manifest zai-org/GLM-5.3-Flash @ B200 (85fecedca608f809)

zai-org__GLM-5.3-Flash/B200/tp4-seq262144-lora128x1/85fecedca608f809/manifest.json

@@ -0,0 +1,145 @@
+{
+  "model": "zai-org/GLM-5.3-Flash",
+  "gpu_type": "B200",
+  "tensor_parallel_size": 4,
+  "max_seq_length": 262144,
+  "enable_lora": true,
+  "max_lora_rank": 128,
+  "max_loras": 1,
+  "lora_target_modules": [
+    "q_a_proj",
+    "kv_a_proj_with_mqa",
+    "q_b_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "b_proj",
+    "f_a_proj",
+    "f_b_proj",
+    "g_a_proj",
+    "g_b_proj",
+    "o_proj",
+    "gate_proj",
+    "up_proj",
+    "down_proj",
+    "lm_head"
+  ],
+  "lora_target_module_preset": null,
+  "weights_source": null,
+  "cudagraph_capture_sizes": [
+    1,
+    2,
+    4,
+    8,
+    16,
+    32,
+    64,
+    128,
+    192,
+    256,
+    384,
+    512,
+    640,
+    768,
+    896,
+    1000
+  ],
+  "use_mega_aot_artifact": true,
+  "deep_gemm_warmup": "skip",
+  "enable_prefix_caching": true,
+  "moe_backend": null,
+  "enable_expert_parallel": false,
+  "vllm_version": "0.1.dev20051+g487ecf187",
+  "torch_version": "2.13.0+cu130",
+  "torch": "2.13.0+cu130",
+  "torch_cuda": "13.0",
+  "vllm": "0.1.dev20051+g487ecf187",
+  "image_tag": "baseten/baseten-weight-sync-inference:jerry-glm53-flash-h200-sampler-b4c1939-glm53-flash",
+  "caller": "github-actions:github-actions[bot]",
+  "model_revision": "eb9eb208eb0d988989d07a6a12d0fdeb5f52574a",
+  "build_id": "34430498285-1",
+  "opaque_sampler_payload": {
+    "tensor_parallel_size": 4,
+    "max_seq_length": 262144,
+    "enable_lora": true,
+    "max_lora_rank": 128,
+    "max_loras": 1,
+    "lora_target_modules": [
+      "q_a_proj",
+      "kv_a_proj_with_mqa",
+      "q_b_proj",
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "b_proj",
+      "f_a_proj",
+      "f_b_proj",
+      "g_a_proj",
+      "g_b_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj",
+      "lm_head"
+    ],
+    "lora_target_module_preset": null,
+    "cudagraph_capture_sizes": [
+      1,
+      2,
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      192,
+      256,
+      384,
+      512,
+      640,
+      768,
+      896,
+      1000
+    ],
+    "use_mega_aot_artifact": true,
+    "deep_gemm_warmup": "skip",
+    "enable_prefix_caching": true,
+    "load_format": "auto",
+    "disable_custom_all_reduce": false,
+    "gpu_memory_utilization": 0.9,
+    "max_num_seqs": null,
+    "moe_backend": null,
+    "enable_expert_parallel": false
+  },
+  "build_profile": {
+    "cudagraph_capture_sizes": [
+      1,
+      2,
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      192,
+      256,
+      384,
+      512,
+      640,
+      768,
+      896,
+      1000
+    ],
+    "use_mega_aot_artifact": true,
+    "deep_gemm_warmup": "skip"
+  },
+  "ready_in_seconds_no_cache": 1776.5,
+  "kv_cache_max_tokens": 12121135,
+  "kv_cache_max_concurrency": 46.238461538461536,
+  "kv_cache_gpu_memory_utilization": 0.9,
+  "cache_size_uncompressed_bytes": 4188160,
+  "cache_size_compressed_bytes": 198983,
+  "compression": "zstd -9 -T0 (multithreaded)",
+  "compress_time_seconds": 0.05,
+  "upload_time_seconds": 3.62
+}