Hugging Face's logo

Datasets:
hf-internal-testing
/
benchmark_results_merge_event

Dataset card Files Community
hf-transformers-bot commited on
Commit
cca31db
•
1 Parent(s): 4472f8d

Upload folder using huggingface_hub

Browse files
Files changed (32) hide show
  1. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/config.yaml +93 -0
  2. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/hydra.yaml +141 -0
  3. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/overrides.yaml +3 -0
  4. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.json +473 -0
  5. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.log +0 -0
  6. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_config.json +104 -0
  7. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_report.json +367 -0
  8. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/summary.json +13 -0
  9. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/config.yaml +93 -0
  10. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/hydra.yaml +141 -0
  11. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/overrides.yaml +3 -0
  12. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark.log +2 -0
  13. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark_config.json +104 -0
  14. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/config.yaml +93 -0
  15. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/hydra.yaml +141 -0
  16. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/overrides.yaml +3 -0
  17. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.json +473 -0
  18. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.log +0 -0
  19. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_config.json +104 -0
  20. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_report.json +367 -0
  21. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/summary.json +13 -0
  22. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/config.yaml +93 -0
  23. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/hydra.yaml +141 -0
  24. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/overrides.yaml +3 -0
  25. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark.json +473 -0
  26. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark.log +0 -0
  27. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark_config.json +104 -0
  28. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark_report.json +367 -0
  29. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/summary.json +13 -0
  30. 2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/multirun.yaml +234 -0
  31. 2024-09-09/summaries.json +41 -0
  32. 2024-09-09/summary.json +40 -0
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/config.yaml ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: pytorch_generate
2
+ backend:
3
+ name: pytorch
4
+ version: 2.4.0+cu121
5
+ _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
6
+ task: null
7
+ library: null
8
+ model_type: null
9
+ model: google/gemma-2b
10
+ processor: null
11
+ device: cuda
12
+ device_ids: '0'
13
+ seed: 42
14
+ inter_op_num_threads: null
15
+ intra_op_num_threads: null
16
+ model_kwargs: {}
17
+ processor_kwargs: {}
18
+ no_weights: true
19
+ device_map: null
20
+ torch_dtype: float16
21
+ eval_mode: true
22
+ to_bettertransformer: false
23
+ low_cpu_mem_usage: null
24
+ attn_implementation: null
25
+ cache_implementation: null
26
+ autocast_enabled: false
27
+ autocast_dtype: null
28
+ torch_compile: false
29
+ torch_compile_target: forward
30
+ torch_compile_config:
31
+ backend: inductor
32
+ mode: reduce-overhead
33
+ fullgraph: true
34
+ quantization_scheme: null
35
+ quantization_config: {}
36
+ deepspeed_inference: false
37
+ deepspeed_inference_config: {}
38
+ peft_type: null
39
+ peft_config: {}
40
+ scenario:
41
+ name: inference
42
+ _target_: optimum_benchmark.scenarios.inference.scenario.InferenceScenario
43
+ iterations: 2
44
+ duration: 0
45
+ warmup_runs: 10
46
+ input_shapes:
47
+ batch_size: 1
48
+ sequence_length: 7
49
+ new_tokens: null
50
+ memory: true
51
+ latency: true
52
+ energy: false
53
+ forward_kwargs: {}
54
+ generate_kwargs:
55
+ max_new_tokens: 128
56
+ min_new_tokens: 128
57
+ do_sample: false
58
+ call_kwargs: {}
59
+ launcher:
60
+ name: process
61
+ _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
62
+ device_isolation: true
63
+ device_isolation_action: warn
64
+ numactl: false
65
+ numactl_kwargs: {}
66
+ start_method: spawn
67
+ environment:
68
+ cpu: ' AMD EPYC 7R32'
69
+ cpu_count: 16
70
+ cpu_ram_mb: 66697.261056
71
+ system: Linux
72
+ machine: x86_64
73
+ platform: Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29
74
+ processor: x86_64
75
+ python_version: 3.8.10
76
+ gpu:
77
+ - NVIDIA A10G
78
+ gpu_count: 1
79
+ gpu_vram_mb: 24146608128
80
+ optimum_benchmark_version: 0.4.0
81
+ optimum_benchmark_commit: null
82
+ transformers_version: 4.45.0.dev0
83
+ transformers_commit: 65bb28444849976f853063edb958b3ef3dd59d12
84
+ accelerate_version: 0.35.0.dev0
85
+ accelerate_commit: null
86
+ diffusers_version: null
87
+ diffusers_commit: null
88
+ optimum_version: 1.22.0.dev0
89
+ optimum_commit: null
90
+ timm_version: 0.9.16
91
+ timm_commit: null
92
+ peft_version: 0.12.1.dev0
93
+ peft_commit: null
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/hydra.yaml ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
4
+ sweep:
5
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12
6
+ subdir: ${hydra.job.override_dirname}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ root:
71
+ level: ERROR
72
+ disable_existing_loggers: true
73
+ job_logging:
74
+ version: 1
75
+ root:
76
+ level: ERROR
77
+ disable_existing_loggers: true
78
+ env: {}
79
+ mode: MULTIRUN
80
+ searchpath: []
81
+ callbacks: {}
82
+ output_subdir: .hydra
83
+ overrides:
84
+ hydra:
85
+ - hydra/job_logging=disabled
86
+ - hydra/hydra_logging=disabled
87
+ - hydra.sweep.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12
88
+ - hydra.run.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
89
+ - hydra.mode=MULTIRUN
90
+ task:
91
+ - backend.model=google/gemma-2b
92
+ - backend.cache_implementation=null
93
+ - backend.torch_compile=False
94
+ job:
95
+ name: cli
96
+ chdir: true
97
+ override_dirname: backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False
98
+ id: '0'
99
+ num: 0
100
+ config_name: generation
101
+ env_set:
102
+ OVERRIDE_BENCHMARKS: '1'
103
+ LOG_LEVEL: WARN
104
+ env_copy: []
105
+ config:
106
+ override_dirname:
107
+ kv_sep: '='
108
+ item_sep: ','
109
+ exclude_keys: []
110
+ runtime:
111
+ version: 1.3.2
112
+ version_base: '1.3'
113
+ cwd: /transformers
114
+ config_sources:
115
+ - path: hydra.conf
116
+ schema: pkg
117
+ provider: hydra
118
+ - path: optimum_benchmark
119
+ schema: pkg
120
+ provider: main
121
+ - path: /transformers/benchmark/config
122
+ schema: file
123
+ provider: command-line
124
+ - path: ''
125
+ schema: structured
126
+ provider: schema
127
+ output_dir: /transformers/_benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False
128
+ choices:
129
+ backend: pytorch
130
+ launcher: process
131
+ scenario: inference
132
+ hydra/env: default
133
+ hydra/callbacks: null
134
+ hydra/job_logging: disabled
135
+ hydra/hydra_logging: disabled
136
+ hydra/hydra_help: default
137
+ hydra/help: default
138
+ hydra/sweeper: basic
139
+ hydra/launcher: basic
140
+ hydra/output: default
141
+ verbose: false
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/overrides.yaml ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ - backend.model=google/gemma-2b
2
+ - backend.cache_implementation=null
3
+ - backend.torch_compile=False
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.json ADDED
@@ -0,0 +1,473 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "name": "pytorch_generate",
4
+ "backend": {
5
+ "name": "pytorch",
6
+ "version": "2.4.0+cu121",
7
+ "_target_": "optimum_benchmark.backends.pytorch.backend.PyTorchBackend",
8
+ "task": "text-generation",
9
+ "library": "transformers",
10
+ "model_type": "gemma",
11
+ "model": "google/gemma-2b",
12
+ "processor": "google/gemma-2b",
13
+ "device": "cuda",
14
+ "device_ids": "0",
15
+ "seed": 42,
16
+ "inter_op_num_threads": null,
17
+ "intra_op_num_threads": null,
18
+ "model_kwargs": {},
19
+ "processor_kwargs": {},
20
+ "no_weights": true,
21
+ "device_map": null,
22
+ "torch_dtype": "float16",
23
+ "eval_mode": true,
24
+ "to_bettertransformer": false,
25
+ "low_cpu_mem_usage": null,
26
+ "attn_implementation": null,
27
+ "cache_implementation": null,
28
+ "autocast_enabled": false,
29
+ "autocast_dtype": null,
30
+ "torch_compile": false,
31
+ "torch_compile_target": "forward",
32
+ "torch_compile_config": {
33
+ "backend": "inductor",
34
+ "mode": "reduce-overhead",
35
+ "fullgraph": true
36
+ },
37
+ "quantization_scheme": null,
38
+ "quantization_config": {},
39
+ "deepspeed_inference": false,
40
+ "deepspeed_inference_config": {},
41
+ "peft_type": null,
42
+ "peft_config": {}
43
+ },
44
+ "scenario": {
45
+ "name": "inference",
46
+ "_target_": "optimum_benchmark.scenarios.inference.scenario.InferenceScenario",
47
+ "iterations": 2,
48
+ "duration": 0,
49
+ "warmup_runs": 10,
50
+ "input_shapes": {
51
+ "batch_size": 1,
52
+ "num_choices": 2,
53
+ "sequence_length": 7
54
+ },
55
+ "new_tokens": null,
56
+ "memory": true,
57
+ "latency": true,
58
+ "energy": false,
59
+ "forward_kwargs": {},
60
+ "generate_kwargs": {
61
+ "max_new_tokens": 128,
62
+ "min_new_tokens": 128,
63
+ "do_sample": false
64
+ },
65
+ "call_kwargs": {}
66
+ },
67
+ "launcher": {
68
+ "name": "process",
69
+ "_target_": "optimum_benchmark.launchers.process.launcher.ProcessLauncher",
70
+ "device_isolation": true,
71
+ "device_isolation_action": "warn",
72
+ "numactl": false,
73
+ "numactl_kwargs": {},
74
+ "start_method": "spawn"
75
+ },
76
+ "environment": {
77
+ "cpu": " AMD EPYC 7R32",
78
+ "cpu_count": 16,
79
+ "cpu_ram_mb": 66697.261056,
80
+ "system": "Linux",
81
+ "machine": "x86_64",
82
+ "platform": "Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29",
83
+ "processor": "x86_64",
84
+ "python_version": "3.8.10",
85
+ "gpu": [
86
+ "NVIDIA A10G"
87
+ ],
88
+ "gpu_count": 1,
89
+ "gpu_vram_mb": 24146608128,
90
+ "optimum_benchmark_version": "0.4.0",
91
+ "optimum_benchmark_commit": null,
92
+ "transformers_version": "4.45.0.dev0",
93
+ "transformers_commit": "65bb28444849976f853063edb958b3ef3dd59d12",
94
+ "accelerate_version": "0.35.0.dev0",
95
+ "accelerate_commit": null,
96
+ "diffusers_version": null,
97
+ "diffusers_commit": null,
98
+ "optimum_version": "1.22.0.dev0",
99
+ "optimum_commit": null,
100
+ "timm_version": "0.9.16",
101
+ "timm_commit": null,
102
+ "peft_version": "0.12.1.dev0",
103
+ "peft_commit": null
104
+ }
105
+ },
106
+ "report": {
107
+ "load": {
108
+ "memory": {
109
+ "unit": "MB",
110
+ "max_ram": 1335.021568,
111
+ "max_global_vram": 6775.373824,
112
+ "max_process_vram": 0.0,
113
+ "max_reserved": 6138.363904,
114
+ "max_allocated": 6060.931072
115
+ },
116
+ "latency": {
117
+ "unit": "s",
118
+ "count": 1,
119
+ "total": 11.54030078125,
120
+ "mean": 11.54030078125,
121
+ "stdev": 0.0,
122
+ "p50": 11.54030078125,
123
+ "p90": 11.54030078125,
124
+ "p95": 11.54030078125,
125
+ "p99": 11.54030078125,
126
+ "values": [
127
+ 11.54030078125
128
+ ]
129
+ },
130
+ "throughput": null,
131
+ "energy": null,
132
+ "efficiency": null
133
+ },
134
+ "prefill": {
135
+ "memory": {
136
+ "unit": "MB",
137
+ "max_ram": 1771.278336,
138
+ "max_global_vram": 6796.345344,
139
+ "max_process_vram": 0.0,
140
+ "max_reserved": 6142.558208,
141
+ "max_allocated": 5028.431872
142
+ },
143
+ "latency": {
144
+ "unit": "s",
145
+ "count": 2,
146
+ "total": 0.04255244827270508,
147
+ "mean": 0.02127622413635254,
148
+ "stdev": 0.0004439353942871094,
149
+ "p50": 0.02127622413635254,
150
+ "p90": 0.02163137245178223,
151
+ "p95": 0.02167576599121094,
152
+ "p99": 0.021711280822753906,
153
+ "values": [
154
+ 0.02083228874206543,
155
+ 0.02172015953063965
156
+ ]
157
+ },
158
+ "throughput": {
159
+ "unit": "tokens/s",
160
+ "value": 329.00574627994285
161
+ },
162
+ "energy": null,
163
+ "efficiency": null
164
+ },
165
+ "decode": {
166
+ "memory": {
167
+ "unit": "MB",
168
+ "max_ram": 1829.13024,
169
+ "max_global_vram": 6800.539648,
170
+ "max_process_vram": 0.0,
171
+ "max_reserved": 6146.752512,
172
+ "max_allocated": 5031.801344
173
+ },
174
+ "latency": {
175
+ "unit": "s",
176
+ "count": 2,
177
+ "total": 5.054838867187501,
178
+ "mean": 2.5274194335937503,
179
+ "stdev": 0.008804931640624902,
180
+ "p50": 2.5274194335937503,
181
+ "p90": 2.53446337890625,
182
+ "p95": 2.5353438720703125,
183
+ "p99": 2.5360482666015627,
184
+ "values": [
185
+ 2.518614501953125,
186
+ 2.536224365234375
187
+ ]
188
+ },
189
+ "throughput": {
190
+ "unit": "tokens/s",
191
+ "value": 50.248881650568805
192
+ },
193
+ "energy": null,
194
+ "efficiency": null
195
+ },
196
+ "per_token": {
197
+ "memory": null,
198
+ "latency": {
199
+ "unit": "s",
200
+ "count": 254,
201
+ "total": 5.054522373199463,
202
+ "mean": 0.01989969438267505,
203
+ "stdev": 0.00042909596730604846,
204
+ "p50": 0.0199552001953125,
205
+ "p90": 0.020370635986328125,
206
+ "p95": 0.020507484245300293,
207
+ "p99": 0.02102564880371094,
208
+ "values": [
209
+ 0.019333120346069335,
210
+ 0.019314687728881837,
211
+ 0.019346431732177736,
212
+ 0.01939967918395996,
213
+ 0.01946214485168457,
214
+ 0.019532800674438477,
215
+ 0.01942937660217285,
216
+ 0.01940275192260742,
217
+ 0.01938739204406738,
218
+ 0.019408895492553712,
219
+ 0.01939148712158203,
220
+ 0.01997209548950195,
221
+ 0.019555328369140625,
222
+ 0.019357696533203125,
223
+ 0.01940275192260742,
224
+ 0.019357696533203125,
225
+ 0.01940275192260742,
226
+ 0.01938742446899414,
227
+ 0.019435487747192382,
228
+ 0.019414016723632813,
229
+ 0.01945497512817383,
230
+ 0.01944063949584961,
231
+ 0.01949286460876465,
232
+ 0.01943756866455078,
233
+ 0.01944371223449707,
234
+ 0.01939455986022949,
235
+ 0.01942732810974121,
236
+ 0.02022809600830078,
237
+ 0.02026393508911133,
238
+ 0.019405824661254883,
239
+ 0.01944473648071289,
240
