Skip to content

Fix A4X MAX Kimi-k2 directory path naming #269

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tonyjohnchen merged 2 commits into from
Jul 23, 2026
Merged

Fix A4X MAX Kimi-k2 directory path naming #269

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4d10db5
Fix kimi-k2 nemo directory path
ngu3 Jul 22, 2026
4907ee6
Add A4X MAX training benchmarks table to README.md
ngu3 Jul 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,6 +112,22 @@ Models | GPU Machine Type
**Qwen-3-30B** | [A4X (NVIDIA GB200)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (25.11) | Pre-training | Slurm | [Link](./training/a4x/qwen3_30b_a3b/megatron-bridge-slurm/nemo2511/)


### Training benchmarks A4X MAX

Models | GPU Machine Type | Framework | Workload Type | Orchestrator | Link to the recipe
------------------ | ---------------------------------------------------------------------------------------------------- | --------- | ------------- | ------------ | ------------------
**DeepSeek-V3** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (25.11) | Pre-training | GKE | [Link](./training/a4x-max/deepseek-v3/megatron-bridge-gke/nemo2511/)
**DeepSeek-V3** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.04) | Pre-training | GKE | [Link](./training/a4x-max/deepseek-v3/megatron-bridge-gke/nemo2604/)
**DeepSeek-V3** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.06) | Pre-training | GKE | [Link](./training/a4x-max/deepseek-v3/megatron-bridge-gke/nemo2606/)
**GPT OSS 120B** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (25.11) | Pre-training | GKE | [Link](./training/a4x-max/gpt-oss-120b/megatron-bridge-gke/nemo2511/)
**GPT OSS 120B** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.04) | Pre-training | GKE | [Link](./training/a4x-max/gpt-oss-120b/megatron-bridge-gke/nemo2604/)
**Kimi-k2** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.04) | Pre-training | GKE | [Link](./training/a4x-max/kimi-k2/megatron-bridge-gke/nemo2604/)
**Llama-3.1-405B** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.02) | Pre-training | GKE | [Link](./training/a4x-max/llama3-1-405b/megatron-bridge-gke/nemo2602/)
**Llama-3.1-405B** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.04) | Pre-training | GKE | [Link](./training/a4x-max/llama3-1-405b/megatron-bridge-gke/nemo2604/)
**Qwen-3-235B** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.02) | Pre-training | GKE | [Link](./training/a4x-max/qwen3-235b-a22b/megatron-bridge-gke/nemo2602/)
**Qwen-3-235B** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.04) | Pre-training | GKE | [Link](./training/a4x-max/qwen3-235b-a22b/megatron-bridge-gke/nemo2604/)
**Wan-2.1-14B** | [A4X MAX (NVIDIA GB300)](https://cloud.google.com/compute/docs/accelerator-optimized-machines#a4x-vms) | Megatron-Bridge (26.02) | Pre-training | GKE | [Link](./training/a4x-max/wan21-14b/megatron-bridge-gke/nemo2602/)

