fix llama buck build (#13169)

Summary:
Pull Request resolved: #13169

Some recent changes break the llama buck build

Reviewed By: rohansjoshi

Differential Revision: D79753385

Author: cccclai

examples/qualcomm/oss_scripts/llama/TARGETS

@@ -15,10 +15,30 @@ python_library(
     ],
 )
 
+python_library(
+    name = "decoder_utils",
+    srcs = [
+        "decoder_utils.py",
+    ],
+    deps = [
+        "//caffe2:torch",
+        "//executorch/examples/models/llama:eval_library",
+    ],
+)
+
+python_library(
+    name = "decoder_constants",
+    srcs = [
+        "decoder_constants.py",
+    ],
+)
+
 python_library(
     name = "llama_lib",
     srcs = ["llama.py"],
     deps = [
+        ":decoder_constants",
+        ":decoder_utils",
         "//executorch/examples/models/llama:source_transformation",
         "//caffe2:torch",
         "//executorch/backends/qualcomm/partition:partition",

examples/qualcomm/oss_scripts/llama/decoder_utils.py

@@ -52,6 +52,7 @@ def __init__(
         use_i64_token: bool,
     ):
         # n seq len = n-1 cache len, so we len(inps) = n-1 during _model_call
+        # #pyre-ignore
         super().__init__(
             model=model, tokenizer=tokenizer, max_seq_length=max_seq_length - 1
         )