Bump version to 0.1.5.1

# What does this PR do?
- using `eval` is a security risk

[//]: # (If resolving an issue, uncomment and update the line below)
[//]: # (Closes #[issue-number])

## Test Plan

- see https://github.com/meta-llama/llama-stack/pull/1327

cc @SLR722 we will need to update the corresponding dataset via

```python
def update_to_json_str():
        
dataset = datasets.load_dataset(...)
processed_dataset = dataset[split].map(
        lambda x: {
                "column": json.dumps(eval(x["column"]))
       }
)
processed_dataset.push_to_hub(...)
```
[//]: # (## Documentation)

llama_stack/providers/inline/eval/meta_reference/eval.py

@@ -3,6 +3,7 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
+import json
 from typing import Any, Dict, List, Optional
 
 from tqdm import tqdm
@@ -117,7 +118,7 @@ class MetaReferenceEvalImpl(
         generations = []
         for i, x in tqdm(enumerate(input_rows)):
             assert ColumnName.chat_completion_input.value in x, "Invalid input row"
-            input_messages = eval(str(x[ColumnName.chat_completion_input.value]))
+            input_messages = json.loads(x[ColumnName.chat_completion_input.value])
             input_messages = [UserMessage(**x) for x in input_messages]
 
             # NOTE: only single-turn agent generation is supported. Create a new session for each input row
@@ -159,7 +160,7 @@ class MetaReferenceEvalImpl(
         generations = []
         for x in tqdm(input_rows):
             if ColumnName.completion_input.value in x:
-                input_content = eval(str(x[ColumnName.completion_input.value]))
+                input_content = json.loads(x[ColumnName.completion_input.value])
                 response = await self.inference_api.completion(
                     model=candidate.model,
                     content=input_content,
@@ -167,9 +168,8 @@ class MetaReferenceEvalImpl(
                 )
                 generations.append({ColumnName.generated_answer.value: response.completion_message.content})
             elif ColumnName.chat_completion_input.value in x:
-                chat_completion_input_str = str(x[ColumnName.chat_completion_input.value])
+                chat_completion_input_json = json.loads(x[ColumnName.chat_completion_input.value])
-                input_messages = eval(chat_completion_input_str)
+                input_messages = [UserMessage(**x) for x in chat_completion_input_json]
-                input_messages = [UserMessage(**x) for x in input_messages]
                 messages = []
                 if candidate.system_message:
                     messages.append(candidate.system_message)

llama_stack/providers/inline/post_training/torchtune/datasets/format_adapter.py

@@ -10,16 +10,19 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
+import json
 from typing import Any, Mapping
 
 from llama_stack.providers.utils.common.data_schema_validator import ColumnName
 
 
-def llama_stack_instruct_to_torchtune_instruct(sample: Mapping[str, Any]) -> Mapping[str, Any]:
+def llama_stack_instruct_to_torchtune_instruct(
+    sample: Mapping[str, Any],
+) -> Mapping[str, Any]:
     assert ColumnName.chat_completion_input.value in sample and ColumnName.expected_answer.value in sample, (
         "Invalid input row"
     )
-    input_messages = eval(str(sample[ColumnName.chat_completion_input.value]))
+    input_messages = json.loads(sample[ColumnName.chat_completion_input.value])
 
     assert len(input_messages) == 1, "llama stack intruct dataset format only supports 1 user message"
     input_message = input_messages[0]
@@ -37,7 +40,7 @@ def llama_stack_instruct_to_torchtune_instruct(sample: Mapping[str, Any]) -> Map
 def llama_stack_chat_to_torchtune_chat(sample: Mapping[str, Any]) -> Mapping[str, Any]:
     assert ColumnName.dialog.value in sample, "Invalid input row"
     role_map = {"user": "human", "assistant": "gpt"}
-    dialog = eval(str(sample[ColumnName.dialog.value]))
+    dialog = json.loads(sample[ColumnName.dialog.value])
 
     assert len(dialog) > 1, "dialog must have at least 2 messagse"
     roles = []

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "llama_stack"
-version = "0.1.4"
+version = "0.1.5.1"
 authors = [{ name = "Meta Llama", email = "llama-oss@meta.com" }]
 description = "Llama Stack"
 readme = "README.md"
@@ -26,8 +26,8 @@ dependencies = [
     "httpx",
     "huggingface-hub",
     "jsonschema",
-    "llama-models>=0.1.4",
+    "llama-models>=0.1.5rc3",
-    "llama-stack-client>=0.1.4",
+    "llama-stack-client>=0.1.5rc3",
     "prompt-toolkit",
     "python-dotenv",
     "pydantic>=2",