+ 0.019508224487304687,
241
+ 0.019382272720336914,
242
+ 0.019412992477416992,
243
+ 0.019787776947021486,
244
+ 0.020000768661499024,
245
+ 0.02018611145019531,
246
+ 0.01946009635925293,
247
+ 0.019367935180664063,
248
+ 0.019362815856933592,
249
+ 0.01944473648071289,
250
+ 0.019393535614013673,
251
+ 0.01941913604736328,
252
+ 0.019414016723632813,
253
+ 0.019404800415039062,
254
+ 0.02016972732543945,
255
+ 0.020093952178955078,
256
+ 0.019377151489257814,
257
+ 0.01940275192260742,
258
+ 0.019405824661254883,
259
+ 0.01939148712158203,
260
+ 0.019405824661254883,
261
+ 0.019392511367797852,
262
+ 0.01939148712158203,
263
+ 0.01938739204406738,
264
+ 0.01947238349914551,
265
+ 0.020206592559814454,
266
+ 0.02038374328613281,
267
+ 0.020134912490844727,
268
+ 0.02021990394592285,
269
+ 0.020196352005004883,
270
+ 0.02019430351257324,
271
+ 0.020357120513916017,
272
+ 0.02043801689147949,
273
+ 0.020754432678222655,
274
+ 0.02028339195251465,
275
+ 0.020249599456787108,
276
+ 0.020322303771972656,
277
+ 0.020387840270996094,
278
+ 0.01957683181762695,
279
+ 0.019529727935791014,
280
+ 0.019555328369140625,
281
+ 0.019532800674438477,
282
+ 0.0196177921295166,
283
+ 0.019531776428222656,
284
+ 0.019528703689575197,
285
+ 0.019533824920654298,
286
+ 0.019805183410644533,
287
+ 0.02110054397583008,
288
+ 0.020959232330322267,
289
+ 0.02037555122375488,
290
+ 0.020345855712890625,
291
+ 0.02024140739440918,
292
+ 0.020172800064086914,
293
+ 0.020609024047851563,
294
+ 0.02025369644165039,
295
+ 0.01960857582092285,
296
+ 0.01965977668762207,
297
+ 0.019510271072387696,
298
+ 0.019500032424926757,
299
+ 0.01962700843811035,
300
+ 0.019525632858276368,
301
+ 0.019728384017944335,
302
+ 0.019680255889892577,
303
+ 0.019725311279296876,
304
+ 0.019567615509033204,
305
+ 0.019562496185302734,
306
+ 0.019548160552978516,
307
+ 0.019594240188598632,
308
+ 0.019808256149291992,
309
+ 0.019507200241088866,
310
+ 0.01948057556152344,
311
+ 0.019558399200439454,
312
+ 0.019521535873413084,
313
+ 0.021210111618041993,
314
+ 0.020182016372680665,
315
+ 0.020162559509277343,
316
+ 0.020341760635375978,
317
+ 0.020320255279541014,
318
+ 0.02047385597229004,
319
+ 0.020230144500732423,
320
+ 0.02019430351257324,
321
+ 0.020384767532348632,
322
+ 0.02026700782775879,
323
+ 0.02027212715148926,
324
+ 0.020700159072875975,
325
+ 0.020864000320434572,
326
+ 0.020384767532348632,
327
+ 0.020290559768676757,
328
+ 0.020208639144897463,
329
+ 0.020287488937377928,
330
+ 0.02035916709899902,
331
+ 0.020274175643920898,
332
+ 0.02047385597229004,
333
+ 0.020313087463378905,
334
+ 0.02029363250732422,
335
+ 0.02028646469116211,
336
+ 0.020818944931030273,
337
+ 0.02006220817565918,
338
+ 0.02007244873046875,
339
+ 0.020161535263061522,
340
+ 0.020135936737060548,
341
+ 0.019761152267456054,
342
+ 0.019393535614013673,
343
+ 0.01943142318725586,
344
+ 0.01960960006713867,
345
+ 0.020074495315551756,
346
+ 0.020008960723876954,
347
+ 0.019713024139404296,
348
+ 0.020192256927490236,
349
+ 0.020184064865112306,
350
+ 0.020116479873657226,
351
+ 0.020151296615600587,
352
+ 0.020157440185546875,
353
+ 0.020154367446899413,
354
+ 0.020147199630737304,
355
+ 0.020108287811279296,
356
+ 0.020090879440307616,
357
+ 0.020117504119873047,
358
+ 0.020082687377929686,
359
+ 0.020032512664794923,
360
+ 0.020105215072631837,
361
+ 0.020017152786254884,
362
+ 0.02011136054992676,
363
+ 0.020151296615600587,
364
+ 0.01964646339416504,
365
+ 0.019595264434814453,
366
+ 0.019458047866821288,
367
+ 0.01943244743347168,
368
+ 0.020008960723876954,
369
+ 0.021184511184692383,
370
+ 0.020694015502929687,
371
+ 0.02025881576538086,
372
+ 0.02001919937133789,
373
+ 0.020149248123168945,
374
+ 0.020150272369384766,
375
+ 0.020384767532348632,
376
+ 0.020340736389160157,
377
+ 0.02026905632019043,
378
+ 0.020204544067382812,
379
+ 0.020171775817871093,
380
+ 0.02028441619873047,
381
+ 0.020172800064086914,
382
+ 0.020242431640625,
383
+ 0.020213760375976563,
384
+ 0.020146175384521483,
385
+ 0.020154367446899413,
386
+ 0.02046771240234375,
387
+ 0.020147199630737304,
388
+ 0.02007142448425293,
389
+ 0.020100095748901366,
390
+ 0.020264959335327147,
391
+ 0.02009702491760254,
392
+ 0.02008678436279297,
393
+ 0.02021785545349121,
394
+ 0.020339712142944336,
395
+ 0.019577856063842772,
396
+ 0.01958502388000488,
397
+ 0.019474431991577147,
398
+ 0.020552703857421875,
399
+ 0.019667999267578125,
400
+ 0.019559392929077147,
401
+ 0.019526655197143555,
402
+ 0.02015439987182617,
403
+ 0.02029052734375,
404
+ 0.020229120254516602,
405
+ 0.02024038314819336,
406
+ 0.02028339195251465,
407
+ 0.019528703689575197,
408
+ 0.01945395278930664,
409
+ 0.019539968490600586,
410
+ 0.019560447692871095,
411
+ 0.019578880310058593,
412
+ 0.019594240188598632,
413
+ 0.019546112060546874,
414
+ 0.020262912750244142,
415
+ 0.02031718444824219,
416
+ 0.020342784881591795,
417
+ 0.02032537651062012,
418
+ 0.02048313522338867,
419
+ 0.020289472579956055,
420
+ 0.02030284881591797,
421
+ 0.02028441619873047,
422
+ 0.020160512924194338,
423
+ 0.020178943634033202,
424
+ 0.020183040618896485,
425
+ 0.020288511276245116,
426
+ 0.020379648208618165,
427
+ 0.02003660774230957,
428
+ 0.019581951141357423,
429
+ 0.019535871505737306,
430
+ 0.0194703369140625,
431
+ 0.019582975387573243,
432
+ 0.019611648559570313,
433
+ 0.019583999633789064,
434
+ 0.019513343811035155,
435
+ 0.019516416549682617,
436
+ 0.020944896697998046,
437
+ 0.020271104812622072,
438
+ 0.02028441619873047,
439
+ 0.02041651153564453,
440
+ 0.020239360809326173,
441
+ 0.019938304901123048,
442
+ 0.02091007995605469,
443
+ 0.019565568923950196,
444
+ 0.01963315200805664,
445
+ 0.019494911193847657,
446
+ 0.019563520431518554,
447
+ 0.019500032424926757,
448
+ 0.019534847259521485,
449
+ 0.01964339256286621,
450
+ 0.019545087814331053,
451
+ 0.01948057556152344,
452
+ 0.019557376861572266,
453
+ 0.019495935440063478,
454
+ 0.019574783325195313,
455
+ 0.019511295318603517,
456
+ 0.019561471939086913,
457
+ 0.019506175994873046,
458
+ 0.019538944244384765,
459
+ 0.019542015075683594,
460
+ 0.019524608612060547,
461
+ 0.019490816116333007,
462
+ 0.019487743377685548
463
+ ]
464
+ },
465
+ "throughput": {
466
+ "unit": "tokens/s",
467
+ "value": 50.25202803469252
468
+ },
469
+ "energy": null,
470
+ "efficiency": null
471
+ }
472
+ }
473
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.log ADDED
File without changes
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "pytorch_generate",
3
+ "backend": {
4
+ "name": "pytorch",
5
+ "version": "2.4.0+cu121",
6
+ "_target_": "optimum_benchmark.backends.pytorch.backend.PyTorchBackend",
7
+ "task": "text-generation",
8
+ "library": "transformers",
9
+ "model_type": "gemma",
10
+ "model": "google/gemma-2b",
11
+ "processor": "google/gemma-2b",
12
+ "device": "cuda",
13
+ "device_ids": "0",
14
+ "seed": 42,
15
+ "inter_op_num_threads": null,
16
+ "intra_op_num_threads": null,
17
+ "model_kwargs": {},
18
+ "processor_kwargs": {},
19
+ "no_weights": true,
20
+ "device_map": null,
21
+ "torch_dtype": "float16",
22
+ "eval_mode": true,
23
+ "to_bettertransformer": false,
24
+ "low_cpu_mem_usage": null,
25
+ "attn_implementation": null,
26
+ "cache_implementation": null,
27
+ "autocast_enabled": false,
28
+ "autocast_dtype": null,
29
+ "torch_compile": false,
30
+ "torch_compile_target": "forward",
31
+ "torch_compile_config": {
32
+ "backend": "inductor",
33
+ "mode": "reduce-overhead",
34
+ "fullgraph": true
35
+ },
36
+ "quantization_scheme": null,
37
+ "quantization_config": {},
38
+ "deepspeed_inference": false,
39
+ "deepspeed_inference_config": {},
40
+ "peft_type": null,
41
+ "peft_config": {}
42
+ },
43
+ "scenario": {
44
+ "name": "inference",
45
+ "_target_": "optimum_benchmark.scenarios.inference.scenario.InferenceScenario",
46
+ "iterations": 2,
47
+ "duration": 0,
48
+ "warmup_runs": 10,
49
+ "input_shapes": {
50
+ "batch_size": 1,
51
+ "num_choices": 2,
52
+ "sequence_length": 7
53
+ },
54
+ "new_tokens": null,
55
+ "memory": true,
56
+ "latency": true,
57
+ "energy": false,
58
+ "forward_kwargs": {},
59
+ "generate_kwargs": {
60
+ "max_new_tokens": 128,
61
+ "min_new_tokens": 128,
62
+ "do_sample": false
63
+ },
64
+ "call_kwargs": {}
65
+ },
66
+ "launcher": {
67
+ "name": "process",
68
+ "_target_": "optimum_benchmark.launchers.process.launcher.ProcessLauncher",
69
+ "device_isolation": true,
70
+ "device_isolation_action": "warn",
71
+ "numactl": false,
72
+ "numactl_kwargs": {},
73
+ "start_method": "spawn"
74
+ },
75
+ "environment": {
76
+ "cpu": " AMD EPYC 7R32",
77
+ "cpu_count": 16,
78
+ "cpu_ram_mb": 66697.261056,
79
+ "system": "Linux",
80
+ "machine": "x86_64",
81
+ "platform": "Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29",
82
+ "processor": "x86_64",
83
+ "python_version": "3.8.10",
84
+ "gpu": [
85
+ "NVIDIA A10G"
86
+ ],
87
+ "gpu_count": 1,
88
+ "gpu_vram_mb": 24146608128,
89
+ "optimum_benchmark_version": "0.4.0",
90
+ "optimum_benchmark_commit": null,
91
+ "transformers_version": "4.45.0.dev0",
92
+ "transformers_commit": "65bb28444849976f853063edb958b3ef3dd59d12",
93
+ "accelerate_version": "0.35.0.dev0",
94
+ "accelerate_commit": null,
95
+ "diffusers_version": null,
96
+ "diffusers_commit": null,
97
+ "optimum_version": "1.22.0.dev0",
98
+ "optimum_commit": null,
99
+ "timm_version": "0.9.16",
100
+ "timm_commit": null,
101
+ "peft_version": "0.12.1.dev0",
102
+ "peft_commit": null
103
+ }
104
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_report.json ADDED
@@ -0,0 +1,367 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "load": {
3
+ "memory": {
4
+ "unit": "MB",
5
+ "max_ram": 1335.021568,
6
+ "max_global_vram": 6775.373824,
7
+ "max_process_vram": 0.0,
8
+ "max_reserved": 6138.363904,
9
+ "max_allocated": 6060.931072
10
+ },
11
+ "latency": {
12
+ "unit": "s",
13
+ "count": 1,
14
+ "total": 11.54030078125,
15
+ "mean": 11.54030078125,
16
+ "stdev": 0.0,
17
+ "p50": 11.54030078125,
18
+ "p90": 11.54030078125,
19
+ "p95": 11.54030078125,
20
+ "p99": 11.54030078125,
21
+ "values": [
22
+ 11.54030078125
23
+ ]
24
+ },
25
+ "throughput": null,
26
+ "energy": null,
27
+ "efficiency": null
28
+ },
29
+ "prefill": {
30
+ "memory": {
31
+ "unit": "MB",
32
+ "max_ram": 1771.278336,
33
+ "max_global_vram": 6796.345344,
34
+ "max_process_vram": 0.0,
35
+ "max_reserved": 6142.558208,
36
+ "max_allocated": 5028.431872
37
+ },
38
+ "latency": {
39
+ "unit": "s",
40
+ "count": 2,
41
+ "total": 0.04255244827270508,
42
+ "mean": 0.02127622413635254,
43
+ "stdev": 0.0004439353942871094,
44
+ "p50": 0.02127622413635254,
45
+ "p90": 0.02163137245178223,
46
+ "p95": 0.02167576599121094,
47
+ "p99": 0.021711280822753906,
48
+ "values": [
49
+ 0.02083228874206543,
50
+ 0.02172015953063965
51
+ ]
52
+ },
53
+ "throughput": {
54
+ "unit": "tokens/s",
55
+ "value": 329.00574627994285
56
+ },
57
+ "energy": null,
58
+ "efficiency": null
59
+ },
60
+ "decode": {
61
+ "memory": {
62
+ "unit": "MB",
63
+ "max_ram": 1829.13024,
64
+ "max_global_vram": 6800.539648,
65
+ "max_process_vram": 0.0,
66
+ "max_reserved": 6146.752512,
67
+ "max_allocated": 5031.801344
68
+ },
69
+ "latency": {
70
+ "unit": "s",
71
+ "count": 2,
72
+ "total": 5.054838867187501,
73
+ "mean": 2.5274194335937503,
74
+ "stdev": 0.008804931640624902,
75
+ "p50": 2.5274194335937503,
76
+ "p90": 2.53446337890625,
77
+ "p95": 2.5353438720703125,
78
+ "p99": 2.5360482666015627,
79
+ "values": [
80
+ 2.518614501953125,
81
+ 2.536224365234375
82
+ ]
83
+ },
84
+ "throughput": {
85
+ "unit": "tokens/s",
86
+ "value": 50.248881650568805
87
+ },
88
+ "energy": null,
89
+ "efficiency": null
90
+ },
91
+ "per_token": {
92
+ "memory": null,
93
+ "latency": {
94
+ "unit": "s",
95
+ "count": 254,
96
+ "total": 5.054522373199463,
97
+ "mean": 0.01989969438267505,
98
+ "stdev": 0.00042909596730604846,
99
+ "p50": 0.0199552001953125,
100
+ "p90": 0.020370635986328125,
101
+ "p95": 0.020507484245300293,
102
+ "p99": 0.02102564880371094,
103
+ "values": [
104
+ 0.019333120346069335,
105
+ 0.019314687728881837,
106
+ 0.019346431732177736,
107
+ 0.01939967918395996,
108
+ 0.01946214485168457,
109
+ 0.019532800674438477,
110
+ 0.01942937660217285,
111
+ 0.01940275192260742,
112
+ 0.01938739204406738,
113
+ 0.019408895492553712,
114
+ 0.01939148712158203,
115
+ 0.01997209548950195,
116
+ 0.019555328369140625,
117
+ 0.019357696533203125,
118
+ 0.01940275192260742,
119
+ 0.019357696533203125,
120
+ 0.01940275192260742,
121
+ 0.01938742446899414,
122
+ 0.019435487747192382,
123
+ 0.019414016723632813,
124
+ 0.01945497512817383,
125
+ 0.01944063949584961,
126
+ 0.01949286460876465,
127
+ 0.01943756866455078,
128
+ 0.01944371223449707,
129
+ 0.01939455986022949,
130
+ 0.01942732810974121,
131
+ 0.02022809600830078,
132
+ 0.02026393508911133,
133
+ 0.019405824661254883,
134
+ 0.01944473648071289,
135
+ 0.019508224487304687,
136
+ 0.019382272720336914,
137
+ 0.019412992477416992,
138
+ 0.019787776947021486,
139
+ 0.020000768661499024,
140
+ 0.02018611145019531,
141
+ 0.01946009635925293,
142
+ 0.019367935180664063,
143
+ 0.019362815856933592,
144
+ 0.01944473648071289,
145
+ 0.019393535614013673,
146
+ 0.01941913604736328,
147
+ 0.019414016723632813,
148
+ 0.019404800415039062,
149
+ 0.02016972732543945,
150
+ 0.020093952178955078,
151
+ 0.019377151489257814,
152
+ 0.01940275192260742,
153
+ 0.019405824661254883,
154
+ 0.01939148712158203,
155
+ 0.019405824661254883,
156
+ 0.019392511367797852,
157
+ 0.01939148712158203,
158
+ 0.01938739204406738,
159
+ 0.01947238349914551,
160
+ 0.020206592559814454,
161
+ 0.02038374328613281,
162
+ 0.020134912490844727,
163
+ 0.02021990394592285,
164
+ 0.020196352005004883,
165
+ 0.02019430351257324,
166
+ 0.020357120513916017,
167
+ 0.02043801689147949,
168
+ 0.020754432678222655,
169
+ 0.02028339195251465,
170
+ 0.020249599456787108,
171
+ 0.020322303771972656,
172
+ 0.020387840270996094,
173
+ 0.01957683181762695,
174
+ 0.019529727935791014,
175
+ 0.019555328369140625,
176
+ 0.019532800674438477,
177
+ 0.0196177921295166,
178
+ 0.019531776428222656,
179
+ 0.019528703689575197,
180
+ 0.019533824920654298,
181
+ 0.019805183410644533,
182
+ 0.02110054397583008,
183
