: 2.1.2+cu118
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: bfloat16
+  eval_mode: true
+  disable_grad: true
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: null
+  quantization_config: {}
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: false
+  can_generate: true
+  forward_kwargs: {}
+  generate_kwargs:
+    num_return_sequences: 1
+    max_new_tokens: 256
+    min_new_tokens: 256
+    do_sample: false
+    use_cache: true
+    pad_token_id: 0
+    temperature: 1.0
+    num_beams: 1
+experiment_name: pytorch+cuda+bfloat16
+device: cuda
+model: bigcode/starcoderbase
+task: text-generation
+library: transformers
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+bfloat16/bigcode/starcoderbase/inference_results.csv

@@ -0,0 +1,2 @@
+forward.latency(s),forward.throughput(samples/s),forward.peak_memory(MB),forward.max_memory_used(MB),forward.max_memory_allocated(MB),forward.max_memory_reserved(MB),forward.energy_consumption(kWh/sample),forward.carbon_emissions(kgCO2eq/sample),generate.latency(s),generate.throughput(tokens/s),decode.latency(s),decode.throughput(tokens/s),generate.peak_memory(MB),generate.max_memory_used(MB),generate.max_memory_allocated(MB),generate.max_memory_reserved(MB),generate.energy_consumption(kWh/token),generate.carbon_emissions(kgCO2eq/token)
+0.0534,18.7,32868,32868,31327,31411,6.03e-06,4.06e-07,6.0,42.7,5.95,42.9,32893,32893,31327,31434,2.61e-06,1.76e-07

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-1b/.hydra/config.yaml

@@ -0,0 +1,80 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: float16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: awq
+  quantization_config:
+    bits: 4
+    version: gemm
+    do_fuse: false
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+float16+awq-4bit+gemm
+device: cuda
+model: bigcode/starcoderbase-1b
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-1b/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase-1b
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase-1b
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+float16+awq-4bit+gemm
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-1b
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-1b/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- model=bigcode/starcoderbase-1b

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-1b/cli.log

@@ -0,0 +1,26 @@
+[2024-02-14 20:11:51,714][launcher][INFO] - Configuring process launcher
+[2024-02-14 20:11:51,715][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 20:11:51,717][process][INFO] - 	+ Launched worker process with PID 2000347.
+[2024-02-14 20:11:51,718][isolation][INFO] - 	+ Launched device(s) isolation process 2000348.
+[2024-02-14 20:11:58,493][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2024-02-14 20:11:58,494][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2024-02-14 20:11:58,494][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2024-02-14 20:11:58,661][datasets][INFO] - PyTorch version 2.1.2+cu118 available.
+[2024-02-14 20:11:58,985][isolation][ERROR] - Found non-permitted process(es) running on system device(s): {1989630}
+[2024-02-14 20:11:58,985][isolation][ERROR] - Terminating benchmark process 2000113
+[2024-02-14 20:11:59,883][backend][INFO] - Configuring pytorch backend
+[2024-02-14 20:11:59,883][pytorch][INFO] - 	+ Inferred class AutoModelForCausalLM for task text-generation and model_type gpt_bigcode
+[2024-02-14 20:11:59,883][pytorch][INFO] - 	+ Disabling gradients
+[2024-02-14 20:11:59,883][pytorch][INFO] - 	+ Processing quantization config
+[2024-02-14 20:11:59,883][pytorch][INFO] - 	+ Processing AWQ config
+[2024-02-14 20:12:01,241][pytorch][INFO] - 	+ Loading model with no weights
+[2024-02-14 20:12:01,242][pytorch][INFO] - 	+ Creating no weights model directory
+[2024-02-14 20:12:01,242][pytorch][INFO] - 	+ Saving pretrained config
+[2024-02-14 20:12:01,243][pytorch][INFO] - 	+ Creating no weights model
+[2024-02-14 20:12:01,247][pytorch][INFO] - 	+ Saving no weights model
+[2024-02-14 20:12:01,247][pytorch][INFO] - 	+ Loading no weights model
+[2024-02-14 20:12:01,247][pytorch][INFO] - 	+ Loading quantized model
+[2024-02-14 20:12:15,038][experiment][ERROR] - Error during backend configuration: 'WQLinear_GEMM' object has no attribute 'weight'
+[2024-02-14 20:12:15,038][backend][INFO] - Cleaning pytorch backend
+[2024-02-14 20:12:15,229][pytorch][INFO] - 	+ Cleaning temporary directory
+[2024-02-14 20:12:15,229][pytorch][INFO] - 	+ Emptying CUDA cache

