[ROCm] update ci_expected_accuracy for dynamo benchmarks (#163256)

Some tests that were already failing changed status to skipped.  Some model entries were missing.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/163256
Approved by: https://github.com/malfet

Co-authored-by: Jeff Daily <jeff.daily@amd.com>

benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_torchbench_inference.csv

@@ -205,7 +205,7 @@ llama,pass,0
 
 
 
-llama_v2_7b_16h,model_fail_to_load,0
+llama_v2_7b_16h,pass_due_to_skip,0
 
 
 

benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_torchbench_inference.csv

@@ -178,7 +178,7 @@ llama,fail_to_run,0
 
 
 
-llama_v2_7b_16h,model_fail_to_load,0
+llama_v2_7b_16h,pass_due_to_skip,0
 
 
 

benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_torchbench_inference.csv

@@ -198,7 +198,7 @@ llama,pass,0
 
 
 
-llama_v2_7b_16h,model_fail_to_load,0
+llama_v2_7b_16h,pass_due_to_skip,0
 
 
 

benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_huggingface_training.csv

@@ -171,3 +171,23 @@ XLNetLMHeadModel,pass,5
 
 
 YituTechConvBert,pass,5
+
+
+
+meta-llama/Llama-3.2-1B,eager_failed_to_run,0
+
+
+
+google/gemma-2-2b,eager_failed_to_run,0
+
+
+
+google/gemma-3-4b-it,eager_failed_to_run,0
+
+
+
+openai/whisper-tiny,eager_failed_to_run,0
+
+
+
+Qwen/Qwen3-0.6B,eager_failed_to_run,0

benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_torchbench_inference.csv

@@ -198,7 +198,7 @@ llama,pass,0
 
 
 
-llama_v2_7b_16h,model_fail_to_load,0
+llama_v2_7b_16h,pass_due_to_skip,0