Skip invalid gen config (#618)

* chore: bump dev version

* feat(decoder): rebuild invalid generation config

* docs(tgi): use more recent model

docs/source/guides/neuronx_tgi.mdx

@@ -87,14 +87,15 @@ possible to export it dynamically, pending some conditions:
 The snippet below shows how you can deploy a service from a hub standard model:
 
 ```
+export HF_TOKEN=<YOUR_TOKEN>
 docker run -p 8080:80 \
        -v $(pwd)/data:/data \
        --privileged \
        -e HF_TOKEN=${HF_TOKEN} \
        -e HF_AUTO_CAST_TYPE="fp16" \
        -e HF_NUM_CORES=2 \
        ghcr.io/huggingface/neuronx-tgi:latest \
-       --model-id NousResearch/Llama-2-7b-chat-hf \
+       --model-id meta-llama/Meta-Llama-3-8B \
        --max-batch-size 1 \
        --max-input-length 3164 \
        --max-total-tokens 4096

optimum/neuron/modeling_decoder.py

@@ -20,6 +20,7 @@
 import os
 import re
 import shutil
+import warnings
 from pathlib import Path
 from tempfile import TemporaryDirectory
 from typing import TYPE_CHECKING, Optional, Tuple, Union
@@ -250,6 +251,13 @@ def _create_checkpoint(
             **kwargs,
         )
 
+        if model.generation_config is not None:
+            with warnings.catch_warnings(record=True) as caught_warnings:
+                model.generation_config.validate()
+            if len(caught_warnings) > 0:
+                logger.warning("Invalid generation config: recreating it from model config.")
+                model.generation_config = GenerationConfig.from_model_config(model.config)
+
         # Save the model checkpoint in a temporary directory
         checkpoint_dir = TemporaryDirectory()
         os.chmod(checkpoint_dir.name, 0o775)

optimum/neuron/version.py

@@ -12,6 +12,6 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 
-__version__ = "0.0.23.dev0"
+__version__ = "0.0.24.dev0"
 
 __sdk_version__ = "2.18.0"