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-1b/hydra_config.yaml

@@ -0,0 +1,88 @@
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+backend:
+  name: pytorch
+  version: 2.1.2+cu118
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: float16
+  eval_mode: true
+  disable_grad: true
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: awq
+  quantization_config:
+    bits: 4
+    version: gemm
+    do_fuse: false
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: false
+  can_generate: true
+  forward_kwargs: {}
+  generate_kwargs:
+    num_return_sequences: 1
+    max_new_tokens: 256
+    min_new_tokens: 256
+    do_sample: false
+    use_cache: true
+    pad_token_id: 0
+    temperature: 1.0
+    num_beams: 1
+experiment_name: pytorch+cuda+float16+awq-4bit+gemm
+device: cuda
+model: bigcode/starcoderbase-1b
+task: text-generation
+library: transformers
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-3b/.hydra/config.yaml

@@ -0,0 +1,80 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: float16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: awq
+  quantization_config:
+    bits: 4
+    version: gemm
+    do_fuse: false
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+float16+awq-4bit+gemm
+device: cuda
+model: bigcode/starcoderbase-3b
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-3b/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase-3b
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase-3b
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+float16+awq-4bit+gemm
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-3b
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-3b/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- model=bigcode/starcoderbase-3b

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-3b/cli.log

@@ -0,0 +1,26 @@
+[2024-02-14 20:12:03,843][launcher][INFO] - Configuring process launcher
+[2024-02-14 20:12:03,843][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 20:12:03,846][process][INFO] - 	+ Launched worker process with PID 2000920.
+[2024-02-14 20:12:03,846][isolation][INFO] - 	+ Launched device(s) isolation process 2000921.
+[2024-02-14 20:12:10,044][isolation][ERROR] - Found non-permitted process(es) running on system device(s): {1989630}
+[2024-02-14 20:12:10,044][isolation][ERROR] - Terminating benchmark process 2000665
+[2024-02-14 20:12:10,643][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2024-02-14 20:12:10,643][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2024-02-14 20:12:10,643][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2024-02-14 20:12:10,759][datasets][INFO] - PyTorch version 2.1.2+cu118 available.
+[2024-02-14 20:12:12,453][backend][INFO] - Configuring pytorch backend
+[2024-02-14 20:12:12,453][pytorch][INFO] - 	+ Inferred class AutoModelForCausalLM for task text-generation and model_type gpt_bigcode
+[2024-02-14 20:12:12,453][pytorch][INFO] - 	+ Disabling gradients
+[2024-02-14 20:12:12,453][pytorch][INFO] - 	+ Processing quantization config
+[2024-02-14 20:12:12,453][pytorch][INFO] - 	+ Processing AWQ config
+[2024-02-14 20:12:12,878][pytorch][INFO] - 	+ Loading model with no weights
+[2024-02-14 20:12:12,879][pytorch][INFO] - 	+ Creating no weights model directory
+[2024-02-14 20:12:12,879][pytorch][INFO] - 	+ Saving pretrained config
+[2024-02-14 20:12:12,880][pytorch][INFO] - 	+ Creating no weights model
+[2024-02-14 20:12:12,885][pytorch][INFO] - 	+ Saving no weights model
+[2024-02-14 20:12:12,885][pytorch][INFO] - 	+ Loading no weights model
+[2024-02-14 20:12:12,886][pytorch][INFO] - 	+ Loading quantized model
+[2024-02-14 20:12:33,600][experiment][ERROR] - Error during backend configuration: 'WQLinear_GEMM' object has no attribute 'weight'
+[2024-02-14 20:12:33,600][backend][INFO] - Cleaning pytorch backend
+[2024-02-14 20:12:33,716][pytorch][INFO] - 	+ Cleaning temporary directory
+[2024-02-14 20:12:33,717][pytorch][INFO] - 	+ Emptying CUDA cache

