add base docker image with pytorch 2.7.0 and variant for cuda 12.8 (axolotl-ai-cloud#2551)

* add base docker image with pytorch 2.7.0 and variant for cuda 12.8

* my bash is terrible

Author: winglian

.github/workflows/base.yml

@@ -46,6 +46,18 @@ jobs:
             python_version: "3.11"
             pytorch: 2.6.0
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "126"
+            cuda_version: 12.6.3
+            cudnn_version: ""
+            python_version: "3.11"
+            pytorch: 2.7.0
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "128"
+            cuda_version: 12.6.3
+            cudnn_version: ""
+            python_version: "3.11"
+            pytorch: 2.7.0
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
           - cuda: "128"
             cuda_version: 12.8.1
             cudnn_version: ""

docker/Dockerfile-base

@@ -37,3 +37,7 @@ RUN git lfs install --skip-repo && \
     pip3 install awscli && \
     # The base image ships with `pydantic==1.8.2` which is not working
     pip3 install -U --no-cache-dir pydantic==1.10.10
+
+RUN if [ "$PYTORCH_VERSION" = "2.7.0" ] ; then \
+        pip3 install flash-attn==2.7.4.post1; \
+    fi

requirements.txt

@@ -19,6 +19,7 @@ datasets==3.5.0
 deepspeed>=0.15.4
 trl==0.16.1
 hf_xet==1.0.0
+hqq==0.2.5
 
 optimum==1.16.2
 hf_transfer