Initial commit
This commit is contained in:
26
configs/a100.yaml
Normal file
26
configs/a100.yaml
Normal file
@@ -0,0 +1,26 @@
|
||||
# A100 Configuration
|
||||
gpu_type: a100
|
||||
gpu_model: "NVIDIA A100 80GB"
|
||||
|
||||
# Default attention implementation
|
||||
default_attention: flash_attention_2
|
||||
|
||||
# Pretraining defaults
|
||||
pretrain:
|
||||
batch_size: 8
|
||||
sequence_length: 8192
|
||||
num_steps: 10
|
||||
warmup_steps: 3
|
||||
|
||||
# Inference defaults
|
||||
inference:
|
||||
num_requests: 10
|
||||
prompt_length: 512
|
||||
generation_length: 100
|
||||
warmup_requests: 2
|
||||
|
||||
# Hardware specs (for reference)
|
||||
hardware:
|
||||
memory_gb: 80
|
||||
tdp_watts: 400
|
||||
compute_capability: "8.0"
|
||||
26
configs/h100.yaml
Normal file
26
configs/h100.yaml
Normal file
@@ -0,0 +1,26 @@
|
||||
# H100 Configuration
|
||||
gpu_type: h100
|
||||
gpu_model: "NVIDIA H100 80GB"
|
||||
|
||||
# Default attention implementation
|
||||
default_attention: flash_attention_3_hopper
|
||||
|
||||
# Pretraining defaults
|
||||
pretrain:
|
||||
batch_size: 8
|
||||
sequence_length: 8192
|
||||
num_steps: 10
|
||||
warmup_steps: 3
|
||||
|
||||
# Inference defaults
|
||||
inference:
|
||||
num_requests: 10
|
||||
prompt_length: 512
|
||||
generation_length: 100
|
||||
warmup_requests: 2
|
||||
|
||||
# Hardware specs (for reference)
|
||||
hardware:
|
||||
memory_gb: 80
|
||||
tdp_watts: 700
|
||||
compute_capability: "9.0"
|
||||
26
configs/h200.yaml
Normal file
26
configs/h200.yaml
Normal file
@@ -0,0 +1,26 @@
|
||||
# H200 Configuration
|
||||
gpu_type: h200
|
||||
gpu_model: "NVIDIA H200 141GB"
|
||||
|
||||
# Default attention implementation
|
||||
default_attention: flash_attention_3_hopper
|
||||
|
||||
# Pretraining defaults
|
||||
pretrain:
|
||||
batch_size: 8
|
||||
sequence_length: 8192
|
||||
num_steps: 10
|
||||
warmup_steps: 3
|
||||
|
||||
# Inference defaults
|
||||
inference:
|
||||
num_requests: 10
|
||||
prompt_length: 512
|
||||
generation_length: 100
|
||||
warmup_requests: 2
|
||||
|
||||
# Hardware specs (for reference)
|
||||
hardware:
|
||||
memory_gb: 141
|
||||
tdp_watts: 700
|
||||
compute_capability: "9.0"
|
||||
26
configs/mi300x.yaml
Normal file
26
configs/mi300x.yaml
Normal file
@@ -0,0 +1,26 @@
|
||||
# MI300X Configuration
|
||||
gpu_type: mi300x
|
||||
gpu_model: "AMD Instinct MI300X"
|
||||
|
||||
# Default attention implementation
|
||||
default_attention: flash_attention_2
|
||||
|
||||
# Pretraining defaults
|
||||
pretrain:
|
||||
batch_size: 8
|
||||
sequence_length: 8192
|
||||
num_steps: 10
|
||||
warmup_steps: 3
|
||||
|
||||
# Inference defaults
|
||||
inference:
|
||||
num_requests: 10
|
||||
prompt_length: 512
|
||||
generation_length: 100
|
||||
warmup_requests: 2
|
||||
|
||||
# Hardware specs (for reference)
|
||||
hardware:
|
||||
memory_gb: 192
|
||||
tdp_watts: 750
|
||||
compute_capability: "gfx940"
|
||||
Reference in New Issue
Block a user