[cherry-pick][secure-mode] Update options allowlist for 0.29.0 (#2310) (

#2311)
deepjavalibrary · Aug 14, 2024 · da971d6 · da971d6
1 parent 316032a
commit da971d6
Showing 1 changed file with 37 additions and 1 deletion.
diff --git a/plugins/secure-mode/src/main/java/ai/djl/serving/plugins/securemode/SecureModeAllowList.java b/plugins/secure-mode/src/main/java/ai/djl/serving/plugins/securemode/SecureModeAllowList.java
@@ -34,6 +34,7 @@ interface SecureModeAllowList {
                     "option.model_id",
                     "option.batch_size",
                     "option.tensor_parallel_degree",
+                    "option.pipeline_parallel_degree",
                     "option.rolling_batch",
                     "option.dtype",
                     "option.trust_remote_code",
@@ -68,5 +69,40 @@ interface SecureModeAllowList {
                     "option.max_loras",
                     "option.max_lora_rank",
                     "option.lora_extra_vocab_size",
-                    "option.max_cpu_loras");
+                    "option.max_cpu_loras",
+                    "option.max_input_len",
+                    "option.max_output_len",
+                    "option.max_num_tokens",
+                    "option.use_custom_all_reduce",
+                    "option.tokens_per_block",
+                    "option.batch_scheduler_policy",
+                    "option.kv_cache_free_gpu_mem_fraction",
+                    "option.max_num_sequences",
+                    "option.enable_trt_overlap",
+                    "option.enable_kv_cache_reuse",
+                    "option.baichuan_model_version",
+                    "option.chatglm_model_version",
+                    "option.gpt_model_version",
+                    "option.multi_block_mode",
+                    "option.use_fused_mlp",
+                    "option.rotary_base",
+                    "option.rotary_dim",
+                    "option.rotary_scaling_type",
+                    "option.rotary_scaling_factor",
+                    "option.logits_dtype",
+                    "option.trtllm_checkpoint_path",
+                    "option.load_by_shard",
+                    "option.smoothquant_alpha",
+                    "option.smoothquant_per_token",
+                    "option.smoothquant_per_channel",
+                    "option.multi_query_mode",
+                    "option.awq_format",
+                    "option.awq_calib_size",
+                    "option.q_format",
+                    "option.calib_size",
+                    "option.calib_batch_size",
+                    "option.use_fp8_context_fmha",
+                    "option.enable_chunked_prefill",
+                    "option.cpu_offload_gb_per_gpu",
+                    "option.enable_prefix_caching");
 }