meta-llama/Meta-Llama-3-8B-Instruct

Explore meta-llama/Meta-Llama-3-8B-Instruct: model specifications, capabilities, benchmark results, platform pricing and original sources.

huggingface/meta-llama/Meta-Llama-3-8B-Instruct

Specifications & capabilities

Published model ID
Meta-Llama-3-8B-Instruct

Evaluations

1–20 / 4,366

LLM-Perf

44,284,123.28

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-bnb-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
sdpa
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
bnb
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
1.0.0
config.environment.transformers_version
4.45.2
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0.dev0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.load_in_8bit
True
config.backend.quantization_config.llm_int8_threshold
0.0
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-bnb-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
8881dbe7854f05e69b9aab4dcbea2787440807926f27673ae979b31b27a8ac3e
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

39,457,320.002

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xT4.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
config.environment.gpu
['Tesla T4']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
8
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
33163.759616
config.environment.gpu_vram_mb
16106127360
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
sdpa
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
bnb
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.load_in_8bit
True
config.backend.quantization_config.llm_int8_threshold
0.0
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xT4.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
523288fcaa18357c5c83defd9fb4112c1638bd5c1b76ab9682f4ca2bfb46f5c1
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

35,219,133.848

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
sdpa
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
exllama
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.exllama_config.version
2
config.backend.quantization_config.exllama_config.max_input_len
64
config.backend.quantization_config.exllama_config.max_batch_size
1
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
79e9da6383189d5ee83387f2eb30bbfec80a4ea666266526460190d34968cd2b
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

34,274,275.338

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.1
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
1.0.0
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.2
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
exllama
config.environment.optimum_benchmark_version
0.5.0.dev0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.exllama_config.version
2
config.backend.quantization_config.exllama_config.max_input_len
64
config.backend.quantization_config.exllama_config.max_batch_size
1
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
b910ba8235e1f08fe71ca3832dccd92b2c4e6ce2533d1c48199050038aaca459
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

33,902,255.952

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
exllama
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.exllama_config.version
2
config.backend.quantization_config.exllama_config.max_input_len
64
config.backend.quantization_config.exllama_config.max_batch_size
1
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
3f47ce2494d905f467255de468c13dc93792cc5cfd7385cf6f67a7a4af6e7066
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

27,897,675.164

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.1
config.environment.python_version
3.10.12
config.backend.attn_implementation
flash_attention_2
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
1.0.0
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.2
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemv
config.environment.optimum_benchmark_version
0.5.0.dev0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
0a300caf14ac7a5b385649cce842847f7f0f246bda755f7b3c310ee031ae022e
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

27,266,341.74

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
flash_attention_2
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemv
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
27d5e1ffec68a58b03b1b38b52338a395f884ac022653d0435e2f766bb137d50
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

25,397,644.071

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.1
config.environment.python_version
3.10.12
config.backend.attn_implementation
flash_attention_2
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
1.0.0
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.2
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemm
config.environment.optimum_benchmark_version
0.5.0.dev0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
7b856e216071869cede1312bbae5da2e417e91bb007027cd918fff63cebc66a5
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

25,014,034.664

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
sdpa
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemm
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
f75501615157672e4742b962c735a9d09616dfd3721b243a1fbd44d9c8900a9b
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

24,831,409.16

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.1
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
1.0.0
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.2
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemm
config.environment.optimum_benchmark_version
0.5.0.dev0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
1181f5223ca360664de0df820b7a59aa829522fee8c92d966b9b4aaba44f1ef9
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

24,642,868.945

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
flash_attention_2
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemm
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
26df603cb9085faa38ccb24b3d03627c699215d3afcb074291ba0bcf7c7f401a
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

24,553,378.401

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemm
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-awq-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
2
benchmark_evaluation
true
comparison_configuration_sha256
db7563284a83fd438e1d7c5283fb69ef19218056e86eb0efa33ae540a3fbed0f
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

12,013,614.142

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xT4.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
config.environment.gpu
['Tesla T4']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
8
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
33163.759616
config.environment.gpu_vram_mb
16106127360
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
awq
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
1.0.0
config.backend.quantization_config.bits
4
config.environment.transformers_version
4.45.2
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.backend.quantization_config.version
gemm
config.environment.optimum_benchmark_version
0.5.0.dev0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cuda-awq-1xT4.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
d145b9dfe3cb0ae5f99b010414e8cb5a1d70e1845266cb695ad822432d7130bc
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