+ 0.020959232330322267,
184
+ 0.02037555122375488,
185
+ 0.020345855712890625,
186
+ 0.02024140739440918,
187
+ 0.020172800064086914,
188
+ 0.020609024047851563,
189
+ 0.02025369644165039,
190
+ 0.01960857582092285,
191
+ 0.01965977668762207,
192
+ 0.019510271072387696,
193
+ 0.019500032424926757,
194
+ 0.01962700843811035,
195
+ 0.019525632858276368,
196
+ 0.019728384017944335,
197
+ 0.019680255889892577,
198
+ 0.019725311279296876,
199
+ 0.019567615509033204,
200
+ 0.019562496185302734,
201
+ 0.019548160552978516,
202
+ 0.019594240188598632,
203
+ 0.019808256149291992,
204
+ 0.019507200241088866,
205
+ 0.01948057556152344,
206
+ 0.019558399200439454,
207
+ 0.019521535873413084,
208
+ 0.021210111618041993,
209
+ 0.020182016372680665,
210
+ 0.020162559509277343,
211
+ 0.020341760635375978,
212
+ 0.020320255279541014,
213
+ 0.02047385597229004,
214
+ 0.020230144500732423,
215
+ 0.02019430351257324,
216
+ 0.020384767532348632,
217
+ 0.02026700782775879,
218
+ 0.02027212715148926,
219
+ 0.020700159072875975,
220
+ 0.020864000320434572,
221
+ 0.020384767532348632,
222
+ 0.020290559768676757,
223
+ 0.020208639144897463,
224
+ 0.020287488937377928,
225
+ 0.02035916709899902,
226
+ 0.020274175643920898,
227
+ 0.02047385597229004,
228
+ 0.020313087463378905,
229
+ 0.02029363250732422,
230
+ 0.02028646469116211,
231
+ 0.020818944931030273,
232
+ 0.02006220817565918,
233
+ 0.02007244873046875,
234
+ 0.020161535263061522,
235
+ 0.020135936737060548,
236
+ 0.019761152267456054,
237
+ 0.019393535614013673,
238
+ 0.01943142318725586,
239
+ 0.01960960006713867,
240
+ 0.020074495315551756,
241
+ 0.020008960723876954,
242
+ 0.019713024139404296,
243
+ 0.020192256927490236,
244
+ 0.020184064865112306,
245
+ 0.020116479873657226,
246
+ 0.020151296615600587,
247
+ 0.020157440185546875,
248
+ 0.020154367446899413,
249
+ 0.020147199630737304,
250
+ 0.020108287811279296,
251
+ 0.020090879440307616,
252
+ 0.020117504119873047,
253
+ 0.020082687377929686,
254
+ 0.020032512664794923,
255
+ 0.020105215072631837,
256
+ 0.020017152786254884,
257
+ 0.02011136054992676,
258
+ 0.020151296615600587,
259
+ 0.01964646339416504,
260
+ 0.019595264434814453,
261
+ 0.019458047866821288,
262
+ 0.01943244743347168,
263
+ 0.020008960723876954,
264
+ 0.021184511184692383,
265
+ 0.020694015502929687,
266
+ 0.02025881576538086,
267
+ 0.02001919937133789,
268
+ 0.020149248123168945,
269
+ 0.020150272369384766,
270
+ 0.020384767532348632,
271
+ 0.020340736389160157,
272
+ 0.02026905632019043,
273
+ 0.020204544067382812,
274
+ 0.020171775817871093,
275
+ 0.02028441619873047,
276
+ 0.020172800064086914,
277
+ 0.020242431640625,
278
+ 0.020213760375976563,
279
+ 0.020146175384521483,
280
+ 0.020154367446899413,
281
+ 0.02046771240234375,
282
+ 0.020147199630737304,
283
+ 0.02007142448425293,
284
+ 0.020100095748901366,
285
+ 0.020264959335327147,
286
+ 0.02009702491760254,
287
+ 0.02008678436279297,
288
+ 0.02021785545349121,
289
+ 0.020339712142944336,
290
+ 0.019577856063842772,
291
+ 0.01958502388000488,
292
+ 0.019474431991577147,
293
+ 0.020552703857421875,
294
+ 0.019667999267578125,
295
+ 0.019559392929077147,
296
+ 0.019526655197143555,
297
+ 0.02015439987182617,
298
+ 0.02029052734375,
299
+ 0.020229120254516602,
300
+ 0.02024038314819336,
301
+ 0.02028339195251465,
302
+ 0.019528703689575197,
303
+ 0.01945395278930664,
304
+ 0.019539968490600586,
305
+ 0.019560447692871095,
306
+ 0.019578880310058593,
307
+ 0.019594240188598632,
308
+ 0.019546112060546874,
309
+ 0.020262912750244142,
310
+ 0.02031718444824219,
311
+ 0.020342784881591795,
312
+ 0.02032537651062012,
313
+ 0.02048313522338867,
314
+ 0.020289472579956055,
315
+ 0.02030284881591797,
316
+ 0.02028441619873047,
317
+ 0.020160512924194338,
318
+ 0.020178943634033202,
319
+ 0.020183040618896485,
320
+ 0.020288511276245116,
321
+ 0.020379648208618165,
322
+ 0.02003660774230957,
323
+ 0.019581951141357423,
324
+ 0.019535871505737306,
325
+ 0.0194703369140625,
326
+ 0.019582975387573243,
327
+ 0.019611648559570313,
328
+ 0.019583999633789064,
329
+ 0.019513343811035155,
330
+ 0.019516416549682617,
331
+ 0.020944896697998046,
332
+ 0.020271104812622072,
333
+ 0.02028441619873047,
334
+ 0.02041651153564453,
335
+ 0.020239360809326173,
336
+ 0.019938304901123048,
337
+ 0.02091007995605469,
338
+ 0.019565568923950196,
339
+ 0.01963315200805664,
340
+ 0.019494911193847657,
341
+ 0.019563520431518554,
342
+ 0.019500032424926757,
343
+ 0.019534847259521485,
344
+ 0.01964339256286621,
345
+ 0.019545087814331053,
346
+ 0.01948057556152344,
347
+ 0.019557376861572266,
348
+ 0.019495935440063478,
349
+ 0.019574783325195313,
350
+ 0.019511295318603517,
351
+ 0.019561471939086913,
352
+ 0.019506175994873046,
353
+ 0.019538944244384765,
354
+ 0.019542015075683594,
355
+ 0.019524608612060547,
356
+ 0.019490816116333007,
357
+ 0.019487743377685548
358
+ ]
359
+ },
360
+ "throughput": {
361
+ "unit": "tokens/s",
362
+ "value": 50.25202803469252
363
+ },
364
+ "energy": null,
365
+ "efficiency": null
366
+ }
367
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=False/summary.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "google/gemma-2b",
3
+ "commit": "65bb28444849976f853063edb958b3ef3dd59d12",
4
+ "config": "backend.cache_implementation=null,backend.torch_compile=False",
5
+ "metrics": {
6
+ "prefill.latency.mean": 0.02127622413635254,
7
+ "prefill.throughput.value": 329.00574627994285,
8
+ "decode.latency.mean": 2.5274194335937503,
9
+ "decode.throughput.value": 50.248881650568805,
10
+ "per_token.latency.mean": 0.01989969438267505,
11
+ "per_token.throughput.value": 50.25202803469252
12
+ }
13
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/config.yaml ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: pytorch_generate
2
+ backend:
3
+ name: pytorch
4
+ version: 2.4.0+cu121
5
+ _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
6
+ task: null
7
+ library: null
8
+ model_type: null
9
+ model: google/gemma-2b
10
+ processor: null
11
+ device: cuda
12
+ device_ids: '0'
13
+ seed: 42
14
+ inter_op_num_threads: null
15
+ intra_op_num_threads: null
16
+ model_kwargs: {}
17
+ processor_kwargs: {}
18
+ no_weights: true
19
+ device_map: null
20
+ torch_dtype: float16
21
+ eval_mode: true
22
+ to_bettertransformer: false
23
+ low_cpu_mem_usage: null
24
+ attn_implementation: null
25
+ cache_implementation: null
26
+ autocast_enabled: false
27
+ autocast_dtype: null
28
+ torch_compile: true
29
+ torch_compile_target: forward
30
+ torch_compile_config:
31
+ backend: inductor
32
+ mode: reduce-overhead
33
+ fullgraph: true
34
+ quantization_scheme: null
35
+ quantization_config: {}
36
+ deepspeed_inference: false
37
+ deepspeed_inference_config: {}
38
+ peft_type: null
39
+ peft_config: {}
40
+ scenario:
41
+ name: inference
42
+ _target_: optimum_benchmark.scenarios.inference.scenario.InferenceScenario
43
+ iterations: 2
44
+ duration: 0
45
+ warmup_runs: 10
46
+ input_shapes:
47
+ batch_size: 1
48
+ sequence_length: 7
49
+ new_tokens: null
50
+ memory: true
51
+ latency: true
52
+ energy: false
53
+ forward_kwargs: {}
54
+ generate_kwargs:
55
+ max_new_tokens: 128
56
+ min_new_tokens: 128
57
+ do_sample: false
58
+ call_kwargs: {}
59
+ launcher:
60
+ name: process
61
+ _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
62
+ device_isolation: true
63
+ device_isolation_action: warn
64
+ numactl: false
65
+ numactl_kwargs: {}
66
+ start_method: spawn
67
+ environment:
68
+ cpu: ' AMD EPYC 7R32'
69
+ cpu_count: 16
70
+ cpu_ram_mb: 66697.261056
71
+ system: Linux
72
+ machine: x86_64
73
+ platform: Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29
74
+ processor: x86_64
75
+ python_version: 3.8.10
76
+ gpu:
77
+ - NVIDIA A10G
78
+ gpu_count: 1
79
+ gpu_vram_mb: 24146608128
80
+ optimum_benchmark_version: 0.4.0
81
+ optimum_benchmark_commit: null
82
+ transformers_version: 4.45.0.dev0
83
+ transformers_commit: 65bb28444849976f853063edb958b3ef3dd59d12
84
+ accelerate_version: 0.35.0.dev0
85
+ accelerate_commit: null
86
+ diffusers_version: null
87
+ diffusers_commit: null
88
+ optimum_version: 1.22.0.dev0
89
+ optimum_commit: null
90
+ timm_version: 0.9.16
91
+ timm_commit: null
92
+ peft_version: 0.12.1.dev0
93
+ peft_commit: null
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/hydra.yaml ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
4
+ sweep:
5
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12
6
+ subdir: ${hydra.job.override_dirname}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ root:
71
+ level: ERROR
72
+ disable_existing_loggers: true
73
+ job_logging:
74
+ version: 1
75
+ root:
76
+ level: ERROR
77
+ disable_existing_loggers: true
78
+ env: {}
79
+ mode: MULTIRUN
80
+ searchpath: []
81
+ callbacks: {}
82
+ output_subdir: .hydra
83
+ overrides:
84
+ hydra:
85
+ - hydra/job_logging=disabled
86
+ - hydra/hydra_logging=disabled
87
+ - hydra.sweep.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12
88
+ - hydra.run.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
89
+ - hydra.mode=MULTIRUN
90
+ task:
91
+ - backend.model=google/gemma-2b
92
+ - backend.cache_implementation=null
93
+ - backend.torch_compile=True
94
+ job:
95
+ name: cli
96
+ chdir: true
97
+ override_dirname: backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True
98
+ id: '1'
99
+ num: 1
100
+ config_name: generation
101
+ env_set:
102
+ OVERRIDE_BENCHMARKS: '1'
103
+ LOG_LEVEL: WARN
104
+ env_copy: []
105
+ config:
106
+ override_dirname:
107
+ kv_sep: '='
108
+ item_sep: ','
109
+ exclude_keys: []
110
+ runtime:
111
+ version: 1.3.2
112
+ version_base: '1.3'
113
+ cwd: /transformers
114
+ config_sources:
115
+ - path: hydra.conf
116
+ schema: pkg
117
+ provider: hydra
118
+ - path: optimum_benchmark
119
+ schema: pkg
120
+ provider: main
121
+ - path: /transformers/benchmark/config
122
+ schema: file
123
+ provider: command-line
124
+ - path: ''
125
+ schema: structured
126
+ provider: schema
127
+ output_dir: /transformers/_benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True
128
+ choices:
129
+ backend: pytorch
130
+ launcher: process
131
+ scenario: inference
132
+ hydra/env: default
133
+ hydra/callbacks: null
134
+ hydra/job_logging: disabled
135
+ hydra/hydra_logging: disabled
136
+ hydra/hydra_help: default
137
+ hydra/help: default
138
+ hydra/sweeper: basic
139
+ hydra/launcher: basic
140
+ hydra/output: default
141
+ verbose: false
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/overrides.yaml ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ - backend.model=google/gemma-2b
2
+ - backend.cache_implementation=null
3
+ - backend.torch_compile=True
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark.log ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ [ISOLATED-PROCESS][2024-09-09 09:07:51,876][process][ERROR] - + Sending traceback to main process
2
+ [MAIN-PROCESS][2024-09-09 09:07:54,281][process][ERROR] - + Received traceback from isolated process
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=null,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark_config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "pytorch_generate",
3
+ "backend": {
4
+ "name": "pytorch",
5
+ "version": "2.4.0+cu121",
6
+ "_target_": "optimum_benchmark.backends.pytorch.backend.PyTorchBackend",
7
+ "task": "text-generation",
8
+ "library": "transformers",
9
+ "model_type": "gemma",
10
+ "model": "google/gemma-2b",
11
+ "processor": "google/gemma-2b",
12
+ "device": "cuda",
13
+ "device_ids": "0",
14
+ "seed": 42,
15
+ "inter_op_num_threads": null,
16
+ "intra_op_num_threads": null,
17
+ "model_kwargs": {},
18
+ "processor_kwargs": {},
19
+ "no_weights": true,
20
+ "device_map": null,
21
+ "torch_dtype": "float16",
22
+ "eval_mode": true,
23
+ "to_bettertransformer": false,
24
+ "low_cpu_mem_usage": null,
25
+ "attn_implementation": null,
26
+ "cache_implementation": null,
27
+ "autocast_enabled": false,
28
+ "autocast_dtype": null,
29
+ "torch_compile": true,
30
+ "torch_compile_target": "forward",
31
+ "torch_compile_config": {
32
+ "backend": "inductor",
33
+ "mode": "reduce-overhead",
34
+ "fullgraph": true
35
+ },
36
+ "quantization_scheme": null,
37
+ "quantization_config": {},
38
+ "deepspeed_inference": false,
39
+ "deepspeed_inference_config": {},
40
+ "peft_type": null,
41
+ "peft_config": {}
42
+ },
43
+ "scenario": {
44
+ "name": "inference",
45
+ "_target_": "optimum_benchmark.scenarios.inference.scenario.InferenceScenario",
46
+ "iterations": 2,
47
+ "duration": 0,
48
+ "warmup_runs": 10,
49
+ "input_shapes": {
50
+ "batch_size": 1,
51
+ "num_choices": 2,
52
+ "sequence_length": 7
53
+ },
54
+ "new_tokens": null,
55
+ "memory": true,
56
+ "latency": true,
57
+ "energy": false,
58
+ "forward_kwargs": {},
59
+ "generate_kwargs": {
60
+ "max_new_tokens": 128,
61
+ "min_new_tokens": 128,
62
+ "do_sample": false
63
+ },
64
+ "call_kwargs": {}
65
+ },
66
+ "launcher": {
67
+ "name": "process",
68
+ "_target_": "optimum_benchmark.launchers.process.launcher.ProcessLauncher",
69
+ "device_isolation": true,
70
+ "device_isolation_action": "warn",
71
+ "numactl": false,
72
+ "numactl_kwargs": {},
73
+ "start_method": "spawn"
74
+ },
75
+ "environment": {
76
+ "cpu": " AMD EPYC 7R32",
77
+ "cpu_count": 16,
78
+ "cpu_ram_mb": 66697.261056,
79
+ "system": "Linux",
80
+ "machine": "x86_64",
81
+ "platform": "Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29",
82
+ "processor": "x86_64",
83
+ "python_version": "3.8.10",
84
+ "gpu": [
85
+ "NVIDIA A10G"
86
+ ],
87
+ "gpu_count": 1,
88
+ "gpu_vram_mb": 24146608128,
89
+ "optimum_benchmark_version": "0.4.0",
90
+ "optimum_benchmark_commit": null,
91
+ "transformers_version": "4.45.0.dev0",
92
+ "transformers_commit": "65bb28444849976f853063edb958b3ef3dd59d12",
93
+ "accelerate_version": "0.35.0.dev0",
94
+ "accelerate_commit": null,
95
+ "diffusers_version": null,
96
+ "diffusers_commit": null,
97
+ "optimum_version": "1.22.0.dev0",
98
+ "optimum_commit": null,
99
+ "timm_version": "0.9.16",
100
+ "timm_commit": null,
101
+ "peft_version": "0.12.1.dev0",
102
+ "peft_commit": null
103
+ }
104
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/config.yaml ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: pytorch_generate
