[Misc] Add compressed-tensors NVFP4A16 emulation support (#17914)

Signed-off-by: Dipika Sikka <dipikasikka1@gmail.com>
Signed-off-by: Dipika <dipikasikka1@gmail.com>

tests/quantization/test_compressed_tensors.py

@@ -13,9 +13,9 @@ from compressed_tensors.quantization import QuantizationType
 from tests.models.utils import check_logprobs_close
 from vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors import (  # noqa: E501
     CompressedTensors24, CompressedTensorsLinearMethod,
-    CompressedTensorsW4A16Sparse24, CompressedTensorsW8A8Fp8,
-    CompressedTensorsW8A8Int8, CompressedTensorsW8A16Fp8,
-    CompressedTensorsWNA16)
+    CompressedTensorsW4A16Fp4, CompressedTensorsW4A16Sparse24,
+    CompressedTensorsW8A8Fp8, CompressedTensorsW8A8Int8,
+    CompressedTensorsW8A16Fp8, CompressedTensorsWNA16)
 from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
     sparse_cutlass_supported)
 from vllm.platforms import current_platform
@@ -648,3 +648,23 @@ def test_compressed_tensors_2of4_sparse_compressed(vllm_runner, args_2of4):
         output = llm.generate_greedy("Hello my name is", max_tokens=20)
         print(output)
         assert output
+
+
+def test_compressed_tensors_nvfp4a16(vllm_runner):
+    # run weight only example
+    model = "nm-testing/TinyLlama-1.1B-Chat-v1.0-FP4"
+    with vllm_runner(model, enforce_eager=True) as llm:
+
+        def check_model(model):
+            layer = model.model.layers[0]
+
+            qkv_proj = layer.self_attn.qkv_proj
+            assert isinstance(qkv_proj.quant_method,
+                              CompressedTensorsLinearMethod)
+            assert isinstance(qkv_proj.scheme, CompressedTensorsW4A16Fp4)
+            assert qkv_proj.scheme.group_size == 16
+
+        llm.apply_model(check_model)
+        output = llm.generate_greedy("Hello my name is", max_tokens=20)
+        print(output)
+        assert output