nvfp4_dynamic is default

Signed-off-by: Shengliang Xu <shengliangx@nvidia.com>

Author: shengliangxu

docs/source/guides/10_recipes.rst

@@ -205,7 +205,7 @@ list order), then inline keys last.
    cfg:
      $import: nvfp4
 
-   # Import + override — import nvfp4_dynamic, then override type inline
+   # Import + override — import nvfp4, then override type inline
    cfg:
      $import: nvfp4    # imports {num_bits: e2m1, block_sizes: {-1: 16, type: dynamic, ...}}
      block_sizes:
@@ -301,8 +301,8 @@ Reusable snippets are stored under ``modelopt_recipes/configs/``:
      - Description
    * - ``configs/numerics/fp8``
      - FP8 E4M3 quantizer attributes
-   * - ``configs/numerics/nvfp4_dynamic``
-     - NVFP4 E2M1 blockwise, dynamic calibration, FP8 scales
+   * - ``configs/numerics/nvfp4``
+     - NVFP4 E2M1 blockwise, dynamic calibration, FP8 scales (default)
    * - ``configs/numerics/nvfp4_static``
      - NVFP4 E2M1 blockwise, static calibration, FP8 scales
    * - ``configs/ptq/base_disable_all``
@@ -597,8 +597,8 @@ The ``modelopt_recipes/`` package is organized as follows:
    +-- configs/                    # Reusable config snippets (imported via $import)
        +-- numerics/               # Numeric format definitions
        |   +-- fp8.yml
-       |   +-- nvfp4_dynamic.yml
        |   +-- nvfp4_static.yml
+       |   +-- nvfp4.yml
        +-- ptq/                    # PTQ-specific entry snippets
            +-- base_disable_all.yaml
            +-- default_disabled_quantizers.yaml

…cipes/configs/numerics/nvfp4_dynamic.yml …elopt_recipes/configs/numerics/nvfp4.ymlmodelopt_recipes/configs/numerics/nvfp4_dynamic.yml renamed to modelopt_recipes/configs/numerics/nvfp4.yml modelopt_recipes/configs/numerics/nvfp4_dynamic.yml renamed to modelopt_recipes/configs/numerics/nvfp4.yml

@@ -13,7 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-# NVFP4 E2M1 blockwise with dynamic calibration and FP8 E4M3 scales.
+# NVFP4 E2M1 blockwise quantizer attributes with FP8 E4M3 scales (dynamic calibration, the default).
 num_bits: e2m1
 block_sizes:
   -1: 16

modelopt_recipes/configs/numerics/nvfp4_static.yml

@@ -13,7 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-# NVFP4 E2M1 blockwise with static calibration and FP8 E4M3 scales.
+# NVFP4 E2M1 blockwise quantizer attributes with FP8 E4M3 scales (static calibration).
 num_bits: e2m1
 block_sizes:
   -1: 16

modelopt_recipes/configs/ptq/w4a4_nvfp4_nvfp4.yaml

@@ -0,0 +1,25 @@
+# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# W4A4 NVFP4: NVFP4 E2M1 dynamic weight and activation quantizers.
+imports:
+  nvfp4: configs/numerics/nvfp4
+---
+  - quantizer_name: '*weight_quantizer'
+    cfg:
+      $import: nvfp4
+  - quantizer_name: '*input_quantizer'
+    cfg:
+      $import: nvfp4

modelopt_recipes/general/ptq/nvfp4_default-fp8_kv.yaml

@@ -16,7 +16,7 @@
 imports:
   base_disable_all: configs/ptq/base_disable_all
   default_disabled: configs/ptq/default_disabled_quantizers
-  nvfp4: configs/numerics/nvfp4_dynamic
+  w4a4_nvfp4_nvfp4: configs/ptq/w4a4_nvfp4_nvfp4
   fp8_kv: configs/ptq/fp8_kv
 
 metadata:
@@ -26,11 +26,6 @@ quantize:
   algorithm: max
   quant_cfg:
     - $import: base_disable_all
-    - quantizer_name: '*weight_quantizer'
-      cfg:
-        $import: nvfp4
-    - quantizer_name: '*input_quantizer'
-      cfg:
-        $import: nvfp4
+    - $import: w4a4_nvfp4_nvfp4
     - $import: fp8_kv
     - $import: default_disabled

modelopt_recipes/general/ptq/nvfp4_default-none_kv_gptq.yaml

@@ -17,7 +17,7 @@ imports:
   base_disable_all: configs/ptq/base_disable_all
   default_disabled: configs/ptq/default_disabled_quantizers
   nvfp4_static: configs/numerics/nvfp4_static
-  nvfp4_dynamic: configs/numerics/nvfp4_dynamic
+  nvfp4: configs/numerics/nvfp4
 
 metadata:
   recipe_type: ptq
@@ -33,7 +33,7 @@ quantize:
         $import: nvfp4_static
     - quantizer_name: '*input_quantizer'
       cfg:
-        $import: nvfp4_dynamic
+        $import: nvfp4
     - quantizer_name: '*[kv]_bmm_quantizer'
       enable: false
     - $import: default_disabled

modelopt_recipes/general/ptq/nvfp4_experts_only-fp8_kv.yaml

@@ -16,7 +16,7 @@
 imports:
   base_disable_all: configs/ptq/base_disable_all
   default_disabled: configs/ptq/default_disabled_quantizers
-  nvfp4: configs/numerics/nvfp4_dynamic
+  nvfp4: configs/numerics/nvfp4
   fp8_kv: configs/ptq/fp8_kv
 
 metadata:

modelopt_recipes/general/ptq/nvfp4_mlp_only-fp8_kv.yaml

@@ -16,7 +16,7 @@
 imports:
   base_disable_all: configs/ptq/base_disable_all
   default_disabled: configs/ptq/default_disabled_quantizers
-  nvfp4: configs/numerics/nvfp4_dynamic
+  nvfp4: configs/numerics/nvfp4
   fp8_kv: configs/ptq/fp8_kv
 
 metadata:

modelopt_recipes/general/ptq/nvfp4_omlp_only-fp8_kv.yaml

@@ -16,7 +16,7 @@
 imports:
   base_disable_all: configs/ptq/base_disable_all
   default_disabled: configs/ptq/default_disabled_quantizers
-  nvfp4: configs/numerics/nvfp4_dynamic
+  nvfp4: configs/numerics/nvfp4
   fp8_kv: configs/ptq/fp8_kv
 
 metadata: