upgrade cog and use new format to specify target concurrency

replicate · May 17, 2024 · 3e555aa · 3e555aa
1 parent 2f893d2
commit 3e555aa
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 3 deletions.
diff --git a/cog.yaml b/cog.yaml
@@ -2,4 +2,6 @@ build:
    gpu: true
    python_version: "3.10"
 predict: "predict.py:Predictor"
-concurrency: 64
+concurrency:
+   max: 64
+   default_target: 16
diff --git a/default.nix b/default.nix
@@ -12,7 +12,7 @@ in
   imports = [ ./interface.nix ];
   cog.build = {
     python_version = "3.10";
-    cog_version = "0.10.0-alpha5";
+    cog_version = "0.10.0-alpha7";
     cuda = "12.1"; # todo: 12.2
     gpu = true;
     # inspiration: echo tensorrt_llm==0.8.0 | uv pip compile - --extra-index-url https://pypi.nvidia.com -p 3.10 --prerelease=allow --annotation-style=line
@@ -45,7 +45,7 @@ in
     overrides = [
       "tokenizers==0.19.0"
       "transformers==4.40.0"
-      "cog @ https://r2.drysys.workers.dev/tmp/cog-0.10.0a6-py3-none-any.whl"
+      "cog==0.10.0a7"
     ];
   };
   cognix.includeNix = true;