2
+ backend:
3
+ name: pytorch
4
+ version: 2.4.0+cu121
5
+ _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
6
+ task: null
7
+ library: null
8
+ model_type: null
9
+ model: google/gemma-2b
10
+ processor: null
11
+ device: cuda
12
+ device_ids: '0'
13
+ seed: 42
14
+ inter_op_num_threads: null
15
+ intra_op_num_threads: null
16
+ model_kwargs: {}
17
+ processor_kwargs: {}
18
+ no_weights: true
19
+ device_map: null
20
+ torch_dtype: float16
21
+ eval_mode: true
22
+ to_bettertransformer: false
23
+ low_cpu_mem_usage: null
24
+ attn_implementation: null
25
+ cache_implementation: static
26
+ autocast_enabled: false
27
+ autocast_dtype: null
28
+ torch_compile: false
29
+ torch_compile_target: forward
30
+ torch_compile_config:
31
+ backend: inductor
32
+ mode: reduce-overhead
33
+ fullgraph: true
34
+ quantization_scheme: null
35
+ quantization_config: {}
36
+ deepspeed_inference: false
37
+ deepspeed_inference_config: {}
38
+ peft_type: null
39
+ peft_config: {}
40
+ scenario:
41
+ name: inference
42
+ _target_: optimum_benchmark.scenarios.inference.scenario.InferenceScenario
43
+ iterations: 2
44
+ duration: 0
45
+ warmup_runs: 10
46
+ input_shapes:
47
+ batch_size: 1
48
+ sequence_length: 7
49
+ new_tokens: null
50
+ memory: true
51
+ latency: true
52
+ energy: false
53
+ forward_kwargs: {}
54
+ generate_kwargs:
55
+ max_new_tokens: 128
56
+ min_new_tokens: 128
57
+ do_sample: false
58
+ call_kwargs: {}
59
+ launcher:
60
+ name: process
61
+ _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
62
+ device_isolation: true
63
+ device_isolation_action: warn
64
+ numactl: false
65
+ numactl_kwargs: {}
66
+ start_method: spawn
67
+ environment:
68
+ cpu: ' AMD EPYC 7R32'
69
+ cpu_count: 16
70
+ cpu_ram_mb: 66697.261056
71
+ system: Linux
72
+ machine: x86_64
73
+ platform: Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29
74
+ processor: x86_64
75
+ python_version: 3.8.10
76
+ gpu:
77
+ - NVIDIA A10G
78
+ gpu_count: 1
79
+ gpu_vram_mb: 24146608128
80
+ optimum_benchmark_version: 0.4.0
81
+ optimum_benchmark_commit: null
82
+ transformers_version: 4.45.0.dev0
83
+ transformers_commit: 65bb28444849976f853063edb958b3ef3dd59d12
84
+ accelerate_version: 0.35.0.dev0
85
+ accelerate_commit: null
86
+ diffusers_version: null
87
+ diffusers_commit: null
88
+ optimum_version: 1.22.0.dev0
89
+ optimum_commit: null
90
+ timm_version: 0.9.16
91
+ timm_commit: null
92
+ peft_version: 0.12.1.dev0
93
+ peft_commit: null
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/hydra.yaml ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
4
+ sweep:
5
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12
6
+ subdir: ${hydra.job.override_dirname}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ root:
71
+ level: ERROR
72
+ disable_existing_loggers: true
73
+ job_logging:
74
+ version: 1
75
+ root:
76
+ level: ERROR
77
+ disable_existing_loggers: true
78
+ env: {}
79
+ mode: MULTIRUN
80
+ searchpath: []
81
+ callbacks: {}
82
+ output_subdir: .hydra
83
+ overrides:
84
+ hydra:
85
+ - hydra/job_logging=disabled
86
+ - hydra/hydra_logging=disabled
87
+ - hydra.sweep.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12
88
+ - hydra.run.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
89
+ - hydra.mode=MULTIRUN
90
+ task:
91
+ - backend.model=google/gemma-2b
92
+ - backend.cache_implementation=static
93
+ - backend.torch_compile=False
94
+ job:
95
+ name: cli
96
+ chdir: true
97
+ override_dirname: backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False
98
+ id: '2'
99
+ num: 2
100
+ config_name: generation
101
+ env_set:
102
+ OVERRIDE_BENCHMARKS: '1'
103
+ LOG_LEVEL: WARN
104
+ env_copy: []
105
+ config:
106
+ override_dirname:
107
+ kv_sep: '='
108
+ item_sep: ','
109
+ exclude_keys: []
110
+ runtime:
111
+ version: 1.3.2
112
+ version_base: '1.3'
113
+ cwd: /transformers
114
+ config_sources:
115
+ - path: hydra.conf
116
+ schema: pkg
117
+ provider: hydra
118
+ - path: optimum_benchmark
119
+ schema: pkg
120
+ provider: main
121
+ - path: /transformers/benchmark/config
122
+ schema: file
123
+ provider: command-line
124
+ - path: ''
125
+ schema: structured
126
+ provider: schema
127
+ output_dir: /transformers/_benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False
128
+ choices:
129
+ backend: pytorch
130
+ launcher: process
131
+ scenario: inference
132
+ hydra/env: default
133
+ hydra/callbacks: null
134
+ hydra/job_logging: disabled
135
+ hydra/hydra_logging: disabled
136
+ hydra/hydra_help: default
137
+ hydra/help: default
138
+ hydra/sweeper: basic
139
+ hydra/launcher: basic
140
+ hydra/output: default
141
+ verbose: false
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/.hydra/overrides.yaml ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ - backend.model=google/gemma-2b
2
+ - backend.cache_implementation=static
3
+ - backend.torch_compile=False
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.json ADDED
@@ -0,0 +1,473 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "name": "pytorch_generate",
4
+ "backend": {
5
+ "name": "pytorch",
6
+ "version": "2.4.0+cu121",
7
+ "_target_": "optimum_benchmark.backends.pytorch.backend.PyTorchBackend",
8
+ "task": "text-generation",
9
+ "library": "transformers",
10
+ "model_type": "gemma",
11
+ "model": "google/gemma-2b",
12
+ "processor": "google/gemma-2b",
13
+ "device": "cuda",
14
+ "device_ids": "0",
15
+ "seed": 42,
16
+ "inter_op_num_threads": null,
17
+ "intra_op_num_threads": null,
18
+ "model_kwargs": {},
19
+ "processor_kwargs": {},
20
+ "no_weights": true,
21
+ "device_map": null,
22
+ "torch_dtype": "float16",
23
+ "eval_mode": true,
24
+ "to_bettertransformer": false,
25
+ "low_cpu_mem_usage": null,
26
+ "attn_implementation": null,
27
+ "cache_implementation": "static",
28
+ "autocast_enabled": false,
29
+ "autocast_dtype": null,
30
+ "torch_compile": false,
31
+ "torch_compile_target": "forward",
32
+ "torch_compile_config": {
33
+ "backend": "inductor",
34
+ "mode": "reduce-overhead",
35
+ "fullgraph": true
36
+ },
37
+ "quantization_scheme": null,
38
+ "quantization_config": {},
39
+ "deepspeed_inference": false,
40
+ "deepspeed_inference_config": {},
41
+ "peft_type": null,
42
+ "peft_config": {}
43
+ },
44
+ "scenario": {
45
+ "name": "inference",
46
+ "_target_": "optimum_benchmark.scenarios.inference.scenario.InferenceScenario",
47
+ "iterations": 2,
48
+ "duration": 0,
49
+ "warmup_runs": 10,
50
+ "input_shapes": {
51
+ "batch_size": 1,
52
+ "num_choices": 2,
53
+ "sequence_length": 7
54
+ },
55
+ "new_tokens": null,
56
+ "memory": true,
57
+ "latency": true,
58
+ "energy": false,
59
+ "forward_kwargs": {},
60
+ "generate_kwargs": {
61
+ "max_new_tokens": 128,
62
+ "min_new_tokens": 128,
63
+ "do_sample": false
64
+ },
65
+ "call_kwargs": {}
66
+ },
67
+ "launcher": {
68
+ "name": "process",
69
+ "_target_": "optimum_benchmark.launchers.process.launcher.ProcessLauncher",
70
+ "device_isolation": true,
71
+ "device_isolation_action": "warn",
72
+ "numactl": false,
73
+ "numactl_kwargs": {},
74
+ "start_method": "spawn"
75
+ },
76
+ "environment": {
77
+ "cpu": " AMD EPYC 7R32",
78
+ "cpu_count": 16,
79
+ "cpu_ram_mb": 66697.261056,
80
+ "system": "Linux",
81
+ "machine": "x86_64",
82
+ "platform": "Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29",
83
+ "processor": "x86_64",
84
+ "python_version": "3.8.10",
85
+ "gpu": [
86
+ "NVIDIA A10G"
87
+ ],
88
+ "gpu_count": 1,
89
+ "gpu_vram_mb": 24146608128,
90
+ "optimum_benchmark_version": "0.4.0",
91
+ "optimum_benchmark_commit": null,
92
+ "transformers_version": "4.45.0.dev0",
93
+ "transformers_commit": "65bb28444849976f853063edb958b3ef3dd59d12",
94
+ "accelerate_version": "0.35.0.dev0",
95
+ "accelerate_commit": null,
96
+ "diffusers_version": null,
97
+ "diffusers_commit": null,
98
+ "optimum_version": "1.22.0.dev0",
99
+ "optimum_commit": null,
100
+ "timm_version": "0.9.16",
101
+ "timm_commit": null,
102
+ "peft_version": "0.12.1.dev0",
103
+ "peft_commit": null
104
+ }
105
+ },
106
+ "report": {
107
+ "load": {
108
+ "memory": {
109
+ "unit": "MB",
110
+ "max_ram": 1336.307712,
111
+ "max_global_vram": 6775.373824,
112
+ "max_process_vram": 0.0,
113
+ "max_reserved": 6138.363904,
114
+ "max_allocated": 6060.931072
115
+ },
116
+ "latency": {
117
+ "unit": "s",
118
+ "count": 1,
119
+ "total": 11.531244140625,
120
+ "mean": 11.531244140625,
121
+ "stdev": 0.0,
122
+ "p50": 11.531244140625,
123
+ "p90": 11.531244140625,
124
+ "p95": 11.531244140625,
125
+ "p99": 11.531244140625,
126
+ "values": [
127
+ 11.531244140625
128
+ ]
129
+ },
130
+ "throughput": null,
131
+ "energy": null,
132
+ "efficiency": null
133
+ },
134
+ "prefill": {
135
+ "memory": {
136
+ "unit": "MB",
137
+ "max_ram": 1815.650304,
138
+ "max_global_vram": 6796.345344,
139
+ "max_process_vram": 0.0,
140
+ "max_reserved": 6142.558208,
141
+ "max_allocated": 5028.450816
142
+ },
143
+ "latency": {
144
+ "unit": "s",
145
+ "count": 2,
146
+ "total": 0.043629503250122076,
147
+ "mean": 0.021814751625061038,
148
+ "stdev": 5.913639068603517e-05,
149
+ "p50": 0.021814751625061038,
150
+ "p90": 0.021862060737609865,
151
+ "p95": 0.021867974376678468,
152
+ "p99": 0.02187270528793335,
153
+ "values": [
154
+ 0.021755615234375,
155
+ 0.02187388801574707
156
+ ]
157
+ },
158
+ "throughput": {
159
+ "unit": "tokens/s",
160
+ "value": 320.88378177812115
161
+ },
162
+ "energy": null,
163
+ "efficiency": null
164
+ },
165
+ "decode": {
166
+ "memory": {
167
+ "unit": "MB",
168
+ "max_ram": 1842.93376,
169
+ "max_global_vram": 6800.539648,
170
+ "max_process_vram": 0.0,
171
+ "max_reserved": 6146.752512,
172
+ "max_allocated": 5031.820288
173
+ },
174
+ "latency": {
175
+ "unit": "s",
176
+ "count": 2,
177
+ "total": 5.278009765625,
178
+ "mean": 2.6390048828125,
179
+ "stdev": 0.009649902343749961,
180
+ "p50": 2.6390048828125,
181
+ "p90": 2.6467248046875,
182
+ "p95": 2.6476897949218747,
183
+ "p99": 2.6484617871093747,
184
+ "values": [
185
+ 2.62935498046875,
186
+ 2.64865478515625
187
+ ]
188
+ },
189
+ "throughput": {
190
+ "unit": "tokens/s",
191
+ "value": 48.1242004617478
192
+ },
193
+ "energy": null,
194
+ "efficiency": null
195
+ },
196
+ "per_token": {
197
+ "memory": null,
198
+ "latency": {
199
+ "unit": "s",
200
+ "count": 254,
201
+ "total": 5.277753339767452,
202
+ "mean": 0.020778556455777385,
203
+ "stdev": 0.0003865362065575658,
204
+ "p50": 0.020905472755432128,
205
+ "p90": 0.021012172508239748,
206
+ "p95": 0.021161625480651855,
207
+ "p99": 0.02166473762512207,
208
+ "values": [
209
+ 0.02003046417236328,
210
+ 0.020065280914306642,
211
+ 0.019990528106689453,
212
+ 0.020926464080810548,
213
+ 0.021012479782104493,
214
+ 0.020948991775512696,
215
+ 0.020947967529296875,
216
+ 0.021011455535888672,
217
+ 0.021008384704589843,
218
+ 0.021201919555664063,
219
+ 0.02150297546386719,
220
+ 0.02108415985107422,
221
+ 0.020997119903564454,
222
+ 0.020933631896972657,
223
+ 0.02101862335205078,
224
+ 0.021005311965942384,
225
+ 0.020943872451782225,
226
+ 0.020888576507568358,
227
+ 0.020926464080810548,
228
+ 0.020997119903564454,
229
+ 0.020960256576538085,
230
+ 0.020934656143188478,
231
+ 0.020958208084106447,
232
+ 0.020967424392700194,
233
+ 0.020971519470214844,
234
+ 0.020068351745605468,
235
+ 0.02005606460571289,
236
+ 0.020105215072631837,
237
+ 0.020083711624145507,
238
+ 0.02008166313171387,
239
+ 0.020124671936035156,
240
+ 0.020814847946166993,
241
+ 0.020990976333618162,
242
+ 0.020924415588378906,
243
+ 0.02089779281616211,
244
+ 0.020908031463623047,
245
+ 0.020933631896972657,
246
+ 0.020952064514160155,
247
+ 0.020933631896972657,
248
+ 0.02147327995300293,
249
+ 0.020937728881835937,
250
+ 0.02092748832702637,
251
+ 0.020888576507568358,
252
+ 0.020839424133300782,
253
+ 0.02085990333557129,
254
+ 0.020916223526000977,
255
+ 0.02089779281616211,
256
+ 0.02088652801513672,
257
+ 0.02087731170654297,
258
+ 0.020890623092651366,
259
+ 0.020940799713134766,
260
+ 0.021316608428955077,
261
+ 0.020915199279785156,
262
+ 0.02086092758178711,
263
+ 0.021187583923339845,
264
+ 0.020996095657348633,
265
+ 0.02089369583129883,
266
+ 0.020118528366088868,
267
+ 0.020182016372680665,
268
+ 0.020411392211914063,
269
+ 0.020377599716186523,
270
+ 0.020122623443603514,
271
+ 0.020122623443603514,
272
+ 0.020130815505981444,
273
+ 0.02017791938781738,
274
+ 0.020148223876953125,
275
+ 0.020115455627441405,
276
+ 0.020162559509277343,
277
+ 0.020187135696411132,
278
+ 0.020165632247924805,
279
+ 0.020100095748901366,
280
+ 0.02008064079284668,
281
+ 0.020060159683227538,
282
+ 0.019842048645019532,
283
+ 0.0198973445892334,
284
+ 0.020090879440307616,
285
+ 0.02008576011657715,
286
+ 0.020048896789550782,
287
+ 0.02008166313171387,
288
+ 0.02005401611328125,
289
+ 0.020195327758789062,
290
+ 0.020926464080810548,
291
+ 0.020904960632324218,
292
+ 0.020923391342163086,
293
+ 0.020896768569946288,
294
+ 0.02087833595275879,
295
+ 0.020941823959350587,
296
+ 0.020930559158325195,
297
+ 0.021746688842773438,
298
+ 0.020146175384521483,
299
+ 0.020108287811279296,
300
+ 0.019981311798095702,
301
+ 0.01986662483215332,
302
+ 0.019935232162475586,
303
+ 0.020083711624145507,
304
+ 0.02087424087524414,
305
+ 0.021046272277832033,
306
+ 0.02090598487854004,
307
+ 0.020960256576538085,
308
+ 0.020904960632324218,
309
+ 0.020958208084106447,
310
+ 0.020993024826049804,
311
+ 0.020934656143188478,
312
+ 0.02087321662902832,
313
+ 0.020953088760375976,
314
+ 0.020960256576538085,
315
+ 0.020942848205566408,
316
+ 0.020908031463623047,
317
+ 0.020989952087402345,
318
+ 0.020736000061035157,
319
+ 0.021797887802124022,
320
+ 0.02107699203491211,
321
+ 0.020914176940917968,
322
+ 0.020940799713134766,
323
+ 0.020938751220703124,
324
+ 0.02087116813659668,
325
+ 0.020931583404541015,
326
+ 0.020941823959350587,
327
+ 0.020914176940917968,
328
+ 0.020908031463623047,
329
+ 0.020939775466918945,
330
+ 0.020892671585083008,
331
+ 0.020888576507568358,
332
+ 0.02089369583129883,
333
+ 0.020799488067626954,
334
+ 0.02090188789367676,
335
+ 0.020891647338867187,
336
+ 0.020063232421875,
337
+ 0.02004582405090332,
338
+ 0.02003455924987793,
