bridge.recipes.glm.gb200.glm5

GB200 recipes for GLM-5 and GLM-5.2.

Module Contents

Functions

glm52_pretrain_192gpu_gb200_bf16_config	GLM-5.2 bounded pretraining on 192 GB200 GPUs.
glm5_pretrain_192gpu_gb200_bf16_config	GLM-5 bounded pretraining on 192 GB200 GPUs.
_glm_pretrain_config	Build bounded pretraining with the selected variant’s native model config.
glm52_sft_192gpu_gb200_bf16_config	GLM-5.2 bounded full SFT on 192 GB200 GPUs.
glm5_sft_192gpu_gb200_bf16_config	GLM-5 bounded full SFT on 192 GB200 GPUs.
_glm_sft_config	Build bounded full SFT with the selected variant’s native model config.
glm52_sft_192gpu_gb200_bf16_128k_config	GLM-5.2 128K packed SFT with context parallelism on 192 GB200 GPUs.
glm5_sft_192gpu_gb200_bf16_128k_config	GLM-5 128K packed SFT with context parallelism on 192 GB200 GPUs.
_glm_sft_128k_config	Build 128K packed SFT with context parallelism with the selected variant’s native model config.
glm52_peft_192gpu_gb200_bf16_config	GLM-5.2 bounded PEFT on 192 GB200 GPUs.
glm5_peft_192gpu_gb200_bf16_config	GLM-5 bounded PEFT on 192 GB200 GPUs.
_glm_peft_config	Build bounded PEFT with the selected variant’s native model config.

Data

_GLM5_MODEL_ID
_GLM5_MODEL_REVISION
_GLM52_MODEL_ID
_GLM52_MODEL_REVISION
_TULU3_REVISION
glm52_gb200_sft_config
__all__

API

bridge.recipes.glm.gb200.glm5._GLM5_MODEL_ID

‘zai-org/GLM-5’

bridge.recipes.glm.gb200.glm5._GLM5_MODEL_REVISION

‘4e6698ba8e85059d749020e3c4d2123719f23926’

bridge.recipes.glm.gb200.glm5._GLM52_MODEL_ID

‘zai-org/GLM-5.2’

bridge.recipes.glm.gb200.glm5._GLM52_MODEL_REVISION

‘4d67f66cc64d3219133b767c253b2ad1425c6c88’

bridge.recipes.glm.gb200.glm5._TULU3_REVISION

‘b14afda60f1bbebe55d5d2fa1e4df5042f97f8be’

bridge.recipes.glm.gb200.glm5.glm52_pretrain_192gpu_gb200_bf16_config() → megatron.bridge.training.config.ConfigContainer

GLM-5.2 bounded pretraining on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5.glm5_pretrain_192gpu_gb200_bf16_config() → megatron.bridge.training.config.ConfigContainer

GLM-5 bounded pretraining on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5._glm_pretrain_config(

*,

model_id: str,

revision: str,

) → megatron.bridge.training.config.ConfigContainer

Build bounded pretraining with the selected variant’s native model config.

bridge.recipes.glm.gb200.glm5.glm52_sft_192gpu_gb200_bf16_config() → megatron.bridge.training.config.ConfigContainer

GLM-5.2 bounded full SFT on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5.glm5_sft_192gpu_gb200_bf16_config() → megatron.bridge.training.config.ConfigContainer

GLM-5 bounded full SFT on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5._glm_sft_config(

*,

model_id: str,

revision: str,

data_slug: str,

) → megatron.bridge.training.config.ConfigContainer

Build bounded full SFT with the selected variant’s native model config.

bridge.recipes.glm.gb200.glm5.glm52_sft_192gpu_gb200_bf16_128k_config() → megatron.bridge.training.config.ConfigContainer

GLM-5.2 128K packed SFT with context parallelism on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5.glm5_sft_192gpu_gb200_bf16_128k_config() → megatron.bridge.training.config.ConfigContainer

GLM-5 128K packed SFT with context parallelism on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5._glm_sft_128k_config(

*,

model_id: str,

revision: str,

data_slug: str,

) → megatron.bridge.training.config.ConfigContainer

Build 128K packed SFT with context parallelism with the selected variant’s native model config.

bridge.recipes.glm.gb200.glm5.glm52_peft_192gpu_gb200_bf16_config(

peft_scheme: str | megatron.bridge.peft.base.PEFT = 'lora',

) → megatron.bridge.training.config.ConfigContainer

GLM-5.2 bounded PEFT on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5.glm5_peft_192gpu_gb200_bf16_config(

peft_scheme: str | megatron.bridge.peft.base.PEFT = 'lora',

) → megatron.bridge.training.config.ConfigContainer

GLM-5 bounded PEFT on 192 GB200 GPUs.

bridge.recipes.glm.gb200.glm5._glm_peft_config(

*,

model_id: str,

revision: str,

data_slug: str,

peft_scheme: str | megatron.bridge.peft.base.PEFT,

) → megatron.bridge.training.config.ConfigContainer

Build bounded PEFT with the selected variant’s native model config.

bridge.recipes.glm.gb200.glm5.glm52_gb200_sft_config

None

bridge.recipes.glm.gb200.glm5.__all__

[‘glm5_peft_192gpu_gb200_bf16_config’, ‘glm5_pretrain_192gpu_gb200_bf16_config’, ‘glm5_sft_192gpu_gb…