fix(nemotron-vl): align recipe environments and verification docs

Signed-off-by: Chen Cui <chcui@nvidia.com>

Author: cuichenx

docs/_static/model-verification/catalog-v1.json

@@ -130,6 +130,8 @@ navigation:
       path: ./nightly/pages/models/nemotron/nemotron3-super.mdx
     - page: Nemotron 3.5 Lightning
       path: ./nightly/pages/models/nemotron/nemotron3.5-lightning.mdx
+    - page: Nemotron 3.5 Super VL
+      path: ./nightly/pages/models/nemotron/nemotron3.5-super-vl.mdx
     - page: Nemotron 3 Ultra
       path: ./nightly/pages/models/nemotron/nemotron3-ultra.mdx
     - page: Nemotron Nano V2 VL

docs/fern/versions/nightly.yml

@@ -4,7 +4,7 @@
 
 Choose a model to open its interactive import/export and training configurations. Every option comes directly from an authoritative YAML verification card; independent fields are never combined into synthetic commands.
 
-Current inventory: **25 model cards** and **356 concrete configurations**.
+Current inventory: **26 model cards** and **374 concrete configurations**.
 
 Each model page lets you select import/export, pretraining, benchmarking, SFT, LoRA, or long-context SFT and immediately see the exact command and expected result for the recorded precision and GPU.
 
@@ -100,7 +100,7 @@ Choose a model to inspect its recorded import/export, training, and precision co
   <section class="verification-provider-card">
     <header class="verification-provider-heading">
       <strong>NVIDIA</strong>
-      <span>6 models</span>
+      <span>7 models</span>
     </header>
     <div class="verification-provider-models">
       <a class="verification-model-link" href="nemotron/nemotron-3-omni#verified-nemotron-3-nano-omni-30b-a3b-reasoning">
@@ -127,6 +127,10 @@ Choose a model to inspect its recorded import/export, training, and precision co
         <strong title="nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16">NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16</strong>
         <span class="verification-status verification-status--verified" title="Verified">✓ Verified</span>
       </a>
+      <a class="verification-model-link" href="nemotron/nemotron3.5-super-vl#verified-nemotron-3.5-super-vl-120b-a12b">
+        <strong title="nvidia/NVIDIA-Nemotron-3.5-Super-120B-A12B-SourceOfTruth">NVIDIA-Nemotron-3.5-Super-120B-A12B-SourceOfTruth</strong>
+        <span class="verification-status verification-status--partial" title="Partial">◐ Partial</span>
+      </a>
     </div>
   </section>
   <section class="verification-provider-card">

docs/fern/versions/nightly/pages/models/README.mdx

@@ -4,7 +4,7 @@
 
 Choose a model to open its interactive import/export and training configurations. Every option comes directly from an authoritative YAML verification card; independent fields are never combined into synthetic commands.
 
-Current inventory: **25 model cards** and **356 concrete configurations**.
+Current inventory: **26 model cards** and **374 concrete configurations**.
 
 Each model page lets you select import/export, pretraining, benchmarking, SFT, LoRA, or long-context SFT and immediately see the exact command and expected result for the recorded precision and GPU.
 
@@ -100,7 +100,7 @@ Choose a model to inspect its recorded import/export, training, and precision co
   <section class="verification-provider-card">
     <header class="verification-provider-heading">
       <strong>NVIDIA</strong>
-      <span>6 models</span>
+      <span>7 models</span>
     </header>
     <div class="verification-provider-models">
       <a class="verification-model-link" href="nemotron/nemotron-3-omni.html#verified-nemotron-3-nano-omni-30b-a3b-reasoning">
@@ -127,6 +127,10 @@ Choose a model to inspect its recorded import/export, training, and precision co
         <strong title="nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16">NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16</strong>
         <span class="verification-status verification-status--verified" title="Verified">✓ Verified</span>
       </a>
+      <a class="verification-model-link" href="nemotron/nemotron3.5-super-vl.html#verified-nemotron-3.5-super-vl-120b-a12b">
+        <strong title="nvidia/NVIDIA-Nemotron-3.5-Super-120B-A12B-SourceOfTruth">NVIDIA-Nemotron-3.5-Super-120B-A12B-SourceOfTruth</strong>
+        <span class="verification-status verification-status--partial" title="Partial">◐ Partial</span>
+      </a>
     </div>
   </section>
   <section class="verification-provider-card">

docs/fern/versions/nightly/pages/models/nemotron/nemotron3.5-super-vl.mdx

@@ -106,6 +106,7 @@
     "nemotron-3-super-120b-a12b": "models/nemotron/nemotron3-super.md",
     "nemotron-3-ultra-550b-a55b": "models/nemotron/nemotron3-ultra.md",
     "nemotron-3.5-lightning": "models/nemotron/nemotron3.5-lightning.md",
+    "nemotron-3.5-super-vl-120b-a12b": "models/nemotron/nemotron3.5-super-vl.md",
     "qwen3-235b-a22b": "models/qwen/qwen3-235b-a22b.md",
     "qwen3-30b-a3b": "models/qwen/qwen3-30b-a3b.md",
     "qwen3-8b": "models/qwen/qwen3-8b.md",

docs/models/README.md

@@ -184,12 +184,20 @@ def nemotron_35_super_vl_pretrain_64gpu_h100_bf16_config() -> ConfigContainer:
     cfg.model.use_te_rng_tracker = False
     cfg.rng.te_rng_tracker = False
 