8,223,514.327

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xA10.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
AMD EPYC 7R32
config.environment.gpu
['NVIDIA A10G']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
16
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66697.261056
config.environment.gpu_vram_mb
24146608128
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
flash_attention_2
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
bnb
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.load_in_4bit
True
config.backend.quantization_config.llm_int8_threshold
0.0
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xA10.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
53d918a870cfe78d3c227dd3a6de63d1f25fbb9443c56972fc936f291455081c
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

5,530,072.528

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-unquantized-32vCPU-C7i.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cpu
config.backend.library
transformers
config.backend.version
2.4.1+cpu
config.environment.cpu
Intel(R) Xeon(R) Platinum 8488C
config.backend.eval_mode
True
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
96
config.environment.processor
x86_64
config.environment.cpu_ram_mb
199386.689536
config.backend.autocast_enabled
False
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-unquantized-32vCPU-C7i.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
27f26b47dd32bce2b980a84eff54928d82908a2e962c6cabd4106be28187eda7
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

5,497,073.409

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-unquantized-32vCPU-C7i.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cpu
config.backend.library
transformers
config.backend.version
2.4.1+cpu
config.environment.cpu
Intel(R) Xeon(R) Platinum 8488C
config.backend.eval_mode
True
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float32
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
96
config.environment.processor
x86_64
config.environment.cpu_ram_mb
199386.689536
config.backend.autocast_enabled
False
config.environment.python_version
3.10.12
config.backend.attn_implementation
sdpa
config.backend.deepspeed_inference
False
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-unquantized-32vCPU-C7i.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
e8794a68519aa33e291a36fb9b3f341f93efdf6fb0b99f8636fcaaaeac858dbc
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

5,434,090.69

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-unquantized-32vCPU-C7i.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cpu
config.backend.library
transformers
config.backend.version
2.4.1+cpu
config.environment.cpu
Intel(R) Xeon(R) Platinum 8488C
config.backend.eval_mode
True
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float32
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
96
config.environment.processor
x86_64
config.environment.cpu_ram_mb
199386.689536
config.backend.autocast_enabled
False
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-unquantized-32vCPU-C7i.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
e4d13b4d65c7ce90f9eee0991c364d34e78f1ba97fe004b8bf10eed002bbdac6
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

4,990,964.803

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cpu-unquantized-32vCPU-C7i.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cpu
config.backend.library
transformers
config.backend.version
2.4.1+cpu
config.environment.cpu
Intel(R) Xeon(R) Platinum 8488C
config.backend.eval_mode
True
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
32
config.environment.processor
x86_64
config.environment.cpu_ram_mb
66326.188032
config.backend.autocast_enabled
False
config.environment.python_version
3.10.12
config.backend.attn_implementation
sdpa
config.backend.deepspeed_inference
False
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
1.0.0
config.environment.transformers_version
4.45.2
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0.dev0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-pytorch-cpu-unquantized-32vCPU-C7i.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
488c0ec1f46936c3500d9cac5c4dbe636c33d4ca6fd229b0413ed45ae377c1ff
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

4,756,655.266

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xT4.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
config.environment.gpu
['Tesla T4']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
8
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
33163.759616
config.environment.gpu_vram_mb
16106127360
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
eager
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
bnb
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.load_in_4bit
True
config.backend.quantization_config.llm_int8_threshold
0.0
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xT4.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
255b8f5da484f53c40c676cee32db5e7301b45bc21c8ca2cb82ab9e50ba92d32
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

