Merge branch 'main' into feat/autotuner_distribute_tuning_part_2

Author: hyukn

examples/disaggregated/slurm/benchmark/submit.py

@@ -294,14 +294,14 @@ def submit_job(config, log_dir, dry_run):
                 env_config['model_path'],
                 str(allocation["port"]),
                 benchmark_config['mode'],
-                benchmark_config['concurrency_list'],
+                f"'{benchmark_config['concurrency_list']}'",
                 str(slurm_config['numa_bind']).lower(),
                 log_dir,
                 str(profiling_config['nsys_on']).lower(),
-                profiling_config['gen_profile_range'] if server_type == "GEN"
-                else profiling_config['ctx_profile_range'],
+                f"'{profiling_config['gen_profile_range']}'" if server_type
+                == "GEN" else f"'{profiling_config['ctx_profile_range']}'",
                 gen_config_path if server_type == "GEN" else ctx_config_path,
-                f'"{cur_worker_env_var}"',
+                f"'{cur_worker_env_var}'",
                 f"&> {log_dir}/3_output_{server_type}_{server_id}.log &",
             ]
             start_server_cmds.append(" ".join(cmd))
@@ -343,21 +343,21 @@ def submit_job(config, log_dir, dry_run):
     if benchmark_config['use_nv_sa_benchmark']:
         benchmark_cmd = [
             f"bash {env_config['work_dir']}/run_benchmark_nv_sa.sh",
-            f"{env_config['model_path']} {isl} {osl} {benchmark_config['benchmark_ratio']} {benchmark_config['multi_round']} {gen_num} {benchmark_config['concurrency_list']} {benchmark_config['streaming']} {log_dir} {disagg_server_hostname} {disagg_server_port}",
+            f"'{env_config['model_path']}' {isl} {osl} {benchmark_config['benchmark_ratio']} {benchmark_config['multi_round']} {gen_num} '{benchmark_config['concurrency_list']}' {benchmark_config['streaming']} '{log_dir}' {disagg_server_hostname} {disagg_server_port}",
             f"&> {log_dir}/6_bench.log"
         ]
         client_cmds.append(" ".join(client_slurm_prefix + benchmark_cmd))
     else:
         benchmark_cmd = [
             f"bash {env_config['work_dir']}/run_benchmark.sh",
-            f"{env_config['model_path']} {benchmark_config['dataset_file']} {benchmark_config['multi_round']} {gen_num} {benchmark_config['concurrency_list']} {benchmark_config['streaming']} {log_dir} {disagg_server_hostname} {disagg_server_port}",
+            f"'{env_config['model_path']}' '{benchmark_config['dataset_file']}' {benchmark_config['multi_round']} {gen_num} '{benchmark_config['concurrency_list']}' {benchmark_config['streaming']} '{log_dir}' {disagg_server_hostname} {disagg_server_port}",
             f"&> {log_dir}/6_bench.log"
         ]
         client_cmds.append(" ".join(client_slurm_prefix + benchmark_cmd))
     if config['accuracy']['enable_accuracy_test']:
         accuracy_cmd = [
             f"bash {env_config['work_dir']}/accuracy_eval.sh",
-            f"{log_dir} {config['accuracy']['model']} {config['accuracy']['tasks']} {env_config['model_path']} {config['accuracy']['model_args_extra']} {log_dir}/accuracy_eval {disagg_server_hostname} {disagg_server_port}",
+            f"'{log_dir}' '{config['accuracy']['model']}' '{config['accuracy']['tasks']}' '{env_config['model_path']}' '{config['accuracy']['model_args_extra']}' '{log_dir}/accuracy_eval' {disagg_server_hostname} {disagg_server_port}",
             f"&> {log_dir}/7_accuracy_eval.log"
         ]
         client_cmds.append(" ".join(client_slurm_prefix + accuracy_cmd))

requirements.txt

@@ -69,7 +69,7 @@ triton==3.5.0
 tiktoken
 blobfile
 openai-harmony==0.0.4
-nvidia-cutlass-dsl==4.3.1; python_version >= "3.10"
+nvidia-cutlass-dsl==4.3.4; python_version >= "3.10"
 plotly
 numexpr<2.14.0 # WAR for attempted use of nonexistent numpy.typing
 partial_json_parser

security_scanning/examples/models/contrib/grok/poetry.lock

@@ -1,4 +1,4 @@
 {
-  "commit_hash": "066b6539407405bfd33cde628c00a67b88d4d270",
-  "timestamp": "2025-12-22T02:39:59Z"
+  "commit_hash": "1e82ff7a0c5a0b525eb22d57e48efb21fc79087e",
+  "timestamp": "2025-12-23T02:42:08Z"
 }

security_scanning/examples/models/core/qwen/poetry.lock

@@ -72,7 +72,7 @@ triton = "3.5.0"
 tiktoken = "^0.12.0"
 blobfile = "^3.1.0"
 openai-harmony = "0.0.4"
-nvidia-cutlass-dsl = "4.3.1"
+nvidia-cutlass-dsl = "4.3.4"
 plotly = "^6.5.0"
 numexpr = "<2.14.0"
 partial-json-parser = "^0.2.1.1.post7"

security_scanning/examples/models/core/qwenvl/poetry.lock

@@ -81,6 +81,7 @@ transforms:
     sharding_source: ['manual', 'factory', 'heuristic']
     support_partial_config: true
     sharding_dims: ['tp', 'ep', 'bmm']
+    shard_all_unprocessed: true
     allreduce_strategy: 'NCCL'
     dist_backend: auto
     requires_shape_prop: true