docs(models): re-verify Nemotron Super H100 long-context SFT at a recipe-containing revision

The sft_long_context.H100 item was marked verified at 44f3031, a revision
where its --recipe nemotron_3_super_sft_16gpu_h100_bf16_32k_config did not
exist (only the earlier _lowmem_ name did). Drop the item pin so it inherits
the card default ffeb5a7, which contains the recipe, and record the date and
metrics from a fresh 16xH100 run of the published command at that revision.
Regenerate the verification catalog and model pages.

Detected by: megatron-bridge QA (NVBug 6631246)

Signed-off-by: Pruthviraj Prakash <pruprakash@nvidia.com>

Author: pruprakash

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

@@ -4818,7 +4818,7 @@
           "workflow": "sft_export_inference"
         },
         {
-          "bridge_commit": "44f3031a2e77ff455eb128efbaa19322ad6397a6",
+          "bridge_commit": "ffeb5a7ae817a5a6ccfa3850d7dee937faad7f0b",
           "command_topologies": [
             {
               "gpus_per_node": "8",
@@ -4837,15 +4837,15 @@
             "sequence_packing": "offline"
           },
           "entry_id": "nemotron-3-super-120b-a12b-sft-long-context-h100",
-          "expected_result": "On 16 H100s, the dedicated TP1/PP8/CP2/EP2 recipe completes exactly ten packed 32K SFT steps with pad-4 alignment, full recompute, and full optimizer CPU offload. Loss is finite from 1.168852 to 1.008992 with no skipped or NaN iterations. The ten-step metric window includes the first-step cold-start cost by policy, the resolved configuration is persisted, and the final model-only checkpoint saves successfully.\n",
+          "expected_result": "On 16 H100s, the dedicated TP1/PP8/CP2/EP2 recipe completes exactly ten packed 32K SFT steps with pad-4 alignment, full recompute, and full optimizer CPU offload. Loss is finite from 1.168852 to 1.011637 with no skipped or NaN iterations. The ten-step metric window includes the first-step cold-start cost by policy, the resolved configuration is persisted, and the final model-only checkpoint saves successfully.\n",
           "hardware": "H100",
-          "last_verified": "2026-07-24",
+          "last_verified": "2026-10-08",
           "metrics": {
-            "final_loss": 1.008992,
+            "final_loss": 1.011637,
             "initial_loss": 1.168852,
-            "last_10_steps_model_tflops_per_gpu_avg": 5.58,
-            "last_10_steps_step_time_ms_avg": 132859.58,
-            "last_10_steps_tokens_per_second_per_gpu_avg": 30.83
+            "last_10_steps_model_tflops_per_gpu_avg": 6.89,
+            "last_10_steps_step_time_ms_avg": 123114.4,
+            "last_10_steps_tokens_per_second_per_gpu_avg": 33.27
           },
           "precision": "bf16",
           "source_pointer": "items.sft_long_context.H100",

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

@@ -462,7 +462,7 @@ Choose a workflow, precision, and exact recorded combination. The command and ex
       <dl class="verification-model-detail-meta">
         <div><dt>Hardware</dt><dd>H100</dd></div>
         <div><dt>Precision</dt><dd>BF16</dd></div>
-        <div><dt>Last verified</dt><dd>2026-07-24</dd></div>
+        <div><dt>Last verified</dt><dd>2026-10-08</dd></div>
       </dl>
       <section class="verification-recorded-metrics">
         <h5>Recorded metrics</h5>
@@ -473,19 +473,19 @@ Choose a workflow, precision, and exact recorded combination. The command and ex
           </div>
           <div>
             <dt>Final loss</dt>
-            <dd>1.008992</dd>
+            <dd>1.011637</dd>
           </div>
           <div>
             <dt>Step time · last 10 avg</dt>
-            <dd>132,859.580 ms</dd>
+            <dd>123,114.400 ms</dd>
           </div>
           <div>
             <dt>Model throughput · last 10 avg</dt>
-            <dd>5.580 TFLOP/s/GPU</dd>
+            <dd>6.890 TFLOP/s/GPU</dd>
           </div>
           <div>
             <dt>Token throughput · last 10 avg</dt>
-            <dd>30.830 tokens/s/GPU</dd>
+            <dd>33.270 tokens/s/GPU</dd>
           </div>
         </dl>
       </section>
@@ -501,7 +501,7 @@ Choose a workflow, precision, and exact recorded combination. The command and ex
       </section>
       <section class="verification-expected-result">
         <h5>Expected result</h5>
-        <p>On 16 H100s, the dedicated TP1/PP8/CP2/EP2 recipe completes exactly ten packed 32K SFT steps with pad-4 alignment, full recompute, and full optimizer CPU offload. Loss is finite from 1.168852 to 1.008992 with no skipped or NaN iterations. The ten-step metric window includes the first-step cold-start cost by policy, the resolved configuration is persisted, and the final model-only checkpoint saves successfully.</p>
+        <p>On 16 H100s, the dedicated TP1/PP8/CP2/EP2 recipe completes exactly ten packed 32K SFT steps with pad-4 alignment, full recompute, and full optimizer CPU offload. Loss is finite from 1.168852 to 1.011637 with no skipped or NaN iterations. The ten-step metric window includes the first-step cold-start cost by policy, the resolved configuration is persisted, and the final model-only checkpoint saves successfully.</p>
       </section>
     </article>
     <article id="nemotron-3-super-120b-a12b-sft-long-context-gb200" class="verification-model-detail" data-entry-detail="nemotron-3-super-120b-a12b-sft-long-context-gb200" tabindex="-1">