-    cfg.env_vars["NVTE_BWD_LAYERNORM_SM_MARGIN"] = 0
-    cfg.env_vars["NVTE_FWD_LAYERNORM_SM_MARGIN"] = 0
     cfg.checkpoint.async_save = False
     cfg.model.freeze_language_model = False
     cfg.model.freeze_vision_model = False
     cfg.model.freeze_vision_projection = False
+    cfg.env_vars = {
+        **COMMON_RECIPE_ENV_VARS,
+        "CUDA_DEVICE_MAX_CONNECTIONS": 32,
+        "NUM_OF_HYBRID_EP_RANKS_PER_NVLINK_DOMAIN": 8,
+        "NUM_OF_TOKENS_PER_CHUNK_COMBINE_API": 64,
+        "NVLINK_DOMAIN_SIZE": 8,
+        "USE_MNNVL": 0,
+        "NVTE_BWD_LAYERNORM_SM_MARGIN": 0,
+        "NVTE_FWD_LAYERNORM_SM_MARGIN": 0,
+    }
     return cfg
 
 
@@ -231,6 +239,16 @@ def nemotron_35_super_vl_sft_64gpu_h100_bf16_config() -> ConfigContainer:
     cfg.optimizer.exp_avg_sq_dtype = torch.float32
     cfg.mixed_precision.grad_reduce_in_fp32 = True
     cfg.ddp.grad_reduce_in_fp32 = True
+    cfg.env_vars = {
+        **COMMON_RECIPE_ENV_VARS,
+        "CUDA_DEVICE_MAX_CONNECTIONS": 32,
+        "NUM_OF_HYBRID_EP_RANKS_PER_NVLINK_DOMAIN": 8,
+        "NUM_OF_TOKENS_PER_CHUNK_COMBINE_API": 64,
+        "NVLINK_DOMAIN_SIZE": 8,
+        "USE_MNNVL": 0,
+        "NVTE_BWD_LAYERNORM_SM_MARGIN": 20,
+        "NVTE_FWD_LAYERNORM_SM_MARGIN": 20,
+    }
     return cfg
 
 
@@ -347,8 +365,6 @@ def nemotron_35_super_vl_peft_16gpu_h100_bf16_config(
     cfg.env_vars = {
         **COMMON_RECIPE_ENV_VARS,
         "CUDA_DEVICE_MAX_CONNECTIONS": 1,
-        "NVLINK_DOMAIN_SIZE": 8,
-        "USE_MNNVL": 0,
     }
     return cfg
 

docs/models/nemotron/nemotron3.5-super-vl.md

@@ -265,7 +265,18 @@ def _nemotron_3_super_pretrain_64gpu_h100_bf16_config() -> ConfigContainer:
         BF16 pretraining configuration for 64 H100 GPUs.
     """
     cfg = nemotron_3_super_pretrain_16gpu_h100_bf16_config()
-    return _apply_nemotron_3_super_64gpu_h100_training_stack(cfg)
+    cfg = _apply_nemotron_3_super_64gpu_h100_training_stack(cfg)
+    cfg.env_vars = {
+        **COMMON_RECIPE_ENV_VARS,
+        "CUDA_DEVICE_MAX_CONNECTIONS": 32,
+        "NUM_OF_HYBRID_EP_RANKS_PER_NVLINK_DOMAIN": 8,
+        "NUM_OF_TOKENS_PER_CHUNK_COMBINE_API": 64,
+        "NVLINK_DOMAIN_SIZE": 8,
+        "USE_MNNVL": 0,
+        "NVTE_BWD_LAYERNORM_SM_MARGIN": 20,
+        "NVTE_FWD_LAYERNORM_SM_MARGIN": 20,
+    }
+    return cfg
 
 
 # =============================================================================

scripts/docs/generate_model_verification_catalog.py

@@ -271,6 +271,7 @@ def test_models_map_to_canonical_guides(generator: ModuleType, catalog: dict[str
     assert paths["muse-glimmer-30b"] == "models/muse_glimmer/muse-glimmer.md"
     assert paths["nemotron-3-nano-4b"] == "models/nemotron/nemotron3-nano-4b.md"
     assert paths["nemotron-3-super-120b-a12b"] == "models/nemotron/nemotron3-super.md"
+    assert paths["nemotron-3.5-super-vl-120b-a12b"] == "models/nemotron/nemotron3.5-super-vl.md"
     assert paths["nemotron-3-ultra-550b-a55b"] == "models/nemotron/nemotron3-ultra.md"
     assert paths["qwen3.8-27b"] == "models/qwen/qwen3.8-27b.md"