339
+ 0.02005401611328125,
340
+ 0.020075519561767577,
341
+ 0.020065280914306642,
342
+ 0.020048896789550782,
343
+ 0.0200447998046875,
344
+ 0.020810752868652343,
345
+ 0.02105036735534668,
346
+ 0.020876287460327148,
347
+ 0.020918272018432618,
348
+ 0.0208721923828125,
349
+ 0.02086502456665039,
350
+ 0.02087731170654297,
351
+ 0.02083430480957031,
352
+ 0.020855808258056642,
353
+ 0.020892671585083008,
354
+ 0.020899839401245117,
355
+ 0.02084249687194824,
356
+ 0.021789695739746092,
357
+ 0.020948991775512696,
358
+ 0.02090188789367676,
359
+ 0.021147647857666017,
360
+ 0.020988927841186524,
361
+ 0.02088960075378418,
362
+ 0.020891647338867187,
363
+ 0.02090188789367676,
364
+ 0.020876287460327148,
365
+ 0.02083020782470703,
366
+ 0.02088960075378418,
367
+ 0.020776960372924806,
368
+ 0.021592063903808592,
369
+ 0.021112831115722656,
370
+ 0.020931583404541015,
371
+ 0.020896768569946288,
372
+ 0.020992000579833983,
373
+ 0.0208855037689209,
374
+ 0.02089779281616211,
375
+ 0.02086911964416504,
376
+ 0.020887615203857422,
377
+ 0.020877248764038087,
378
+ 0.02090291213989258,
379
+ 0.020858879089355468,
380
+ 0.02087833595275879,
381
+ 0.020858879089355468,
382
+ 0.02090598487854004,
383
+ 0.02088755226135254,
384
+ 0.02091110420227051,
385
+ 0.02085683250427246,
386
+ 0.02092748832702637,
387
+ 0.020916223526000977,
388
+ 0.021145599365234375,
389
+ 0.020899839401245117,
390
+ 0.02088755226135254,
391
+ 0.020855808258056642,
392
+ 0.02150092887878418,
393
+ 0.020937728881835937,
394
+ 0.02086502456665039,
395
+ 0.020958208084106447,
396
+ 0.020918272018432618,
397
+ 0.020939775466918945,
398
+ 0.020955135345458984,
399
+ 0.020900863647460938,
400
+ 0.02101759910583496,
401
+ 0.020954111099243163,
402
+ 0.02089369583129883,
403
+ 0.02103500747680664,
404
+ 0.020959232330322267,
405
+ 0.020941823959350587,
406
+ 0.020884479522705078,
407
+ 0.02085990333557129,
408
+ 0.020926464080810548,
409
+ 0.020938751220703124,
410
+ 0.020920320510864256,
411
+ 0.020928512573242186,
412
+ 0.020974592208862306,
413
+ 0.020943872451782225,
414
+ 0.020909055709838868,
415
+ 0.020898815155029296,
416
+ 0.02084864044189453,
417
+ 0.021112831115722656,
418
+ 0.021527551651000978,
419
+ 0.020682752609252928,
420
+ 0.020924415588378906,
421
+ 0.020992000579833983,
422
+ 0.020946943283081054,
423
+ 0.020940799713134766,
424
+ 0.02092748832702637,
425
+ 0.020992000579833983,
426
+ 0.020956159591674805,
427
+ 0.020907007217407226,
428
+ 0.02091929626464844,
429
+ 0.020948991775512696,
430
+ 0.020941823959350587,
431
+ 0.020907007217407226,
432
+ 0.020722688674926756,
433
+ 0.0208855037689209,
434
+ 0.020913152694702147,
435
+ 0.02087833595275879,
436
+ 0.020900863647460938,
437
+ 0.02089574432373047,
438
+ 0.020987903594970703,
439
+ 0.020931583404541015,
440
+ 0.02151219177246094,
441
+ 0.020955135345458984,
442
+ 0.020951040267944337,
443
+ 0.020944896697998046,
444
+ 0.020923391342163086,
445
+ 0.020967424392700194,
446
+ 0.021014528274536134,
447
+ 0.020958208084106447,
448
+ 0.020954111099243163,
449
+ 0.020674560546875,
450
+ 0.020912128448486327,
451
+ 0.021489664077758788,
452
+ 0.020982784271240236,
453
+ 0.020076543807983398,
454
+ 0.020105215072631837,
455
+ 0.020059135437011717,
456
+ 0.02007244873046875,
457
+ 0.020060159683227538,
458
+ 0.020133888244628906,
459
+ 0.020948991775512696,
460
+ 0.020953088760375976,
461
+ 0.020938751220703124,
462
+ 0.02088652801513672
463
+ ]
464
+ },
465
+ "throughput": {
466
+ "unit": "tokens/s",
467
+ "value": 48.12653863266591
468
+ },
469
+ "energy": null,
470
+ "efficiency": null
471
+ }
472
+ }
473
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark.log ADDED
File without changes
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "pytorch_generate",
3
+ "backend": {
4
+ "name": "pytorch",
5
+ "version": "2.4.0+cu121",
6
+ "_target_": "optimum_benchmark.backends.pytorch.backend.PyTorchBackend",
7
+ "task": "text-generation",
8
+ "library": "transformers",
9
+ "model_type": "gemma",
10
+ "model": "google/gemma-2b",
11
+ "processor": "google/gemma-2b",
12
+ "device": "cuda",
13
+ "device_ids": "0",
14
+ "seed": 42,
15
+ "inter_op_num_threads": null,
16
+ "intra_op_num_threads": null,
17
+ "model_kwargs": {},
18
+ "processor_kwargs": {},
19
+ "no_weights": true,
20
+ "device_map": null,
21
+ "torch_dtype": "float16",
22
+ "eval_mode": true,
23
+ "to_bettertransformer": false,
24
+ "low_cpu_mem_usage": null,
25
+ "attn_implementation": null,
26
+ "cache_implementation": "static",
27
+ "autocast_enabled": false,
28
+ "autocast_dtype": null,
29
+ "torch_compile": false,
30
+ "torch_compile_target": "forward",
31
+ "torch_compile_config": {
32
+ "backend": "inductor",
33
+ "mode": "reduce-overhead",
34
+ "fullgraph": true
35
+ },
36
+ "quantization_scheme": null,
37
+ "quantization_config": {},
38
+ "deepspeed_inference": false,
39
+ "deepspeed_inference_config": {},
40
+ "peft_type": null,
41
+ "peft_config": {}
42
+ },
43
+ "scenario": {
44
+ "name": "inference",
45
+ "_target_": "optimum_benchmark.scenarios.inference.scenario.InferenceScenario",
46
+ "iterations": 2,
47
+ "duration": 0,
48
+ "warmup_runs": 10,
49
+ "input_shapes": {
50
+ "batch_size": 1,
51
+ "num_choices": 2,
52
+ "sequence_length": 7
53
+ },
54
+ "new_tokens": null,
55
+ "memory": true,
56
+ "latency": true,
57
+ "energy": false,
58
+ "forward_kwargs": {},
59
+ "generate_kwargs": {
60
+ "max_new_tokens": 128,
61
+ "min_new_tokens": 128,
62
+ "do_sample": false
63
+ },
64
+ "call_kwargs": {}
65
+ },
66
+ "launcher": {
67
+ "name": "process",
68
+ "_target_": "optimum_benchmark.launchers.process.launcher.ProcessLauncher",
69
+ "device_isolation": true,
70
+ "device_isolation_action": "warn",
71
+ "numactl": false,
72
+ "numactl_kwargs": {},
73
+ "start_method": "spawn"
74
+ },
75
+ "environment": {
76
+ "cpu": " AMD EPYC 7R32",
77
+ "cpu_count": 16,
78
+ "cpu_ram_mb": 66697.261056,
79
+ "system": "Linux",
80
+ "machine": "x86_64",
81
+ "platform": "Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29",
82
+ "processor": "x86_64",
83
+ "python_version": "3.8.10",
84
+ "gpu": [
85
+ "NVIDIA A10G"
86
+ ],
87
+ "gpu_count": 1,
88
+ "gpu_vram_mb": 24146608128,
89
+ "optimum_benchmark_version": "0.4.0",
90
+ "optimum_benchmark_commit": null,
91
+ "transformers_version": "4.45.0.dev0",
92
+ "transformers_commit": "65bb28444849976f853063edb958b3ef3dd59d12",
93
+ "accelerate_version": "0.35.0.dev0",
94
+ "accelerate_commit": null,
95
+ "diffusers_version": null,
96
+ "diffusers_commit": null,
97
+ "optimum_version": "1.22.0.dev0",
98
+ "optimum_commit": null,
99
+ "timm_version": "0.9.16",
100
+ "timm_commit": null,
101
+ "peft_version": "0.12.1.dev0",
102
+ "peft_commit": null
103
+ }
104
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/benchmark_report.json ADDED
@@ -0,0 +1,367 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "load": {
3
+ "memory": {
4
+ "unit": "MB",
5
+ "max_ram": 1336.307712,
6
+ "max_global_vram": 6775.373824,
7
+ "max_process_vram": 0.0,
8
+ "max_reserved": 6138.363904,
9
+ "max_allocated": 6060.931072
10
+ },
11
+ "latency": {
12
+ "unit": "s",
13
+ "count": 1,
14
+ "total": 11.531244140625,
15
+ "mean": 11.531244140625,
16
+ "stdev": 0.0,
17
+ "p50": 11.531244140625,
18
+ "p90": 11.531244140625,
19
+ "p95": 11.531244140625,
20
+ "p99": 11.531244140625,
21
+ "values": [
22
+ 11.531244140625
23
+ ]
24
+ },
25
+ "throughput": null,
26
+ "energy": null,
27
+ "efficiency": null
28
+ },
29
+ "prefill": {
30
+ "memory": {
31
+ "unit": "MB",
32
+ "max_ram": 1815.650304,
33
+ "max_global_vram": 6796.345344,
34
+ "max_process_vram": 0.0,
35
+ "max_reserved": 6142.558208,
36
+ "max_allocated": 5028.450816
37
+ },
38
+ "latency": {
39
+ "unit": "s",
40
+ "count": 2,
41
+ "total": 0.043629503250122076,
42
+ "mean": 0.021814751625061038,
43
+ "stdev": 5.913639068603517e-05,
44
+ "p50": 0.021814751625061038,
45
+ "p90": 0.021862060737609865,
46
+ "p95": 0.021867974376678468,
47
+ "p99": 0.02187270528793335,
48
+ "values": [
49
+ 0.021755615234375,
50
+ 0.02187388801574707
51
+ ]
52
+ },
53
+ "throughput": {
54
+ "unit": "tokens/s",
55
+ "value": 320.88378177812115
56
+ },
57
+ "energy": null,
58
+ "efficiency": null
59
+ },
60
+ "decode": {
61
+ "memory": {
62
+ "unit": "MB",
63
+ "max_ram": 1842.93376,
64
+ "max_global_vram": 6800.539648,
65
+ "max_process_vram": 0.0,
66
+ "max_reserved": 6146.752512,
67
+ "max_allocated": 5031.820288
68
+ },
69
+ "latency": {
70
+ "unit": "s",
71
+ "count": 2,
72
+ "total": 5.278009765625,
73
+ "mean": 2.6390048828125,
74
+ "stdev": 0.009649902343749961,
75
+ "p50": 2.6390048828125,
76
+ "p90": 2.6467248046875,
77
+ "p95": 2.6476897949218747,
78
+ "p99": 2.6484617871093747,
79
+ "values": [
80
+ 2.62935498046875,
81
+ 2.64865478515625
82
+ ]
83
+ },
84
+ "throughput": {
85
+ "unit": "tokens/s",
86
+ "value": 48.1242004617478
87
+ },
88
+ "energy": null,
89
+ "efficiency": null
90
+ },
91
+ "per_token": {
92
+ "memory": null,
93
+ "latency": {
94
+ "unit": "s",
95
+ "count": 254,
96
+ "total": 5.277753339767452,
97
+ "mean": 0.020778556455777385,
98
+ "stdev": 0.0003865362065575658,
99
+ "p50": 0.020905472755432128,
100
+ "p90": 0.021012172508239748,
101
+ "p95": 0.021161625480651855,
102
+ "p99": 0.02166473762512207,
103
+ "values": [
104
+ 0.02003046417236328,
105
+ 0.020065280914306642,
106
+ 0.019990528106689453,
107
+ 0.020926464080810548,
108
+ 0.021012479782104493,
109
+ 0.020948991775512696,
110
+ 0.020947967529296875,
111
+ 0.021011455535888672,
112
+ 0.021008384704589843,
113
+ 0.021201919555664063,
114
+ 0.02150297546386719,
115
+ 0.02108415985107422,
116
+ 0.020997119903564454,
117
+ 0.020933631896972657,
118
+ 0.02101862335205078,
119
+ 0.021005311965942384,
120
+ 0.020943872451782225,
121
+ 0.020888576507568358,
122
+ 0.020926464080810548,
123
+ 0.020997119903564454,
124
+ 0.020960256576538085,
125
+ 0.020934656143188478,
126
+ 0.020958208084106447,
127
+ 0.020967424392700194,
128
+ 0.020971519470214844,
129
+ 0.020068351745605468,
130
+ 0.02005606460571289,
131
+ 0.020105215072631837,
132
+ 0.020083711624145507,
133
+ 0.02008166313171387,
134
+ 0.020124671936035156,
135
+ 0.020814847946166993,
136
+ 0.020990976333618162,
137
+ 0.020924415588378906,
138
+ 0.02089779281616211,
139
+ 0.020908031463623047,
140
+ 0.020933631896972657,
141
+ 0.020952064514160155,
142
+ 0.020933631896972657,
143
+ 0.02147327995300293,
144
+ 0.020937728881835937,
145
+ 0.02092748832702637,
146
+ 0.020888576507568358,
147
+ 0.020839424133300782,
148
+ 0.02085990333557129,
149
+ 0.020916223526000977,
150
+ 0.02089779281616211,
151
+ 0.02088652801513672,
152
+ 0.02087731170654297,
153
+ 0.020890623092651366,
154
+ 0.020940799713134766,
155
+ 0.021316608428955077,
156
+ 0.020915199279785156,
157
+ 0.02086092758178711,
158
+ 0.021187583923339845,
159
+ 0.020996095657348633,
160
+ 0.02089369583129883,
161
+ 0.020118528366088868,
162
+ 0.020182016372680665,
163
+ 0.020411392211914063,
164
+ 0.020377599716186523,
165
+ 0.020122623443603514,
166
+ 0.020122623443603514,
167
+ 0.020130815505981444,
168
+ 0.02017791938781738,
169
+ 0.020148223876953125,
170
+ 0.020115455627441405,
171
+ 0.020162559509277343,
172
+ 0.020187135696411132,
173
+ 0.020165632247924805,
174
+ 0.020100095748901366,
175
+ 0.02008064079284668,
176
+ 0.020060159683227538,
177
+ 0.019842048645019532,
178
+ 0.0198973445892334,
179
+ 0.020090879440307616,
180
+ 0.02008576011657715,
181
+ 0.020048896789550782,
182
+ 0.02008166313171387,
183
+ 0.02005401611328125,
184
+ 0.020195327758789062,
185
+ 0.020926464080810548,
186
+ 0.020904960632324218,
187
+ 0.020923391342163086,
188
+ 0.020896768569946288,
189
+ 0.02087833595275879,
190
+ 0.020941823959350587,
191
+ 0.020930559158325195,
192
+ 0.021746688842773438,
193
+ 0.020146175384521483,
194
+ 0.020108287811279296,
195
+ 0.019981311798095702,
196
+ 0.01986662483215332,
197
+ 0.019935232162475586,
198
+ 0.020083711624145507,
199
+ 0.02087424087524414,
200
+ 0.021046272277832033,
201
+ 0.02090598487854004,
202
+ 0.020960256576538085,
203
+ 0.020904960632324218,
204
+ 0.020958208084106447,
205
+ 0.020993024826049804,
206
+ 0.020934656143188478,
207
+ 0.02087321662902832,
208
+ 0.020953088760375976,
209
+ 0.020960256576538085,
210
+ 0.020942848205566408,
211
+ 0.020908031463623047,
212
+ 0.020989952087402345,
213
+ 0.020736000061035157,
214
+ 0.021797887802124022,
215
+ 0.02107699203491211,
216
+ 0.020914176940917968,
217
+ 0.020940799713134766,
218
+ 0.020938751220703124,
219
+ 0.02087116813659668,
220
+ 0.020931583404541015,
221
+ 0.020941823959350587,
222
+ 0.020914176940917968,
223
+ 0.020908031463623047,
224
+ 0.020939775466918945,
225
+ 0.020892671585083008,
226
+ 0.020888576507568358,
227
+ 0.02089369583129883,
228
+ 0.020799488067626954,
229
+ 0.02090188789367676,
230
+ 0.020891647338867187,
231
+ 0.020063232421875,
232
+ 0.02004582405090332,
233
+ 0.02003455924987793,
234
+ 0.02005401611328125,
235
+ 0.020075519561767577,
236
+ 0.020065280914306642,
237
+ 0.020048896789550782,
238
+ 0.0200447998046875,
239
+ 0.020810752868652343,
240
+ 0.02105036735534668,
241
+ 0.020876287460327148,
242
+ 0.020918272018432618,
243
+ 0.0208721923828125,
244
+ 0.02086502456665039,
245
+ 0.02087731170654297,
246
+ 0.02083430480957031,
247
+ 0.020855808258056642,
248
+ 0.020892671585083008,
249
+ 0.020899839401245117,
250
+ 0.02084249687194824,
251
+ 0.021789695739746092,
252
+ 0.020948991775512696,
253
+ 0.02090188789367676,
254
+ 0.021147647857666017,
255
+ 0.020988927841186524,
256
+ 0.02088960075378418,
257
+ 0.020891647338867187,
258
+ 0.02090188789367676,
259
+ 0.020876287460327148,
260
+ 0.02083020782470703,
261
+ 0.02088960075378418,
262
+ 0.020776960372924806,
263
+ 0.021592063903808592,
264
+ 0.021112831115722656,
265
+ 0.020931583404541015,
266
+ 0.020896768569946288,
267
+ 0.020992000579833983,
268
+ 0.0208855037689209,
269
+ 0.02089779281616211,
270
+ 0.02086911964416504,
271
+ 0.020887615203857422,
272
+ 0.020877248764038087,
273
+ 0.02090291213989258,
274
+ 0.020858879089355468,
275
+ 0.02087833595275879,
276
+ 0.020858879089355468,
277
+ 0.02090598487854004,
278
+ 0.02088755226135254,
279
+ 0.02091110420227051,
280
+ 0.02085683250427246,
281
