Extend docs with quantizers/quantized_tensors/custom_recipe (#2428)

* Extend docs with quantizers/quantized_tensors/custom_recipe
Signed-off-by: Evgeny <etsykunov@nvidia.com>

* Bring structure, reduce redundant members
Signed-off-by: Evgeny <etsykunov@nvidia.com>

---------
Signed-off-by: Evgeny <etsykunov@nvidia.com>

docs/api/common.rst

@@ -17,3 +17,5 @@ Common API
 .. autoapiclass:: transformer_engine.common.recipe.Float8CurrentScaling(fp8_format=Format.HYBRID)
 
 .. autoapiclass:: transformer_engine.common.recipe.Float8BlockScaling(fp8_format=Format.E4M3)
+
+.. autoapiclass:: transformer_engine.common.recipe.CustomRecipe(qfactory, fp8_dpa=False, fp8_mha=False)

docs/api/pytorch.rst

@@ -85,3 +85,51 @@ pyTorch
 
 .. autoapiclass:: transformer_engine.pytorch.UserBufferQuantizationMode
   :members: FP8, NONE
+
+Quantized tensors
+-----------------
+
+.. autoapiclass:: transformer_engine.pytorch.QuantizedTensorStorage
+   :members: update_usage, prepare_for_saving, restore_from_saved
+
+.. autoapiclass:: transformer_engine.pytorch.QuantizedTensor(shape, dtype, *, requires_grad=False, device=None)
+   :members: dequantize, quantize_
+
+.. autoapiclass:: transformer_engine.pytorch.Float8TensorStorage(data, fp8_scale_inv, fp8_dtype, data_transpose=None, quantizer=None)
+
+.. autoapiclass:: transformer_engine.pytorch.MXFP8TensorStorage(rowwise_data, rowwise_scale_inv, columnwise_data, columnwise_scale_inv, fp8_dtype, quantizer)
+
+.. autoapiclass:: transformer_engine.pytorch.Float8BlockwiseQTensorStorage(rowwise_data, rowwise_scale_inv, columnwise_data, columnwise_scale_inv, fp8_dtype, quantizer, is_2D_scaled, data_format)
+
+.. autoapiclass:: transformer_engine.pytorch.NVFP4TensorStorage(rowwise_data, rowwise_scale_inv, columnwise_data, columnwise_scale_inv, amax_rowwise, amax_columnwise, fp4_dtype, quantizer)
+
+.. autoapiclass:: transformer_engine.pytorch.Float8Tensor(shape, dtype, data, fp8_scale_inv, fp8_dtype, requires_grad=False, data_transpose=None, quantizer=None)
+
+.. autoapiclass:: transformer_engine.pytorch.MXFP8Tensor(rowwise_data, rowwise_scale_inv, columnwise_data, columnwise_scale_inv, fp8_dtype, quantizer)
+
+.. autoapiclass:: transformer_engine.pytorch.Float8BlockwiseQTensor(rowwise_data, rowwise_scale_inv, columnwise_data, columnwise_scale_inv, fp8_dtype, quantizer, is_2D_scaled, data_format)
+
+.. autoapiclass:: transformer_engine.pytorch.NVFP4Tensor(rowwise_data, rowwise_scale_inv, columnwise_data, columnwise_scale_inv, amax_rowwise, amax_columnwise, fp4_dtype, quantizer)
+
+Quantizers
+----------
+
+.. autoapiclass:: transformer_engine.pytorch.Quantizer(rowwise, columnwise)
+   :members: update_quantized, quantize
+
+.. autoapiclass:: transformer_engine.pytorch.Float8Quantizer(scale, amax, fp8_dtype, *, rowwise=True, columnwise=True)
+
+.. autoapiclass:: transformer_engine.pytorch.Float8CurrentScalingQuantizer(fp8_dtype, device, *, rowwise=True, columnwise=True, **kwargs)
+
+.. autoapiclass:: transformer_engine.pytorch.MXFP8Quantizer(fp8_dtype, *, rowwise=True, columnwise=True)
+
+.. autoapiclass:: transformer_engine.pytorch.Float8BlockQuantizer(fp8_dtype, *, rowwise, columnwise, **kwargs)
+
+.. autoapiclass:: transformer_engine.pytorch.NVFP4Quantizer(fp4_dtype, *, rowwise=True, columnwise=True, **kwargs)
+
+Tensor saving and restoring functions
+-------------------------------------
+
+.. autoapifunction:: transformer_engine.pytorch.prepare_for_saving
+
+.. autoapifunction:: transformer_engine.pytorch.restore_from_saved