`bridge.perf_recipes.qwen_vl.h100.qwen3_vl`#

H100 performance recipes for Qwen3-VL.

Module Contents#

Functions#

`qwen3_vl_235b_a22b_pretrain_256gpu_h100_bf16_config`	Qwen3-VL 235B-A22B pretrain: 256× H100, BF16, TP=2 PP=8 VP=4 EP=32.
`qwen3_vl_235b_a22b_pretrain_256gpu_h100_fp8cs_config`	Qwen3-VL 235B-A22B pretrain: 256× H100, FP8 current-scaling, TP=2 PP=8 VP=4 EP=32.
`qwen3_vl_30b_a3b_pretrain_16gpu_h100_bf16_config`	Qwen3-VL 30B-A3B pretrain: 16× H100, BF16, PP=2 VP=12 EP=8.
`qwen3_vl_30b_a3b_pretrain_16gpu_h100_fp8cs_config`	Qwen3-VL 30B-A3B pretrain: 16× H100, FP8 current-scaling, PP=2 VP=12 EP=8.

API#

bridge.perf_recipes.qwen_vl.h100.qwen3_vl.qwen3_vl_235b_a22b_pretrain_256gpu_h100_bf16_config() → megatron.bridge.perf_recipes.qwen_vl.common.ConfigContainer#: Qwen3-VL 235B-A22B pretrain: 256× H100, BF16, TP=2 PP=8 VP=4 EP=32.

bridge.perf_recipes.qwen_vl.h100.qwen3_vl.qwen3_vl_235b_a22b_pretrain_256gpu_h100_fp8cs_config() → megatron.bridge.perf_recipes.qwen_vl.common.ConfigContainer#: Qwen3-VL 235B-A22B pretrain: 256× H100, FP8 current-scaling, TP=2 PP=8 VP=4 EP=32.

bridge.perf_recipes.qwen_vl.h100.qwen3_vl.qwen3_vl_30b_a3b_pretrain_16gpu_h100_bf16_config() → megatron.bridge.perf_recipes.qwen_vl.common.ConfigContainer#: Qwen3-VL 30B-A3B pretrain: 16× H100, BF16, PP=2 VP=12 EP=8.

bridge.perf_recipes.qwen_vl.h100.qwen3_vl.qwen3_vl_30b_a3b_pretrain_16gpu_h100_fp8cs_config() → megatron.bridge.perf_recipes.qwen_vl.common.ConfigContainer#: Qwen3-VL 30B-A3B pretrain: 16× H100, FP8 current-scaling, PP=2 VP=12 EP=8.