From 1e4e406d775410efeafcdc32d04a1a75afa65f0e Mon Sep 17 00:00:00 2001
From: David Corvoysier <david@huggingface.co>
Date: Fri, 21 Feb 2025 12:46:57 +0000
Subject: [PATCH] test(neuron): avoid using image sha when exporting models

We now manually evaluate the apparent hash of the neuron backend by
combining the hash of the neuron backend directory and Dockerfile.
This new hash is used to identify exported neuron models instead of the
image sha.
This has two benefits:
- it changes less frequently (only hwen the neuron backend changes),
  which means less neuron models being pushed to the hub,
- it can be evaluated locally, meaning that running the tests once
  locally will export the models before the CI uses them.
---
 integration-tests/fixtures/neuron/model.py | 24 ++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)

diff --git a/integration-tests/fixtures/neuron/model.py b/integration-tests/fixtures/neuron/model.py
index eb1091218..10f1c84eb 100644
--- a/integration-tests/fixtures/neuron/model.py
+++ b/integration-tests/fixtures/neuron/model.py
@@ -6,6 +6,7 @@ from tempfile import TemporaryDirectory
 import huggingface_hub
 import pytest
 import docker
+import hashlib
 import os
 import tempfile
 
@@ -50,9 +51,28 @@ MODEL_CONFIGURATIONS = {
 }
 
 
+def get_neuron_backend_hash():
+    import subprocess
+    res = subprocess.run(["git", "rev-parse", "--show-toplevel"],
+                         capture_output=True,
+                         text=True)
+    root_dir = res.stdout.split('\n')[0]
+    def get_sha(path):
+        res = subprocess.run(["git", "ls-tree", "HEAD", f"{root_dir}/{path}"],
+                             capture_output=True,
+                             text=True)
+        # Output of the command is in the form '040000 tree|blob <SHA>\t<path>\n'
+        sha = res.stdout.split('\t')[0].split(' ')[-1]
+        return sha.encode()
+    # We hash both the neuron backends directory and Dockerfile and create a smaller hash out of that
+    m = hashlib.sha256()
+    m.update(get_sha('backends/neuron'))
+    m.update(get_sha('Dockerfile.neuron'))
+    return m.hexdigest()[:10]
+
+
 def get_neuron_model_name(config_name: str):
-    version = get_tgi_docker_image().split(":")[-1]
-    return f"neuron-tgi-testing-{config_name}-{version}"
+    return f"neuron-tgi-testing-{config_name}-{get_neuron_backend_hash()}"
 
 
 def get_tgi_docker_image():