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-3b/hydra_config.yaml

@@ -0,0 +1,88 @@
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+backend:
+  name: pytorch
+  version: 2.1.2+cu118
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: float16
+  eval_mode: true
+  disable_grad: true
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: awq
+  quantization_config:
+    bits: 4
+    version: gemm
+    do_fuse: false
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: false
+  can_generate: true
+  forward_kwargs: {}
+  generate_kwargs:
+    num_return_sequences: 1
+    max_new_tokens: 256
+    min_new_tokens: 256
+    do_sample: false
+    use_cache: true
+    pad_token_id: 0
+    temperature: 1.0
+    num_beams: 1
+experiment_name: pytorch+cuda+float16+awq-4bit+gemm
+device: cuda
+model: bigcode/starcoderbase-3b
+task: text-generation
+library: transformers
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-7b/.hydra/config.yaml

@@ -0,0 +1,80 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: float16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: awq
+  quantization_config:
+    bits: 4
+    version: gemm
+    do_fuse: false
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+float16+awq-4bit+gemm
+device: cuda
+model: bigcode/starcoderbase-7b
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-7b/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase-7b
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase-7b
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+float16+awq-4bit+gemm
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-7b
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-7b/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- model=bigcode/starcoderbase-7b

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-7b/cli.log

@@ -0,0 +1,26 @@
+[2024-02-14 20:12:14,937][launcher][INFO] - Configuring process launcher
+[2024-02-14 20:12:14,937][process][INFO] - Setting multiprocessing start method to spawn.
+[2024-02-14 20:12:14,939][process][INFO] - 	+ Launched worker process with PID 2001537.
+[2024-02-14 20:12:14,940][isolation][INFO] - 	+ Launched device(s) isolation process 2001538.
+[2024-02-14 20:12:21,080][isolation][ERROR] - Found non-permitted process(es) running on system device(s): {1989630}
+[2024-02-14 20:12:21,080][isolation][ERROR] - Terminating benchmark process 2001272
+[2024-02-14 20:12:21,313][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2024-02-14 20:12:21,313][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2024-02-14 20:12:21,313][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2024-02-14 20:12:21,428][datasets][INFO] - PyTorch version 2.1.2+cu118 available.
+[2024-02-14 20:12:22,919][backend][INFO] - Configuring pytorch backend
+[2024-02-14 20:12:22,919][pytorch][INFO] - 	+ Inferred class AutoModelForCausalLM for task text-generation and model_type gpt_bigcode
+[2024-02-14 20:12:22,919][pytorch][INFO] - 	+ Disabling gradients
+[2024-02-14 20:12:22,919][pytorch][INFO] - 	+ Processing quantization config
+[2024-02-14 20:12:22,919][pytorch][INFO] - 	+ Processing AWQ config
+[2024-02-14 20:12:22,948][pytorch][INFO] - 	+ Loading model with no weights
+[2024-02-14 20:12:22,949][pytorch][INFO] - 	+ Creating no weights model directory
+[2024-02-14 20:12:22,949][pytorch][INFO] - 	+ Saving pretrained config
+[2024-02-14 20:12:22,951][pytorch][INFO] - 	+ Creating no weights model
+[2024-02-14 20:12:22,955][pytorch][INFO] - 	+ Saving no weights model
+[2024-02-14 20:12:22,956][pytorch][INFO] - 	+ Loading no weights model
+[2024-02-14 20:12:22,956][pytorch][INFO] - 	+ Loading quantized model
+[2024-02-14 20:12:53,845][experiment][ERROR] - Error during backend configuration: 'WQLinear_GEMM' object has no attribute 'weight'
+[2024-02-14 20:12:53,845][backend][INFO] - Cleaning pytorch backend
+[2024-02-14 20:12:53,965][pytorch][INFO] - 	+ Cleaning temporary directory
+[2024-02-14 20:12:53,966][pytorch][INFO] - 	+ Emptying CUDA cache

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase-7b/hydra_config.yaml

@@ -0,0 +1,88 @@
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+backend:
+  name: pytorch
+  version: 2.1.2+cu118
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: float16
+  eval_mode: true
+  disable_grad: true
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: awq
+  quantization_config:
+    bits: 4
+    version: gemm
+    do_fuse: false
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: false
+  can_generate: true
+  forward_kwargs: {}
+  generate_kwargs:
+    num_return_sequences: 1
+    max_new_tokens: 256
+    min_new_tokens: 256
+    do_sample: false
+    use_cache: true
+    pad_token_id: 0
+    temperature: 1.0
+    num_beams: 1
+experiment_name: pytorch+cuda+float16+awq-4bit+gemm
+device: cuda
+model: bigcode/starcoderbase-7b
+task: text-generation
+library: transformers
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase/.hydra/config.yaml