### Inference benchmarks A3 Mega

| Models | GPU Machine Type | Framework | Workload Type | Orchestrator | Link to the recipe |
Expand Down
0 ...-fp8mx-seq4096-gbs16384/recipe/Chart.yaml → ...-fp8mx-seq4096-gbs16384/recipe/Chart.yaml
View file
Open in desktop
File renamed without changes.
0 ...s-fp8mx-seq4096-gbs16384/recipe/README.md → ...s-fp8mx-seq4096-gbs16384/recipe/README.md
View file
Open in desktop
File renamed without changes.
0 ...fp8mx-seq4096-gbs16384/recipe/launcher.sh → ...fp8mx-seq4096-gbs16384/recipe/launcher.sh
View file
Open in desktop
File renamed without changes.
0 .../templates/workload-config-configmap.yaml → .../templates/workload-config-configmap.yaml
View file
Open in desktop
File renamed without changes.
0 ...s16384/recipe/templates/workload-job.yaml → ...s16384/recipe/templates/workload-job.yaml
View file
Open in desktop
File renamed without changes.
0 ...emplates/workload-launcher-configmap.yaml → ...emplates/workload-launcher-configmap.yaml
View file
Open in desktop
File renamed without changes.
0 ...s16384/recipe/templates/workload-svc.yaml → ...s16384/recipe/templates/workload-svc.yaml
View file
Open in desktop
File renamed without changes.
0 ...fp8mx-seq4096-gbs16384/recipe/values.yaml → ...fp8mx-seq4096-gbs16384/recipe/values.yaml
View file
Open in desktop
File renamed without changes.
0 ...s-fp8mx-seq4096-gbs4096/recipe/Chart.yaml → ...s-fp8mx-seq4096-gbs4096/recipe/Chart.yaml
View file
Open in desktop
File renamed without changes.
2 changes: 1 addition & 1 deletion ...us-fp8mx-seq4096-gbs4096/recipe/README.md → ...us-fp8mx-seq4096-gbs4096/recipe/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,7 @@ Clone the `gpu-recipes` repository and set a reference to the recipe folder.
git clone https://github.com/ai-hypercomputer/gpu-recipes.git
cd gpu-recipes
export REPO_ROOT=`git rev-parse --show-toplevel`
export RECIPE_ROOT=$REPO_ROOT/training/a4x-max/kimi-k2/megatron-bridge-gke/nemo260401/256gpus-fp8mx-seq4096-gbs4096/recipe
export RECIPE_ROOT=$REPO_ROOT/training/a4x-max/kimi-k2/megatron-bridge-gke/nemo2604/256gpus-fp8mx-seq4096-gbs4096/recipe
cd $RECIPE_ROOT
```

Expand Down
0 ...-fp8mx-seq4096-gbs4096/recipe/launcher.sh → ...-fp8mx-seq4096-gbs4096/recipe/launcher.sh
View file
Open in desktop
File renamed without changes.
0 .../templates/workload-config-configmap.yaml → .../templates/workload-config-configmap.yaml
View file
Open in desktop
File renamed without changes.
0 ...bs4096/recipe/templates/workload-job.yaml → ...bs4096/recipe/templates/workload-job.yaml
View file
Open in desktop
File renamed without changes.
0 ...emplates/workload-launcher-configmap.yaml → ...emplates/workload-launcher-configmap.yaml
View file
Open in desktop
File renamed without changes.
0 ...bs4096/recipe/templates/workload-svc.yaml → ...bs4096/recipe/templates/workload-svc.yaml
View file
Open in desktop
File renamed without changes.
0 ...-fp8mx-seq4096-gbs4096/recipe/values.yaml → ...-fp8mx-seq4096-gbs4096/recipe/values.yaml
View file
Open in desktop
File renamed without changes.
0 ...s-fp8mx-seq4096-gbs8192/recipe/Chart.yaml → ...s-fp8mx-seq4096-gbs8192/recipe/Chart.yaml
View file
Open in desktop
File renamed without changes.
2 changes: 1 addition & 1 deletion ...us-fp8mx-seq4096-gbs8192/recipe/README.md → ...us-fp8mx-seq4096-gbs8192/recipe/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,7 @@ Clone the `gpu-recipes` repository and set a reference to the recipe folder.
git clone https://github.com/ai-hypercomputer/gpu-recipes.git
cd gpu-recipes
export REPO_ROOT=`git rev-parse --show-toplevel`
export RECIPE_ROOT=$REPO_ROOT/training/a4x-max/kimi-k2/megatron-bridge-gke/nemo260401/512gpus-fp8mx-seq4096-gbs8192/recipe
export RECIPE_ROOT=$REPO_ROOT/training/a4x-max/kimi-k2/megatron-bridge-gke/nemo2604/512gpus-fp8mx-seq4096-gbs8192/recipe
cd $RECIPE_ROOT
```

Expand Down
0 ...-fp8mx-seq4096-gbs8192/recipe/launcher.sh → ...-fp8mx-seq4096-gbs8192/recipe/launcher.sh
View file
Open in desktop
File renamed without changes.
0 .../templates/workload-config-configmap.yaml → .../templates/workload-config-configmap.yaml
View file
Open in desktop
File renamed without changes.
0 ...bs8192/recipe/templates/workload-job.yaml → ...bs8192/recipe/templates/workload-job.yaml
View file
Open in desktop
File renamed without changes.
0 ...emplates/workload-launcher-configmap.yaml → ...emplates/workload-launcher-configmap.yaml
View file
Open in desktop
File renamed without changes.
0 ...bs8192/recipe/templates/workload-svc.yaml → ...bs8192/recipe/templates/workload-svc.yaml
View file
Open in desktop
File renamed without changes.
0 ...-fp8mx-seq4096-gbs8192/recipe/values.yaml → ...-fp8mx-seq4096-gbs8192/recipe/values.yaml
View file
Open in desktop
File renamed without changes.