Update TGI copy (#438)

modal-labs · Sep 5, 2023 · 2b1cbe6 · 2b1cbe6
1 parent 45dbdc1
commit 2b1cbe6
Showing 1 changed file with 2 additions and 3 deletions.
diff --git a/06_gpu_and_ml/text_generation_inference.py b/06_gpu_and_ml/text_generation_inference.py
@@ -35,9 +35,8 @@
 # the largest 70B model, the 135GB model can be loaded in as little as 70 seconds.
 #
 # ### Download the weights
-# Since TGI uses `snapshot_download` under the hood, running this function for our image build
-# will place the weights into the cache directly. There are no progress bars as this uses
-# the high-throughput `hf-transfer` library, but expect ~700MB/s for this step.
+# We can use the included utilities to download the model weights (and convert to safetensors, if necessary)
+# as part of the image build.
 #