@@ -0,0 +1,80 @@
+backend:
+  name: pytorch
+  version: ${pytorch_version:}
+  _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
+  seed: 42
+  inter_op_num_threads: null
+  intra_op_num_threads: null
+  delete_cache: false
+  no_weights: true
+  device_map: null
+  torch_dtype: float16
+  eval_mode: ${is_inference:${benchmark.name}}
+  disable_grad: ${is_inference:${benchmark.name}}
+  amp_autocast: false
+  amp_dtype: null
+  torch_compile: false
+  torch_compile_config: {}
+  to_bettertransformer: false
+  use_flash_attention_2: false
+  quantization_scheme: awq
+  quantization_config:
+    bits: 4
+    version: gemm
+    do_fuse: false
+  data_parallel: false
+  deepspeed_inference: false
+  deepspeed_inference_config: {}
+  peft_strategy: null
+  peft_config: {}
+benchmark:
+  name: inference
+  _target_: optimum_benchmark.benchmarks.inference.benchmark.InferenceBenchmark
+  duration: 10
+  warmup_runs: 10
+  memory: true
+  energy: true
+  input_shapes:
+    batch_size: 1
+    sequence_length: 256
+    num_choices: 1
+    feature_size: 80
+    nb_max_frames: 3000
+    audio_sequence_length: 16000
+  new_tokens: 256
+  can_diffuse: ${can_diffuse:${task}}
+  can_generate: ${can_generate:${task}}
+  forward_kwargs: {}
+  generate_kwargs: {}
+launcher:
+  name: process
+  _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
+  device_isolation: true
+  start_method: spawn
+experiment_name: pytorch+cuda+float16+awq-4bit+gemm
+device: cuda
+model: bigcode/starcoderbase
+task: ${infer_task:${model}}
+library: ${infer_library:${model}}
+hub_kwargs:
+  revision: main
+  cache_dir: null
+  force_download: false
+  local_files_only: false
+  trust_remote_code: true
+environment:
+  optimum_version: 1.16.2
+  optimum_commit: null
+  transformers_version: 4.37.2
+  transformers_commit: null
+  accelerate_version: 0.27.2
+  accelerate_commit: null
+  diffusers_version: null
+  diffusers_commit: null
+  python_version: 3.10.12
+  system: Linux
+  cpu: ' AMD EPYC 7742 64-Core Processor'
+  cpu_count: 128
+  cpu_ram_mb: 540671
+  gpus:
+  - NVIDIA A100-SXM4-80GB

hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase/.hydra/hydra.yaml

@@ -0,0 +1,176 @@
+hydra:
+  run:
+    dir: dataset/${oc.env:HOSTNAME}/${experiment_name}/${model}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+      colorlog:
+        (): colorlog.ColoredFormatter
+        format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
+          - %(message)s'
+        log_colors:
+          DEBUG: purple
+          INFO: green
+          WARNING: yellow
+          ERROR: red
+          CRITICAL: red
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: colorlog
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - model=bigcode/starcoderbase
+  job:
+    name: cli
+    chdir: true
+    override_dirname: model=bigcode/starcoderbase
+    id: ???
+    num: ???
+    config_name: pytorch+cuda+float16+awq-4bit+gemm
+    env_set:
+      COUNTRY_ISO_CODE: FRA
+      OVERRIDE_BENCHMARKS: '0'
+      CUDA_VISIBLE_DEVICES: '0'
+      CUDA_DEVICE_ORDER: PCI_BUS_ID
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/opt-bench
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: optimum_benchmark
+      schema: pkg
+      provider: main
+    - path: hydra_plugins.hydra_colorlog.conf
+      schema: pkg
+      provider: hydra-colorlog
+    - path: /workspace/opt-bench/configs
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/opt-bench/dataset/hf-dgx-01/pytorch+cuda+float16+awq-4bit+gemm/bigcode/starcoderbase
+    choices:
+      launcher: process
+      benchmark: inference
+      backend: pytorch
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: colorlog
+      hydra/hydra_logging: colorlog
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false