+ 0.02092748832702637,
282
+ 0.020916223526000977,
283
+ 0.021145599365234375,
284
+ 0.020899839401245117,
285
+ 0.02088755226135254,
286
+ 0.020855808258056642,
287
+ 0.02150092887878418,
288
+ 0.020937728881835937,
289
+ 0.02086502456665039,
290
+ 0.020958208084106447,
291
+ 0.020918272018432618,
292
+ 0.020939775466918945,
293
+ 0.020955135345458984,
294
+ 0.020900863647460938,
295
+ 0.02101759910583496,
296
+ 0.020954111099243163,
297
+ 0.02089369583129883,
298
+ 0.02103500747680664,
299
+ 0.020959232330322267,
300
+ 0.020941823959350587,
301
+ 0.020884479522705078,
302
+ 0.02085990333557129,
303
+ 0.020926464080810548,
304
+ 0.020938751220703124,
305
+ 0.020920320510864256,
306
+ 0.020928512573242186,
307
+ 0.020974592208862306,
308
+ 0.020943872451782225,
309
+ 0.020909055709838868,
310
+ 0.020898815155029296,
311
+ 0.02084864044189453,
312
+ 0.021112831115722656,
313
+ 0.021527551651000978,
314
+ 0.020682752609252928,
315
+ 0.020924415588378906,
316
+ 0.020992000579833983,
317
+ 0.020946943283081054,
318
+ 0.020940799713134766,
319
+ 0.02092748832702637,
320
+ 0.020992000579833983,
321
+ 0.020956159591674805,
322
+ 0.020907007217407226,
323
+ 0.02091929626464844,
324
+ 0.020948991775512696,
325
+ 0.020941823959350587,
326
+ 0.020907007217407226,
327
+ 0.020722688674926756,
328
+ 0.0208855037689209,
329
+ 0.020913152694702147,
330
+ 0.02087833595275879,
331
+ 0.020900863647460938,
332
+ 0.02089574432373047,
333
+ 0.020987903594970703,
334
+ 0.020931583404541015,
335
+ 0.02151219177246094,
336
+ 0.020955135345458984,
337
+ 0.020951040267944337,
338
+ 0.020944896697998046,
339
+ 0.020923391342163086,
340
+ 0.020967424392700194,
341
+ 0.021014528274536134,
342
+ 0.020958208084106447,
343
+ 0.020954111099243163,
344
+ 0.020674560546875,
345
+ 0.020912128448486327,
346
+ 0.021489664077758788,
347
+ 0.020982784271240236,
348
+ 0.020076543807983398,
349
+ 0.020105215072631837,
350
+ 0.020059135437011717,
351
+ 0.02007244873046875,
352
+ 0.020060159683227538,
353
+ 0.020133888244628906,
354
+ 0.020948991775512696,
355
+ 0.020953088760375976,
356
+ 0.020938751220703124,
357
+ 0.02088652801513672
358
+ ]
359
+ },
360
+ "throughput": {
361
+ "unit": "tokens/s",
362
+ "value": 48.12653863266591
363
+ },
364
+ "energy": null,
365
+ "efficiency": null
366
+ }
367
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=False/summary.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "google/gemma-2b",
3
+ "commit": "65bb28444849976f853063edb958b3ef3dd59d12",
4
+ "config": "backend.cache_implementation=static,backend.torch_compile=False",
5
+ "metrics": {
6
+ "prefill.latency.mean": 0.021814751625061038,
7
+ "prefill.throughput.value": 320.88378177812115,
8
+ "decode.latency.mean": 2.6390048828125,
9
+ "decode.throughput.value": 48.1242004617478,
10
+ "per_token.latency.mean": 0.020778556455777385,
11
+ "per_token.throughput.value": 48.12653863266591
12
+ }
13
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/config.yaml ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: pytorch_generate
2
+ backend:
3
+ name: pytorch
4
+ version: 2.4.0+cu121
5
+ _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
6
+ task: null
7
+ library: null
8
+ model_type: null
9
+ model: google/gemma-2b
10
+ processor: null
11
+ device: cuda
12
+ device_ids: '0'
13
+ seed: 42
14
+ inter_op_num_threads: null
15
+ intra_op_num_threads: null
16
+ model_kwargs: {}
17
+ processor_kwargs: {}
18
+ no_weights: true
19
+ device_map: null
20
+ torch_dtype: float16
21
+ eval_mode: true
22
+ to_bettertransformer: false
23
+ low_cpu_mem_usage: null
24
+ attn_implementation: null
25
+ cache_implementation: static
26
+ autocast_enabled: false
27
+ autocast_dtype: null
28
+ torch_compile: true
29
+ torch_compile_target: forward
30
+ torch_compile_config:
31
+ backend: inductor
32
+ mode: reduce-overhead
33
+ fullgraph: true
34
+ quantization_scheme: null
35
+ quantization_config: {}
36
+ deepspeed_inference: false
37
+ deepspeed_inference_config: {}
38
+ peft_type: null
39
+ peft_config: {}
40
+ scenario:
41
+ name: inference
42
+ _target_: optimum_benchmark.scenarios.inference.scenario.InferenceScenario
43
+ iterations: 2
44
+ duration: 0
45
+ warmup_runs: 10
46
+ input_shapes:
47
+ batch_size: 1
48
+ sequence_length: 7
49
+ new_tokens: null
50
+ memory: true
51
+ latency: true
52
+ energy: false
53
+ forward_kwargs: {}
54
+ generate_kwargs:
55
+ max_new_tokens: 128
56
+ min_new_tokens: 128
57
+ do_sample: false
58
+ call_kwargs: {}
59
+ launcher:
60
+ name: process
61
+ _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
62
+ device_isolation: true
63
+ device_isolation_action: warn
64
+ numactl: false
65
+ numactl_kwargs: {}
66
+ start_method: spawn
67
+ environment:
68
+ cpu: ' AMD EPYC 7R32'
69
+ cpu_count: 16
70
+ cpu_ram_mb: 66697.261056
71
+ system: Linux
72
+ machine: x86_64
73
+ platform: Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29
74
+ processor: x86_64
75
+ python_version: 3.8.10
76
+ gpu:
77
+ - NVIDIA A10G
78
+ gpu_count: 1
79
+ gpu_vram_mb: 24146608128
80
+ optimum_benchmark_version: 0.4.0
81
+ optimum_benchmark_commit: null
82
+ transformers_version: 4.45.0.dev0
83
+ transformers_commit: 65bb28444849976f853063edb958b3ef3dd59d12
84
+ accelerate_version: 0.35.0.dev0
85
+ accelerate_commit: null
86
+ diffusers_version: null
87
+ diffusers_commit: null
88
+ optimum_version: 1.22.0.dev0
89
+ optimum_commit: null
90
+ timm_version: 0.9.16
91
+ timm_commit: null
92
+ peft_version: 0.12.1.dev0
93
+ peft_commit: null
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/hydra.yaml ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
4
+ sweep:
5
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12
6
+ subdir: ${hydra.job.override_dirname}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ root:
71
+ level: ERROR
72
+ disable_existing_loggers: true
73
+ job_logging:
74
+ version: 1
75
+ root:
76
+ level: ERROR
77
+ disable_existing_loggers: true
78
+ env: {}
79
+ mode: MULTIRUN
80
+ searchpath: []
81
+ callbacks: {}
82
+ output_subdir: .hydra
83
+ overrides:
84
+ hydra:
85
+ - hydra/job_logging=disabled
86
+ - hydra/hydra_logging=disabled
87
+ - hydra.sweep.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12
88
+ - hydra.run.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
89
+ - hydra.mode=MULTIRUN
90
+ task:
91
+ - backend.model=google/gemma-2b
92
+ - backend.cache_implementation=static
93
+ - backend.torch_compile=True
94
+ job:
95
+ name: cli
96
+ chdir: true
97
+ override_dirname: backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True
98
+ id: '3'
99
+ num: 3
100
+ config_name: generation
101
+ env_set:
102
+ OVERRIDE_BENCHMARKS: '1'
103
+ LOG_LEVEL: WARN
104
+ env_copy: []
105
+ config:
106
+ override_dirname:
107
+ kv_sep: '='
108
+ item_sep: ','
109
+ exclude_keys: []
110
+ runtime:
111
+ version: 1.3.2
112
+ version_base: '1.3'
113
+ cwd: /transformers
114
+ config_sources:
115
+ - path: hydra.conf
116
+ schema: pkg
117
+ provider: hydra
118
+ - path: optimum_benchmark
119
+ schema: pkg
120
+ provider: main
121
+ - path: /transformers/benchmark/config
122
+ schema: file
123
+ provider: command-line
124
+ - path: ''
125
+ schema: structured
126
+ provider: schema
127
+ output_dir: /transformers/_benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True
128
+ choices:
129
+ backend: pytorch
130
+ launcher: process
131
+ scenario: inference
132
+ hydra/env: default
133
+ hydra/callbacks: null
134
+ hydra/job_logging: disabled
135
+ hydra/hydra_logging: disabled
136
+ hydra/hydra_help: default
137
+ hydra/help: default
138
+ hydra/sweeper: basic
139
+ hydra/launcher: basic
140
+ hydra/output: default
141
+ verbose: false
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/.hydra/overrides.yaml ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ - backend.model=google/gemma-2b
2
+ - backend.cache_implementation=static
3
+ - backend.torch_compile=True
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark.json ADDED
@@ -0,0 +1,473 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "name": "pytorch_generate",
4
+ "backend": {
5
+ "name": "pytorch",
6
+ "version": "2.4.0+cu121",
7
+ "_target_": "optimum_benchmark.backends.pytorch.backend.PyTorchBackend",
8
+ "task": "text-generation",
9
+ "library": "transformers",
10
+ "model_type": "gemma",
11
+ "model": "google/gemma-2b",
12
+ "processor": "google/gemma-2b",
13
+ "device": "cuda",
14
+ "device_ids": "0",
15
+ "seed": 42,
16
+ "inter_op_num_threads": null,
17
+ "intra_op_num_threads": null,
18
+ "model_kwargs": {},
19
+ "processor_kwargs": {},
20
+ "no_weights": true,
21
+ "device_map": null,
22
+ "torch_dtype": "float16",
23
+ "eval_mode": true,
24
+ "to_bettertransformer": false,
25
+ "low_cpu_mem_usage": null,
26
+ "attn_implementation": null,
27
+ "cache_implementation": "static",
28
+ "autocast_enabled": false,
29
+ "autocast_dtype": null,
30
+ "torch_compile": true,
31
+ "torch_compile_target": "forward",
32
+ "torch_compile_config": {
33
+ "backend": "inductor",
34
+ "mode": "reduce-overhead",
35
+ "fullgraph": true
36
+ },
37
+ "quantization_scheme": null,
38
+ "quantization_config": {},
39
+ "deepspeed_inference": false,
40
+ "deepspeed_inference_config": {},
41
+ "peft_type": null,
42
+ "peft_config": {}
43
+ },
44
+ "scenario": {
45
+ "name": "inference",
46
+ "_target_": "optimum_benchmark.scenarios.inference.scenario.InferenceScenario",
47
+ "iterations": 2,
48
+ "duration": 0,
49
+ "warmup_runs": 10,
50
+ "input_shapes": {
51
+ "batch_size": 1,
52
+ "num_choices": 2,
53
+ "sequence_length": 7
54
+ },
55
+ "new_tokens": null,
56
+ "memory": true,
57
+ "latency": true,
58
+ "energy": false,
59
+ "forward_kwargs": {},
60
+ "generate_kwargs": {
61
+ "max_new_tokens": 128,
62
+ "min_new_tokens": 128,
63
+ "do_sample": false
64
+ },
65
+ "call_kwargs": {}
66
+ },
67
+ "launcher": {
68
+ "name": "process",
69
+ "_target_": "optimum_benchmark.launchers.process.launcher.ProcessLauncher",
70
+ "device_isolation": true,
71
+ "device_isolation_action": "warn",
72
+ "numactl": false,
73
+ "numactl_kwargs": {},
74
+ "start_method": "spawn"
75
+ },
76
+ "environment": {
77
+ "cpu": " AMD EPYC 7R32",
78
+ "cpu_count": 16,
79
+ "cpu_ram_mb": 66697.261056,
80
+ "system": "Linux",
81
+ "machine": "x86_64",
82
+ "platform": "Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29",
83
+ "processor": "x86_64",
84
+ "python_version": "3.8.10",
85
+ "gpu": [
86
+ "NVIDIA A10G"
87
+ ],
88
+ "gpu_count": 1,
89
+ "gpu_vram_mb": 24146608128,
90
+ "optimum_benchmark_version": "0.4.0",
91
+ "optimum_benchmark_commit": null,
92
+ "transformers_version": "4.45.0.dev0",
93
+ "transformers_commit": "65bb28444849976f853063edb958b3ef3dd59d12",
94
+ "accelerate_version": "0.35.0.dev0",
95
+ "accelerate_commit": null,
96
+ "diffusers_version": null,
97
+ "diffusers_commit": null,
98
+ "optimum_version": "1.22.0.dev0",
99
+ "optimum_commit": null,
100
+ "timm_version": "0.9.16",
101
+ "timm_commit": null,
102
+ "peft_version": "0.12.1.dev0",
103
+ "peft_commit": null
104
+ }
105
+ },
106
+ "report": {
107
+ "load": {
108
+ "memory": {
109
+ "unit": "MB",
110
+ "max_ram": 1352.421376,
111
+ "max_global_vram": 6775.373824,
112
+ "max_process_vram": 0.0,
113
+ "max_reserved": 6138.363904,
114
+ "max_allocated": 6060.931072
115
+ },
116
+ "latency": {
117
+ "unit": "s",
118
+ "count": 1,
119
+ "total": 12.48078125,
120
+ "mean": 12.48078125,
121
+ "stdev": 0.0,
122
+ "p50": 12.48078125,
123
+ "p90": 12.48078125,
124
+ "p95": 12.48078125,
125
+ "p99": 12.48078125,
126
+ "values": [
127
+ 12.48078125
128
+ ]
129
+ },
130
+ "throughput": null,
131
+ "energy": null,
132
+ "efficiency": null
133
+ },
134
+ "prefill": {
135
+ "memory": {
136
+ "unit": "MB",
137
+ "max_ram": 1911.443456,
138
+ "max_global_vram": 6796.345344,
139
+ "max_process_vram": 0.0,
140
+ "max_reserved": 6138.363904,
141
+ "max_allocated": 5278.596608
142
+ },
143
+ "latency": {
144
+ "unit": "s",
145
+ "count": 2,
146
+ "total": 0.028401632308959958,
147
+ "mean": 0.014200816154479979,
148
+ "stdev": 0.00015841579437255848,
149
+ "p50": 0.014200816154479979,
150
+ "p90": 0.014327548789978026,
151
+ "p95": 0.014343390369415282,
152
+ "p99": 0.014356063632965087,
153
+ "values": [
154
+ 0.014042400360107421,
155
+ 0.014359231948852538
156
+ ]
157
+ },
158
+ "throughput": {
159
+ "unit": "tokens/s",
160
+ "value": 492.9294150316626
161
+ },
162
+ "energy": null,
163
+ "efficiency": null
164
+ },
165
+ "decode": {
166
+ "memory": {
167
+ "unit": "MB",
168
+ "max_ram": 2084.560896,
169
+ "max_global_vram": 6060.244992,
170
+ "max_process_vram": 0.0,
171
+ "max_reserved": 5381.292032,
172
+ "max_allocated": 5282.647552
173
+ },
174
+ "latency": {
175
+ "unit": "s",
176
+ "count": 2,
177
+ "total": 3.1285546875000003,
178
+ "mean": 1.5642773437500002,
179
+ "stdev": 0.00018029785156248312,
180
+ "p50": 1.5642773437500002,
181
+ "p90": 1.5644215820312501,
182
+ "p95": 1.5644396118164063,
183
+ "p99": 1.5644540356445313,
184
+ "values": [
185
+ 1.5644576416015625,
186
+ 1.5640970458984376
187
+ ]
188
+ },
189
+ "throughput": {
190
+ "unit": "tokens/s",
191
+ "value": 81.18764904920651
192
+ },
193
+ "energy": null,
194
+ "efficiency": null
195
+ },
196
+ "per_token": {
197
+ "memory": null,
198
+ "latency": {
199
+ "unit": "s",
200
+ "count": 254,
201
+ "total": 3.128261626243589,
202
+ "mean": 0.012315990654502328,
203
+ "stdev": 0.00010128158543273827,
204
+ "p50": 0.012290047645568849,
205
+ "p90": 0.012329983711242675,
206
+ "p95": 0.01245911030769348,
207
+ "p99": 0.012875171537399293,
208
+ "values": [
209
+ 0.012876799583435058,
210
+ 0.012303359985351562,
211
+ 0.012290047645568849,
212
+ 0.012297216415405274,
213
+ 0.012314623832702636,
214
+ 0.012327936172485352,
215
+ 0.012281855583190919,
216
+ 0.012271615982055664,
217
+ 0.012286975860595703,
218
+ 0.012294143676757812,
219
+ 0.012288000106811523,
220
+ 0.012300288200378418,
221
+ 0.012291071891784668,
222
+ 0.012272640228271485,
223
+ 0.012279808044433594,
224
+ 0.012272640228271485,
225
+ 0.012286975860595703,
226
+ 0.012275712013244629,
227
+ 0.012308480262756348,
228
+ 0.012311552047729492,
229
+ 0.012289024353027344,
230
+ 0.012275712013244629,