docs/models/nemotron/nemotron3-super.md

@@ -470,7 +470,7 @@ Choose a workflow, precision, and exact recorded combination. The command and ex
       <dl class="verification-model-detail-meta">
         <div><dt>Hardware</dt><dd>H100</dd></div>
         <div><dt>Precision</dt><dd>BF16</dd></div>
-        <div><dt>Last verified</dt><dd>2026-07-24</dd></div>
+        <div><dt>Last verified</dt><dd>2026-10-08</dd></div>
       </dl>
       <section class="verification-recorded-metrics">
         <h5>Recorded metrics</h5>
@@ -481,19 +481,19 @@ Choose a workflow, precision, and exact recorded combination. The command and ex
           </div>
           <div>
             <dt>Final loss</dt>
-            <dd>1.008992</dd>
+            <dd>1.011637</dd>
           </div>
           <div>
             <dt>Step time · last 10 avg</dt>
-            <dd>132,859.580 ms</dd>
+            <dd>123,114.400 ms</dd>
           </div>
           <div>
             <dt>Model throughput · last 10 avg</dt>
-            <dd>5.580 TFLOP/s/GPU</dd>
+            <dd>6.890 TFLOP/s/GPU</dd>
           </div>
           <div>
             <dt>Token throughput · last 10 avg</dt>
-            <dd>30.830 tokens/s/GPU</dd>
+            <dd>33.270 tokens/s/GPU</dd>
           </div>
         </dl>
       </section>
@@ -509,7 +509,7 @@ Choose a workflow, precision, and exact recorded combination. The command and ex
       </section>
       <section class="verification-expected-result">
         <h5>Expected result</h5>
-        <p>On 16 H100s, the dedicated TP1/PP8/CP2/EP2 recipe completes exactly ten packed 32K SFT steps with pad-4 alignment, full recompute, and full optimizer CPU offload. Loss is finite from 1.168852 to 1.008992 with no skipped or NaN iterations. The ten-step metric window includes the first-step cold-start cost by policy, the resolved configuration is persisted, and the final model-only checkpoint saves successfully.
+        <p>On 16 H100s, the dedicated TP1/PP8/CP2/EP2 recipe completes exactly ten packed 32K SFT steps with pad-4 alignment, full recompute, and full optimizer CPU offload. Loss is finite from 1.168852 to 1.011637 with no skipped or NaN iterations. The ten-step metric window includes the first-step cold-start cost by policy, the resolved configuration is persisted, and the final model-only checkpoint saves successfully.
 </p>
       </section>
     </article>

examples/model_verification_cards/nemotron-3-super-120b-a12b/card.yaml

@@ -399,7 +399,6 @@ items:
     H100:
       status: verified
       precision: bf16
-      bridge_commit: 44f3031a2e77ff455eb128efbaa19322ad6397a6  # pragma: allowlist secret
       enabled_features:
         sequence_packing: offline
         context_parallel_size: 2
@@ -431,17 +430,17 @@ items:
         --save_dir work/model-verification/nemotron-3-super-120b-a12b/h100-long-sft
         --save_interval 10 logger.log_interval=1 logger.log_throughput=true
         logger.tensorboard_dir=null
-      last_verified: 2026-07-24
+      last_verified: 2026-10-08
       metrics:
         initial_loss: 1.168852
-        final_loss: 1.008992
-        last_10_steps_step_time_ms_avg: 132859.580
-        last_10_steps_model_tflops_per_gpu_avg: 5.580
-        last_10_steps_tokens_per_second_per_gpu_avg: 30.830
+        final_loss: 1.011637
+        last_10_steps_step_time_ms_avg: 123114.400
+        last_10_steps_model_tflops_per_gpu_avg: 6.890
+        last_10_steps_tokens_per_second_per_gpu_avg: 33.270
       expected_result: >
         On 16 H100s, the dedicated TP1/PP8/CP2/EP2 recipe completes exactly
         ten packed 32K SFT steps with pad-4 alignment, full recompute, and full
-        optimizer CPU offload. Loss is finite from 1.168852 to 1.008992 with no
+        optimizer CPU offload. Loss is finite from 1.168852 to 1.011637 with no
         skipped or NaN iterations. The ten-step metric window includes the
         first-step cold-start cost by policy, the resolved configuration is
         persisted, and the final model-only checkpoint saves successfully.