Add 2.3 RC for CUDA and CPU. (#106)

Author: cpuhrsch

.github/workflows/regression_test.yml

@@ -18,13 +18,16 @@ jobs:
             torch-spec: 'torch==2.2.2'
           - name: CUDA 2.3 RC
             runs-on: 4-core-ubuntu-gpu-t4
-            torch-spec: '--pre torch --index-url https://download.pytorch.org/whl/nightly/cu121'
+            torch-spec: 'torch==2.3.0 --index-url https://download.pytorch.org/whl/test/cu121'
           - name: CUDA Nightly
             runs-on: 4-core-ubuntu-gpu-t4
             torch-spec: '--pre torch --index-url https://download.pytorch.org/whl/nightly/cu121'
-          - name: CPU
+          - name: CPU 2.2.2
+            runs-on: 32-core-ubuntu
+            torch-spec: 'torch==2.2.2 --index-url https://download.pytorch.org/whl/cpu'
+          - name: CPU 2.3 RC
             runs-on: 32-core-ubuntu
-            torch-spec: 'torch --index-url https://download.pytorch.org/whl/cpu'
+            torch-spec: 'torch==2.3.0 --index-url https://download.pytorch.org/whl/test/cpu'
           - name: Nightly CPU
             runs-on: 32-core-ubuntu
             torch-spec: '--pre torch --index-url https://download.pytorch.org/whl/nightly/cpu'

test/quantization/test_quant_api.py

@@ -26,6 +26,7 @@
 )
 from torchao.quantization.utils import (
     TORCH_VERSION_AFTER_2_3,
+    TORCH_VERSION_AFTER_2_4,
 )
 from pathlib import Path
 from sentencepiece import SentencePieceProcessor
@@ -136,7 +137,7 @@ def test_dynamic_quant_gpu_unified_api_eager_mode_impl(self):
         compiled = m(*example_inputs)
         torch.testing.assert_close(quantized, compiled, atol=0, rtol=0)
 
-    @unittest.skipIf(not TORCH_VERSION_AFTER_2_3, "skipping when torch verion is 2.3 or lower")
+    @unittest.skipIf(not TORCH_VERSION_AFTER_2_4, "skipping when torch verion is 2.4 or lower")
     def test_8da4w_quantizer(self):
         from torchao.quantization.quant_api import Int8DynActInt4WeightQuantizer
         from torchao.quantization.GPTQ import Int8DynActInt4WeightLinear