231
+ 0.012292096138000488,
232
+ 0.012275712013244629,
233
+ 0.012285951614379884,
234
+ 0.012290047645568849,
235
+ 0.012271615982055664,
236
+ 0.012296192169189453,
237
+ 0.012283904075622559,
238
+ 0.012279808044433594,
239
+ 0.012283904075622559,
240
+ 0.012291071891784668,
241
+ 0.012291071891784668,
242
+ 0.012292096138000488,
243
+ 0.012280832290649414,
244
+ 0.012267519950866699,
245
+ 0.012294143676757812,
246
+ 0.012288000106811523,
247
+ 0.012292096138000488,
248
+ 0.012274687767028808,
249
+ 0.012273664474487305,
250
+ 0.01246720027923584,
251
+ 0.012308480262756348,
252
+ 0.012297216415405274,
253
+ 0.012311552047729492,
254
+ 0.012333056449890138,
255
+ 0.012309503555297852,
256
+ 0.012279808044433594,
257
+ 0.012294143676757812,
258
+ 0.012294143676757812,
259
+ 0.012285984039306641,
260
+ 0.012297183990478515,
261
+ 0.012300288200378418,
262
+ 0.012290047645568849,
263
+ 0.012285951614379884,
264
+ 0.012285951614379884,
265
+ 0.012275712013244629,
266
+ 0.012266495704650878,
267
+ 0.012280832290649414,
268
+ 0.012329983711242675,
269
+ 0.012306431770324706,
270
+ 0.01232588768005371,
271
+ 0.012297216415405274,
272
+ 0.01228492832183838,
273
+ 0.012277759552001954,
274
+ 0.012271615982055664,
275
+ 0.012281855583190919,
276
+ 0.012274687767028808,
277
+ 0.012286016464233399,
278
+ 0.01228486442565918,
279
+ 0.012285951614379884,
280
+ 0.012310527801513671,
281
+ 0.012294143676757812,
282
+ 0.012290047645568849,
283
+ 0.012278783798217773,
284
+ 0.01242521572113037,
285
+ 0.01232588768005371,
286
+ 0.012561408042907715,
287
+ 0.012838912010192872,
288
+ 0.012317695617675782,
289
+ 0.012289024353027344,
290
+ 0.012288000106811523,
291
+ 0.012308480262756348,
292
+ 0.012329983711242675,
293
+ 0.012899328231811523,
294
+ 0.012322815895080566,
295
+ 0.012290047645568849,
296
+ 0.01228492832183838,
297
+ 0.012298239707946777,
298
+ 0.012291071891784668,
299
+ 0.012275712013244629,
300
+ 0.012275712013244629,
301
+ 0.012326911926269531,
302
+ 0.012316672325134278,
303
+ 0.012310527801513671,
304
+ 0.012301312446594239,
305
+ 0.012281855583190919,
306
+ 0.012270591735839843,
307
+ 0.012291071891784668,
308
+ 0.012277759552001954,
309
+ 0.012356608390808106,
310
+ 0.012512255668640136,
311
+ 0.012548095703125,
312
+ 0.012496895790100097,
313
+ 0.012336128234863282,
314
+ 0.012388352394104005,
315
+ 0.012313599586486817,
316
+ 0.012270591735839843,
317
+ 0.012279808044433594,
318
+ 0.012289024353027344,
319
+ 0.012279808044433594,
320
+ 0.012296192169189453,
321
+ 0.012314623832702636,
322
+ 0.012288000106811523,
323
+ 0.012288000106811523,
324
+ 0.012288000106811523,
325
+ 0.01226854419708252,
326
+ 0.012317695617675782,
327
+ 0.012299263954162597,
328
+ 0.012344320297241211,
329
+ 0.012280832290649414,
330
+ 0.012296192169189453,
331
+ 0.012285951614379884,
332
+ 0.012274687767028808,
333
+ 0.012289024353027344,
334
+ 0.012286975860595703,
335
+ 0.012281855583190919,
336
+ 0.012308480262756348,
337
+ 0.012296192169189453,
338
+ 0.012456959724426269,
339
+ 0.012314623832702636,
340
+ 0.012289024353027344,
341
+ 0.012286975860595703,
342
+ 0.012312576293945313,
343
+ 0.012314623832702636,
344
+ 0.012361727714538574,
345
+ 0.012463104248046876,
346
+ 0.012866559982299805,
347
+ 0.012321791648864745,
348
+ 0.012299263954162597,
349
+ 0.012305407524108887,
350
+ 0.012301312446594239,
351
+ 0.012299263954162597,
352
+ 0.012280832290649414,
353
+ 0.012281855583190919,
354
+ 0.012277759552001954,
355
+ 0.01228492832183838,
356
+ 0.012309503555297852,
357
+ 0.012301312446594239,
358
+ 0.012298239707946777,
359
+ 0.012299263954162597,
360
+ 0.012288000106811523,
361
+ 0.012282879829406738,
362
+ 0.012282879829406738,
363
+ 0.012274687767028808,
364
+ 0.012282879829406738,
365
+ 0.012278783798217773,
366
+ 0.012301312446594239,
367
+ 0.012292096138000488,
368
+ 0.012294143676757812,
369
+ 0.012290047645568849,
370
+ 0.012301312446594239,
371
+ 0.012289024353027344,
372
+ 0.012322815895080566,
373
+ 0.01228492832183838,
374
+ 0.012307456016540527,
375
+ 0.012298239707946777,
376
+ 0.012873727798461914,
377
+ 0.012294143676757812,
378
+ 0.012280832290649414,
379
+ 0.012285951614379884,
380
+ 0.012290047645568849,
381
+ 0.012299263954162597,
382
+ 0.012308480262756348,
383
+ 0.012296192169189453,
384
+ 0.012288000106811523,
385
+ 0.012281855583190919,
386
+ 0.012271615982055664,
387
+ 0.012275712013244629,
388
+ 0.01228492832183838,
389
+ 0.012272640228271485,
390
+ 0.012292096138000488,
391
+ 0.012294143676757812,
392
+ 0.012288000106811523,
393
+ 0.012297216415405274,
394
+ 0.012786687850952149,
395
+ 0.012385279655456542,
396
+ 0.012288000106811523,
397
+ 0.012285951614379884,
398
+ 0.012286975860595703,
399
+ 0.012356608390808106,
400
+ 0.012301312446594239,
401
+ 0.01228492832183838,
402
+ 0.012280832290649414,
403
+ 0.012271615982055664,
404
+ 0.012295167922973632,
405
+ 0.012300288200378418,
406
+ 0.012288000106811523,
407
+ 0.012311552047729492,
408
+ 0.012303359985351562,
409
+ 0.012286975860595703,
410
+ 0.012282879829406738,
411
+ 0.012423168182373047,
412
+ 0.012306431770324706,
413
+ 0.012306431770324706,
414
+ 0.012283904075622559,
415
+ 0.012285951614379884,
416
+ 0.012288000106811523,
417
+ 0.012286975860595703,
418
+ 0.012288000106811523,
419
+ 0.012301312446594239,
420
+ 0.012318719863891601,
421
+ 0.012298239707946777,
422
+ 0.012296192169189453,
423
+ 0.012288000106811523,
424
+ 0.012303359985351562,
425
+ 0.012295167922973632,
426
+ 0.01227673625946045,
427
+ 0.012283904075622559,
428
+ 0.012272640228271485,
429
+ 0.01228492832183838,
430
+ 0.012281855583190919,
431
+ 0.01228492832183838,
432
+ 0.01228492832183838,
433
+ 0.012297216415405274,
434
+ 0.012312576293945313,
435
+ 0.012290047645568849,
436
+ 0.012274687767028808,
437
+ 0.012273664474487305,
438
+ 0.012278783798217773,
439
+ 0.012282879829406738,
440
+ 0.012286975860595703,
441
+ 0.012290047645568849,
442
+ 0.012291071891784668,
443
+ 0.012270591735839843,
444
+ 0.012283904075622559,
445
+ 0.012281855583190919,
446
+ 0.012278783798217773,
447
+ 0.012282879829406738,
448
+ 0.012319744110107422,
449
+ 0.012295167922973632,
450
+ 0.01226854419708252,
451
+ 0.012272640228271485,
452
+ 0.01228492832183838,
453
+ 0.01227673625946045,
454
+ 0.012294143676757812,
455
+ 0.012281855583190919,
456
+ 0.012271615982055664,
457
+ 0.012281855583190919,
458
+ 0.012300288200378418,
459
+ 0.012281855583190919,
460
+ 0.012920831680297852,
461
+ 0.012331007957458496,
462
+ 0.012274687767028808
463
+ ]
464
+ },
465
+ "throughput": {
466
+ "unit": "tokens/s",
467
+ "value": 81.19525485628981
468
+ },
469
+ "energy": null,
470
+ "efficiency": null
471
+ }
472
+ }
473
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark.log ADDED
File without changes
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark_config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "pytorch_generate",
3
+ "backend": {
4
+ "name": "pytorch",
5
+ "version": "2.4.0+cu121",
6
+ "_target_": "optimum_benchmark.backends.pytorch.backend.PyTorchBackend",
7
+ "task": "text-generation",
8
+ "library": "transformers",
9
+ "model_type": "gemma",
10
+ "model": "google/gemma-2b",
11
+ "processor": "google/gemma-2b",
12
+ "device": "cuda",
13
+ "device_ids": "0",
14
+ "seed": 42,
15
+ "inter_op_num_threads": null,
16
+ "intra_op_num_threads": null,
17
+ "model_kwargs": {},
18
+ "processor_kwargs": {},
19
+ "no_weights": true,
20
+ "device_map": null,
21
+ "torch_dtype": "float16",
22
+ "eval_mode": true,
23
+ "to_bettertransformer": false,
24
+ "low_cpu_mem_usage": null,
25
+ "attn_implementation": null,
26
+ "cache_implementation": "static",
27
+ "autocast_enabled": false,
28
+ "autocast_dtype": null,
29
+ "torch_compile": true,
30
+ "torch_compile_target": "forward",
31
+ "torch_compile_config": {
32
+ "backend": "inductor",
33
+ "mode": "reduce-overhead",
34
+ "fullgraph": true
35
+ },
36
+ "quantization_scheme": null,
37
+ "quantization_config": {},
38
+ "deepspeed_inference": false,
39
+ "deepspeed_inference_config": {},
40
+ "peft_type": null,
41
+ "peft_config": {}
42
+ },
43
+ "scenario": {
44
+ "name": "inference",
45
+ "_target_": "optimum_benchmark.scenarios.inference.scenario.InferenceScenario",
46
+ "iterations": 2,
47
+ "duration": 0,
48
+ "warmup_runs": 10,
49
+ "input_shapes": {
50
+ "batch_size": 1,
51
+ "num_choices": 2,
52
+ "sequence_length": 7
53
+ },
54
+ "new_tokens": null,
55
+ "memory": true,
56
+ "latency": true,
57
+ "energy": false,
58
+ "forward_kwargs": {},
59
+ "generate_kwargs": {
60
+ "max_new_tokens": 128,
61
+ "min_new_tokens": 128,
62
+ "do_sample": false
63
+ },
64
+ "call_kwargs": {}
65
+ },
66
+ "launcher": {
67
+ "name": "process",
68
+ "_target_": "optimum_benchmark.launchers.process.launcher.ProcessLauncher",
69
+ "device_isolation": true,
70
+ "device_isolation_action": "warn",
71
+ "numactl": false,
72
+ "numactl_kwargs": {},
73
+ "start_method": "spawn"
74
+ },
75
+ "environment": {
76
+ "cpu": " AMD EPYC 7R32",
77
+ "cpu_count": 16,
78
+ "cpu_ram_mb": 66697.261056,
79
+ "system": "Linux",
80
+ "machine": "x86_64",
81
+ "platform": "Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29",
82
+ "processor": "x86_64",
83
+ "python_version": "3.8.10",
84
+ "gpu": [
85
+ "NVIDIA A10G"
86
+ ],
87
+ "gpu_count": 1,
88
+ "gpu_vram_mb": 24146608128,
89
+ "optimum_benchmark_version": "0.4.0",
90
+ "optimum_benchmark_commit": null,
91
+ "transformers_version": "4.45.0.dev0",
92
+ "transformers_commit": "65bb28444849976f853063edb958b3ef3dd59d12",
93
+ "accelerate_version": "0.35.0.dev0",
94
+ "accelerate_commit": null,
95
+ "diffusers_version": null,
96
+ "diffusers_commit": null,
97
+ "optimum_version": "1.22.0.dev0",
98
+ "optimum_commit": null,
99
+ "timm_version": "0.9.16",
100
+ "timm_commit": null,
101
+ "peft_version": "0.12.1.dev0",
102
+ "peft_commit": null
103
+ }
104
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/benchmark_report.json ADDED
@@ -0,0 +1,367 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "load": {
3
+ "memory": {
4
+ "unit": "MB",
5
+ "max_ram": 1352.421376,
6
+ "max_global_vram": 6775.373824,
7
+ "max_process_vram": 0.0,
8
+ "max_reserved": 6138.363904,
9
+ "max_allocated": 6060.931072
10
+ },
11
+ "latency": {
12
+ "unit": "s",
13
+ "count": 1,
14
+ "total": 12.48078125,
15
+ "mean": 12.48078125,
16
+ "stdev": 0.0,
17
+ "p50": 12.48078125,
18
+ "p90": 12.48078125,
19
+ "p95": 12.48078125,
20
+ "p99": 12.48078125,
21
+ "values": [
22
+ 12.48078125
23
+ ]
24
+ },
25
+ "throughput": null,
26
+ "energy": null,
27
+ "efficiency": null
28
+ },
29
+ "prefill": {
30
+ "memory": {
31
+ "unit": "MB",
32
+ "max_ram": 1911.443456,
33
+ "max_global_vram": 6796.345344,
34
+ "max_process_vram": 0.0,
35
+ "max_reserved": 6138.363904,
36
+ "max_allocated": 5278.596608
37
+ },
38
+ "latency": {
39
+ "unit": "s",
40
+ "count": 2,
41
+ "total": 0.028401632308959958,
42
+ "mean": 0.014200816154479979,
43
+ "stdev": 0.00015841579437255848,
44
+ "p50": 0.014200816154479979,
45
+ "p90": 0.014327548789978026,
46
+ "p95": 0.014343390369415282,
47
+ "p99": 0.014356063632965087,
48
+ "values": [
49
+ 0.014042400360107421,
50
+ 0.014359231948852538
51
+ ]
52
+ },
53
+ "throughput": {
54
+ "unit": "tokens/s",
55
+ "value": 492.9294150316626
56
+ },
57
+ "energy": null,
58
+ "efficiency": null
59
+ },
60
+ "decode": {
61
+ "memory": {
62
+ "unit": "MB",
63
+ "max_ram": 2084.560896,
64
+ "max_global_vram": 6060.244992,
65
+ "max_process_vram": 0.0,
66
+ "max_reserved": 5381.292032,
67
+ "max_allocated": 5282.647552
68
+ },
69
+ "latency": {
70
+ "unit": "s",
71
+ "count": 2,
72
+ "total": 3.1285546875000003,
73
+ "mean": 1.5642773437500002,
74
+ "stdev": 0.00018029785156248312,
75
+ "p50": 1.5642773437500002,
76
+ "p90": 1.5644215820312501,
77
+ "p95": 1.5644396118164063,
78
+ "p99": 1.5644540356445313,
79
+ "values": [
80
+ 1.5644576416015625,
81
+ 1.5640970458984376
82
+ ]
83
+ },
84
+ "throughput": {
85
+ "unit": "tokens/s",
86
+ "value": 81.18764904920651
87
+ },
88
+ "energy": null,
89
+ "efficiency": null
90
+ },
91
+ "per_token": {
92
+ "memory": null,
93
+ "latency": {
94
+ "unit": "s",
95
+ "count": 254,
96
+ "total": 3.128261626243589,
97
+ "mean": 0.012315990654502328,
98
+ "stdev": 0.00010128158543273827,
99
+ "p50": 0.012290047645568849,
100
+ "p90": 0.012329983711242675,
101
+ "p95": 0.01245911030769348,
102
+ "p99": 0.012875171537399293,
103
+ "values": [
104
+ 0.012876799583435058,
105
+ 0.012303359985351562,
106
+ 0.012290047645568849,
107
+ 0.012297216415405274,
108
+ 0.012314623832702636,
109
+ 0.012327936172485352,
110
+ 0.012281855583190919,
111
+ 0.012271615982055664,
112
+ 0.012286975860595703,
113
+ 0.012294143676757812,
114
+ 0.012288000106811523,
115
+ 0.012300288200378418,
116
+ 0.012291071891784668,
117
+ 0.012272640228271485,
118
+ 0.012279808044433594,
119
+ 0.012272640228271485,
120
+ 0.012286975860595703,
121
+ 0.012275712013244629,
122
+ 0.012308480262756348,
123
+ 0.012311552047729492,
124
+ 0.012289024353027344,
125
+ 0.012275712013244629,
126
+ 0.012292096138000488,
127
+ 0.012275712013244629,
128
+ 0.012285951614379884,
129
+ 0.012290047645568849,
130
+ 0.012271615982055664,
131
+ 0.012296192169189453,
132
+ 0.012283904075622559,
133
+ 0.012279808044433594,
134
+ 0.012283904075622559,
135
+ 0.012291071891784668,
136
+ 0.012291071891784668,
137
+ 0.012292096138000488,
138
+ 0.012280832290649414,
139
+ 0.012267519950866699,
140
+ 0.012294143676757812,
141
+ 0.012288000106811523,
142
+ 0.012292096138000488,
143
+ 0.012274687767028808,
144
+ 0.012273664474487305,
145
+ 0.01246720027923584,
146
+ 0.012308480262756348,
147
+ 0.012297216415405274,
148
+ 0.012311552047729492,
149
+ 0.012333056449890138,
150
+ 0.012309503555297852,
151
+ 0.012279808044433594,
152
+ 0.012294143676757812,
153
+ 0.012294143676757812,
154
+ 0.012285984039306641,
155
+ 0.012297183990478515,
156
+ 0.012300288200378418,
157
+ 0.012290047645568849,
158
+ 0.012285951614379884,
159
+ 0.012285951614379884,
160
+ 0.012275712013244629,
161
+ 0.012266495704650878,
162
+ 0.012280832290649414,
163
+ 0.012329983711242675,
164
+ 0.012306431770324706,
165
+ 0.01232588768005371,
166
+ 0.012297216415405274,
167
+ 0.01228492832183838,