LLM-Perf

4,722,629.477

source id
llmPerf
leaderboard name
prefill efficiency value
dimension
runtime
metric
prefill efficiency value
rank
1
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xT4.csv
execution
config.backend.name
pytorch
config.backend.seed
42
config.backend.task
text-generation
config.backend.device
cuda
config.backend.library
transformers
config.backend.version
2.4.1+cu124
config.environment.cpu
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
config.environment.gpu
['Tesla T4']
config.backend.eval_mode
True
config.backend.device_ids
0
config.backend.model_type
llama
config.backend.no_weights
True
config.environment.system
Linux
config.backend.torch_dtype
float16
config.environment.machine
x86_64
config.backend.torch_compile
False
config.environment.cpu_count
8
config.environment.gpu_count
1
config.environment.processor
x86_64
config.environment.cpu_ram_mb
33163.759616
config.environment.gpu_vram_mb
16106127360
config.backend.autocast_enabled
False
config.environment.peft_version
0.13.0
config.environment.python_version
3.10.12
config.backend.attn_implementation
sdpa
config.backend.deepspeed_inference
False
config.backend.quantization_scheme
bnb
config.environment.optimum_version
1.22.0
config.backend.to_bettertransformer
False
config.backend.torch_compile_target
forward
config.environment.accelerate_version
0.34.2
config.environment.transformers_version
4.45.1
config.scenario.input_shapes.batch_size
1
config.scenario.input_shapes.num_choices
2
config.environment.optimum_benchmark_version
0.5.0
config.scenario.input_shapes.sequence_length
256
config.backend.model_kwargs.trust_remote_code
True
config.scenario.generate_kwargs.max_new_tokens
64
config.scenario.generate_kwargs.min_new_tokens
64
config.backend.quantization_config.load_in_4bit
True
config.backend.quantization_config.llm_int8_threshold
0.0
metric_kind
prefill efficiency value
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
optimum-benchmark/llm-perf-leaderboard/perf-df-bnb-1xT4.csv
published_at
2024-12-13T10:07:03Z
source_revision
2fa8d8c300c4eb6e744d5e35999db926c0c26e90e767e4c38da326eebe21a92b
metric_population
3
benchmark_evaluation
true
comparison_configuration_sha256
1e9c7c02933eee1da96187f067ad9af7c19695beddd813d22aef483af37fd725
complete_configuration_location
source record config and original CSV artifact
source model id
meta-llama/Meta-Llama-3-8B-Instruct
Original result ↗

Provider specifications & access

1–5 / 5

meta-llama/Meta-Llama-3-8B-Instruct ↗

source model id
HuggingFace/meta-llama-meta-llama-3-8b-instruct@4
hosting
foundry-managed-compute
platforms
  • Microsoft Foundry
license
llama3
architecture
input_modalities
  • text
output_modalities
  • text
capabilities

    meta-llama/Meta-Llama-3-8B-Instruct ↗

    source model id
    HuggingFace/meta-llama-meta-llama-3-8b-instruct@5
    hosting
    foundry-managed-compute
    platforms
    • Microsoft Foundry
    license
    llama3
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities

      meta-llama/Meta-Llama-3-8B-Instruct ↗

      source model id
      HuggingFace/meta-llama-meta-llama-3-8b-instruct@6
      hosting
      foundry-managed-compute
      platforms
      • Microsoft Foundry
      license
      llama3
      architecture
      input_modalities
      • text
      output_modalities
      • text
      capabilities

        meta-llama/Meta-Llama-3-8B-Instruct ↗

        source model id
        HuggingFace/meta-llama-meta-llama-3-8b-instruct@7
        hosting
        foundry-managed-compute
        platforms
        • Microsoft Foundry
        license
        llama3
        architecture
        input_modalities
        • text
        output_modalities
        • text
        capabilities

          LLaMA3_8B_Chat ↗

          source model id
          publishers/meta/models/llama3@meta-llama-3-8b-instruct
          hosting
          vertex-self-deployed
          platforms
          • Google Cloud Vertex AI
          architecture
          input_modalities
          • text
          output_modalities
          • text
          capabilities

            Safety assessments

            1–1 / 1

            Foundation Model Transparency Index 2025 ↗

            31 / 100

            source id
            stanfordFMTI
            canonical model id
            huggingface/meta-llama/Meta-Llama-3-8B-Instruct
            taxonomy
            transparency
            hazard category
            transparency
            hazard name
            Organization transparency
            policy area
            Meta
            assessment type
            organization-level disclosure
            metric
            disclosure_score
            result label
            31 / 100
            sample size
            100
            updated at
            2026-11-09T18:08:05.227Z

            Original sources

            1–18 / 18

            Hugging Face · Model reference ↗

            Google Cloud · Model page ↗

            Microsoft Foundry · Model page ↗

            Microsoft Foundry · Model page ↗

            Microsoft Foundry · Model page ↗

            Microsoft Foundry · Model page ↗

            Google Cloud · Model documentation ↗

            Microsoft Foundry · Model documentation ↗

            Epoch AI · Catalog ↗

            GitHub · Catalog ↗

            Google Cloud · Catalog ↗

            Hugging Face · Catalog ↗

            Hugging Face · Catalog ↗

            Hugging Face · Catalog ↗

            Hugging Face · Catalog ↗

            Hugging Face · Catalog ↗

            Hugging Face · Catalog ↗

            Microsoft Foundry · Catalog ↗

            Source links

            1–10 / 10

            Hugging Face model card

            1–1 / 1

            Explore more models

            Models

            Loading
            EVIDENCE / COMPARISON

            Benchmarks

            Explore every benchmark

            Loading

            These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

            Install TensorTrace