168
+ 0.012277759552001954,
169
+ 0.012271615982055664,
170
+ 0.012281855583190919,
171
+ 0.012274687767028808,
172
+ 0.012286016464233399,
173
+ 0.01228486442565918,
174
+ 0.012285951614379884,
175
+ 0.012310527801513671,
176
+ 0.012294143676757812,
177
+ 0.012290047645568849,
178
+ 0.012278783798217773,
179
+ 0.01242521572113037,
180
+ 0.01232588768005371,
181
+ 0.012561408042907715,
182
+ 0.012838912010192872,
183
+ 0.012317695617675782,
184
+ 0.012289024353027344,
185
+ 0.012288000106811523,
186
+ 0.012308480262756348,
187
+ 0.012329983711242675,
188
+ 0.012899328231811523,
189
+ 0.012322815895080566,
190
+ 0.012290047645568849,
191
+ 0.01228492832183838,
192
+ 0.012298239707946777,
193
+ 0.012291071891784668,
194
+ 0.012275712013244629,
195
+ 0.012275712013244629,
196
+ 0.012326911926269531,
197
+ 0.012316672325134278,
198
+ 0.012310527801513671,
199
+ 0.012301312446594239,
200
+ 0.012281855583190919,
201
+ 0.012270591735839843,
202
+ 0.012291071891784668,
203
+ 0.012277759552001954,
204
+ 0.012356608390808106,
205
+ 0.012512255668640136,
206
+ 0.012548095703125,
207
+ 0.012496895790100097,
208
+ 0.012336128234863282,
209
+ 0.012388352394104005,
210
+ 0.012313599586486817,
211
+ 0.012270591735839843,
212
+ 0.012279808044433594,
213
+ 0.012289024353027344,
214
+ 0.012279808044433594,
215
+ 0.012296192169189453,
216
+ 0.012314623832702636,
217
+ 0.012288000106811523,
218
+ 0.012288000106811523,
219
+ 0.012288000106811523,
220
+ 0.01226854419708252,
221
+ 0.012317695617675782,
222
+ 0.012299263954162597,
223
+ 0.012344320297241211,
224
+ 0.012280832290649414,
225
+ 0.012296192169189453,
226
+ 0.012285951614379884,
227
+ 0.012274687767028808,
228
+ 0.012289024353027344,
229
+ 0.012286975860595703,
230
+ 0.012281855583190919,
231
+ 0.012308480262756348,
232
+ 0.012296192169189453,
233
+ 0.012456959724426269,
234
+ 0.012314623832702636,
235
+ 0.012289024353027344,
236
+ 0.012286975860595703,
237
+ 0.012312576293945313,
238
+ 0.012314623832702636,
239
+ 0.012361727714538574,
240
+ 0.012463104248046876,
241
+ 0.012866559982299805,
242
+ 0.012321791648864745,
243
+ 0.012299263954162597,
244
+ 0.012305407524108887,
245
+ 0.012301312446594239,
246
+ 0.012299263954162597,
247
+ 0.012280832290649414,
248
+ 0.012281855583190919,
249
+ 0.012277759552001954,
250
+ 0.01228492832183838,
251
+ 0.012309503555297852,
252
+ 0.012301312446594239,
253
+ 0.012298239707946777,
254
+ 0.012299263954162597,
255
+ 0.012288000106811523,
256
+ 0.012282879829406738,
257
+ 0.012282879829406738,
258
+ 0.012274687767028808,
259
+ 0.012282879829406738,
260
+ 0.012278783798217773,
261
+ 0.012301312446594239,
262
+ 0.012292096138000488,
263
+ 0.012294143676757812,
264
+ 0.012290047645568849,
265
+ 0.012301312446594239,
266
+ 0.012289024353027344,
267
+ 0.012322815895080566,
268
+ 0.01228492832183838,
269
+ 0.012307456016540527,
270
+ 0.012298239707946777,
271
+ 0.012873727798461914,
272
+ 0.012294143676757812,
273
+ 0.012280832290649414,
274
+ 0.012285951614379884,
275
+ 0.012290047645568849,
276
+ 0.012299263954162597,
277
+ 0.012308480262756348,
278
+ 0.012296192169189453,
279
+ 0.012288000106811523,
280
+ 0.012281855583190919,
281
+ 0.012271615982055664,
282
+ 0.012275712013244629,
283
+ 0.01228492832183838,
284
+ 0.012272640228271485,
285
+ 0.012292096138000488,
286
+ 0.012294143676757812,
287
+ 0.012288000106811523,
288
+ 0.012297216415405274,
289
+ 0.012786687850952149,
290
+ 0.012385279655456542,
291
+ 0.012288000106811523,
292
+ 0.012285951614379884,
293
+ 0.012286975860595703,
294
+ 0.012356608390808106,
295
+ 0.012301312446594239,
296
+ 0.01228492832183838,
297
+ 0.012280832290649414,
298
+ 0.012271615982055664,
299
+ 0.012295167922973632,
300
+ 0.012300288200378418,
301
+ 0.012288000106811523,
302
+ 0.012311552047729492,
303
+ 0.012303359985351562,
304
+ 0.012286975860595703,
305
+ 0.012282879829406738,
306
+ 0.012423168182373047,
307
+ 0.012306431770324706,
308
+ 0.012306431770324706,
309
+ 0.012283904075622559,
310
+ 0.012285951614379884,
311
+ 0.012288000106811523,
312
+ 0.012286975860595703,
313
+ 0.012288000106811523,
314
+ 0.012301312446594239,
315
+ 0.012318719863891601,
316
+ 0.012298239707946777,
317
+ 0.012296192169189453,
318
+ 0.012288000106811523,
319
+ 0.012303359985351562,
320
+ 0.012295167922973632,
321
+ 0.01227673625946045,
322
+ 0.012283904075622559,
323
+ 0.012272640228271485,
324
+ 0.01228492832183838,
325
+ 0.012281855583190919,
326
+ 0.01228492832183838,
327
+ 0.01228492832183838,
328
+ 0.012297216415405274,
329
+ 0.012312576293945313,
330
+ 0.012290047645568849,
331
+ 0.012274687767028808,
332
+ 0.012273664474487305,
333
+ 0.012278783798217773,
334
+ 0.012282879829406738,
335
+ 0.012286975860595703,
336
+ 0.012290047645568849,
337
+ 0.012291071891784668,
338
+ 0.012270591735839843,
339
+ 0.012283904075622559,
340
+ 0.012281855583190919,
341
+ 0.012278783798217773,
342
+ 0.012282879829406738,
343
+ 0.012319744110107422,
344
+ 0.012295167922973632,
345
+ 0.01226854419708252,
346
+ 0.012272640228271485,
347
+ 0.01228492832183838,
348
+ 0.01227673625946045,
349
+ 0.012294143676757812,
350
+ 0.012281855583190919,
351
+ 0.012271615982055664,
352
+ 0.012281855583190919,
353
+ 0.012300288200378418,
354
+ 0.012281855583190919,
355
+ 0.012920831680297852,
356
+ 0.012331007957458496,
357
+ 0.012274687767028808
358
+ ]
359
+ },
360
+ "throughput": {
361
+ "unit": "tokens/s",
362
+ "value": 81.19525485628981
363
+ },
364
+ "energy": null,
365
+ "efficiency": null
366
+ }
367
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/backend.cache_implementation=static,backend.model=google/gemma-2b,backend.torch_compile=True/summary.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "google/gemma-2b",
3
+ "commit": "65bb28444849976f853063edb958b3ef3dd59d12",
4
+ "config": "backend.cache_implementation=static,backend.torch_compile=True",
5
+ "metrics": {
6
+ "prefill.latency.mean": 0.014200816154479979,
7
+ "prefill.throughput.value": 492.9294150316626,
8
+ "decode.latency.mean": 1.5642773437500002,
9
+ "decode.throughput.value": 81.18764904920651,
10
+ "per_token.latency.mean": 0.012315990654502328,
11
+ "per_token.throughput.value": 81.19525485628981
12
+ }
13
+ }
2024-09-09/commit=65bb28444849976f853063edb958b3ef3dd59d12/multirun.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
4
+ sweep:
5
+ dir: _benchmark/tmplwp5r740/commit=65bb28444849976f853063edb958b3ef3dd59d12
6
+ subdir: ${hydra.job.override_dirname}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ root:
71
+ level: ERROR
72
+ disable_existing_loggers: true
73
+ job_logging:
74
+ version: 1
75
+ root:
76
+ level: ERROR
77
+ disable_existing_loggers: true
78
+ env: {}
79
+ mode: MULTIRUN
80
+ searchpath: []
81
+ callbacks: {}
82
+ output_subdir: .hydra
83
+ overrides:
84
+ hydra:
85
+ - hydra/job_logging=disabled
86
+ - hydra/hydra_logging=disabled
87
+ - hydra.sweep.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12
88
+ - hydra.run.dir=_benchmark/tmplwp5r740/commit\=65bb28444849976f853063edb958b3ef3dd59d12/${hydra.job.override_dirname}
89
+ - hydra.mode=MULTIRUN
90
+ task:
91
+ - backend.model=google/gemma-2b
92
+ - backend.cache_implementation=null,static
93
+ - backend.torch_compile=false,true
94
+ job:
95
+ name: cli
96
+ chdir: true
97
+ override_dirname: backend.cache_implementation=null,static,backend.model=google/gemma-2b,backend.torch_compile=false,true
98
+ id: ???
99
+ num: ???
100
+ config_name: generation
101
+ env_set:
102
+ OVERRIDE_BENCHMARKS: '1'
103
+ LOG_LEVEL: WARN
104
+ env_copy: []
105
+ config:
106
+ override_dirname:
107
+ kv_sep: '='
108
+ item_sep: ','
109
+ exclude_keys: []
110
+ runtime:
111
+ version: 1.3.2
112
+ version_base: '1.3'
113
+ cwd: /transformers
114
+ config_sources:
115
+ - path: hydra.conf
116
+ schema: pkg
117
+ provider: hydra
118
+ - path: optimum_benchmark
119
+ schema: pkg
120
+ provider: main
121
+ - path: /transformers/benchmark/config
122
+ schema: file
123
+ provider: command-line
124
+ - path: ''
125
+ schema: structured
126
+ provider: schema
127
+ output_dir: ???
128
+ choices:
129
+ backend: pytorch
130
+ launcher: process
131
+ scenario: inference
132
+ hydra/env: default
133
+ hydra/callbacks: null
134
+ hydra/job_logging: disabled
135
+ hydra/hydra_logging: disabled
136
+ hydra/hydra_help: default
137
+ hydra/help: default
138
+ hydra/sweeper: basic
139
+ hydra/launcher: basic
140
+ hydra/output: default
141
+ verbose: false
142
+ name: pytorch_generate
143
+ backend:
144
+ name: pytorch
145
+ version: 2.4.0+cu121
146
+ _target_: optimum_benchmark.backends.pytorch.backend.PyTorchBackend
147
+ task: null
148
+ library: null
149
+ model_type: null
150
+ model: google/gemma-2b
151
+ processor: null
152
+ device: cuda
153
+ device_ids: '0'
154
+ seed: 42
155
+ inter_op_num_threads: null
156
+ intra_op_num_threads: null
157
+ model_kwargs: {}
158
+ processor_kwargs: {}
159
+ no_weights: true
160
+ device_map: null
161
+ torch_dtype: float16
162
+ eval_mode: true
163
+ to_bettertransformer: false
164
+ low_cpu_mem_usage: null
165
+ attn_implementation: null
166
+ cache_implementation: static
167
+ autocast_enabled: false
168
+ autocast_dtype: null
169
+ torch_compile: true
170
+ torch_compile_target: forward
171
+ torch_compile_config:
172
+ backend: inductor
173
+ mode: reduce-overhead
174
+ fullgraph: true
175
+ quantization_scheme: null
176
+ quantization_config: {}
177
+ deepspeed_inference: false
178
+ deepspeed_inference_config: {}
179
+ peft_type: null
180
+ peft_config: {}
181
+ scenario:
182
+ name: inference
183
+ _target_: optimum_benchmark.scenarios.inference.scenario.InferenceScenario
184
+ iterations: 2
185
+ duration: 0
186
+ warmup_runs: 10
187
+ input_shapes:
188
+ batch_size: 1
189
+ sequence_length: 7
190
+ new_tokens: null
191
+ memory: true
192
+ latency: true
193
+ energy: false
194
+ forward_kwargs: {}
195
+ generate_kwargs:
196
+ max_new_tokens: 128
197
+ min_new_tokens: 128
198
+ do_sample: false
199
+ call_kwargs: {}
200
+ launcher:
201
+ name: process
202
+ _target_: optimum_benchmark.launchers.process.launcher.ProcessLauncher
203
+ device_isolation: true
204
+ device_isolation_action: warn
205
+ numactl: false
206
+ numactl_kwargs: {}
207
+ start_method: spawn
208
+ environment:
209
+ cpu: ' AMD EPYC 7R32'
210
+ cpu_count: 16
211
+ cpu_ram_mb: 66697.261056
212
+ system: Linux
213
+ machine: x86_64
214
+ platform: Linux-5.10.223-212.873.amzn2.x86_64-x86_64-with-glibc2.29
215
+ processor: x86_64
216
+ python_version: 3.8.10
217
+ gpu:
218
+ - NVIDIA A10G
219
+ gpu_count: 1
220
+ gpu_vram_mb: 24146608128
221
+ optimum_benchmark_version: 0.4.0
222
+ optimum_benchmark_commit: null
223
+ transformers_version: 4.45.0.dev0
224
+ transformers_commit: 65bb28444849976f853063edb958b3ef3dd59d12
225
+ accelerate_version: 0.35.0.dev0
226
+ accelerate_commit: null
227
+ diffusers_version: null
228
+ diffusers_commit: null
229
+ optimum_version: 1.22.0.dev0
230
+ optimum_commit: null
231
+ timm_version: 0.9.16
232
+ timm_commit: null
233
+ peft_version: 0.12.1.dev0
234
+ peft_commit: null
2024-09-09/summaries.json ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "model": "google/gemma-2b",
4
+ "commit": "65bb28444849976f853063edb958b3ef3dd59d12",
5
+ "config": "backend.cache_implementation=null,backend.torch_compile=False",
6
+ "metrics": {
7
+ "prefill.latency.mean": 0.02127622413635254,
8
+ "prefill.throughput.value": 329.00574627994285,
9
+ "decode.latency.mean": 2.5274194335937503,
10
+ "decode.throughput.value": 50.248881650568805,
11
+ "per_token.latency.mean": 0.01989969438267505,
12
+ "per_token.throughput.value": 50.25202803469252
13
+ }
14
+ },
15
+ {
16
+ "model": "google/gemma-2b",
17
+ "commit": "65bb28444849976f853063edb958b3ef3dd59d12",
18
+ "config": "backend.cache_implementation=static,backend.torch_compile=False",
19
+ "metrics": {
20
+ "prefill.latency.mean": 0.021814751625061038,
21
+ "prefill.throughput.value": 320.88378177812115,
22
+ "decode.latency.mean": 2.6390048828125,
23
+ "decode.throughput.value": 48.1242004617478,
24
+ "per_token.latency.mean": 0.020778556455777385,
25
+ "per_token.throughput.value": 48.12653863266591
26
+ }
27
+ },
28
+ {
29
+ "model": "google/gemma-2b",
30
+ "commit": "65bb28444849976f853063edb958b3ef3dd59d12",
31
+ "config": "backend.cache_implementation=static,backend.torch_compile=True",
32
+ "metrics": {
33
+ "prefill.latency.mean": 0.014200816154479979,
34
+ "prefill.throughput.value": 492.9294150316626,
35
+ "decode.latency.mean": 1.5642773437500002,
36
+ "decode.throughput.value": 81.18764904920651,
37
+ "per_token.latency.mean": 0.012315990654502328,
38
+ "per_token.throughput.value": 81.19525485628981
39
+ }
40
+ }
41
+ ]
2024-09-09/summary.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "google/gemma-2b": {
3
+ "backend.cache_implementation=null,backend.torch_compile=False": {
4
+ "65bb28444849976f853063edb958b3ef3dd59d12": {
5
+ "metrics": {
6
+ "prefill.latency.mean": 0.02127622413635254,
7
+ "prefill.throughput.value": 329.00574627994285,
8
+ "decode.latency.mean": 2.5274194335937503,
9
+ "decode.throughput.value": 50.248881650568805,
10
+ "per_token.latency.mean": 0.01989969438267505,
11
+ "per_token.throughput.value": 50.25202803469252
12
+ }
13
+ }
14
+ },
15
+ "backend.cache_implementation=static,backend.torch_compile=False": {
16
+ "65bb28444849976f853063edb958b3ef3dd59d12": {
17
+ "metrics": {
18
+ "prefill.latency.mean": 0.021814751625061038,
19
+ "prefill.throughput.value": 320.88378177812115,
20
+ "decode.latency.mean": 2.6390048828125,
21
+ "decode.throughput.value": 48.1242004617478,
22
+ "per_token.latency.mean": 0.020778556455777385,
23
+ "per_token.throughput.value": 48.12653863266591
24
+ }
25
+ }
26
+ },
27
+ "backend.cache_implementation=static,backend.torch_compile=True": {
28
+ "65bb28444849976f853063edb958b3ef3dd59d12": {
29
+ "metrics": {
30
+ "prefill.latency.mean": 0.014200816154479979,
31
+ "prefill.throughput.value": 492.9294150316626,
32
+ "decode.latency.mean": 1.5642773437500002,
33
+ "decode.throughput.value": 81.18764904920651,
34
+ "per_token.latency.mean": 0.012315990654502328,
35
+ "per_token.throughput.value": 81.19525485628981
36
+ }
37
+ }
38
+ }
39
+ }
40
+ }