Add B300 GPU preset and GLM-5-NVFP4 on 8xB300 config

Add NVIDIA B300 (Blackwell Ultra) to hardware presets: same GB202 die as
B200 (2.25 PFLOPS BF16 dense) but with HBM3e 12-Hi stacks (288 GB,
12 TB/s — 50% more capacity and bandwidth than B200).

Add nvidia/GLM-5-NVFP4 HuggingFace config.json and a matching simulation
config for 8xB300: FP4 weights (~372 GB) leave ~1.9 TB for KV cache,
yielding 82k blocks per instance (3.8x more than the BF16-on-B200 setup).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

configs/glm5-nvfp4-8xb300.yaml

@@ -0,0 +1,31 @@
+# GLM-5-NVFP4 (nvidia/GLM-5-NVFP4) on 8 x B300 (Blackwell Ultra, 288GB each).
+# Architecture auto-loaded from HuggingFace config.json.
+#
+# FP4 weights: ~744B params * 0.5 bytes = ~372 GB across 8 GPUs.
+# Total HBM: 8 * 288 GB = 2304 GB.  KV budget: ~1900 GB after weights.
+
+model:
+  config_json: ../models/GLM-5-NVFP4/config.json
+  name: glm-5-nvfp4
+  dtype_bytes: 1               # FP8 KV cache
+  block_size_tokens: 512
+
+hardware:
+  type: 8xb300
+  hbm_bytes: 1900.0e9          # KV budget after FP4 weights (~372 GB)
+
+cluster:
+  num_instances: 32
+  meta_store:
+    ttl_seconds: 120.0
+  router:
+    mode: prefix_affinity
+    prefix_k: 8
+    load_alpha: 1.0
+
+sim:
+  trace_path: bailian-traces/glm_coder_blksz_512_040915-040917.jsonl
+  max_requests: null
+  output_dir: runs/glm5_nvfp4_8xb300
+  sample_interval_s: 1.0
+  seed: 42

models/GLM-5-NVFP4/config.json

@@ -0,0 +1,59 @@
+{
+  "architectures": [
+    "GlmMoeDsaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "ep_size": 1,
+  "first_k_dense_replace": 3,
+  "hidden_act": "silu",
+  "head_dim": 64,
+  "hidden_size": 6144,
+  "index_head_dim": 128,
+  "index_n_heads": 32,
+  "index_topk": 2048,
+  "indexer_rope_interleave": true,
+  "initializer_range": 0.02,
+  "intermediate_size": 12288,
+  "kv_lora_rank": 512,
+  "max_position_embeddings": 202752,
+  "moe_intermediate_size": 2048,
+  "moe_layer_freq": 1,
+  "model_type": "glm_moe_dsa",
+  "n_group": 1,
+  "n_routed_experts": 256,
+  "n_shared_experts": 1,
+  "norm_topk_prob": true,
+  "num_attention_heads": 64,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 78,
+  "num_key_value_heads": 64,
+  "num_nextn_predict_layers": 1,
+  "pad_token_id": 154820,
+  "pretraining_tp": 1,
+  "q_lora_rank": 2048,
+  "qk_head_dim": 256,
+  "qk_nope_head_dim": 192,
+  "qk_rope_head_dim": 64,
+  "rms_norm_eps": 1e-05,
+  "rope_interleave": true,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "routed_scaling_factor": 2.5,
+  "scoring_func": "sigmoid",
+  "tie_word_embeddings": false,
+  "topk_group": 1,
+  "topk_method": "noaux_tc",
+  "transformers_version": "5.0.2.dev0",
+  "use_cache": true,
+  "v_head_dim": 256,
+  "vocab_size": 154880
+}

src/hardware_presets.rs

@@ -20,6 +20,7 @@ pub const AVAILABLE: &[&str] = &[
     "a100-80gb",
     "a100-40gb",
     "b200",
+    "b300",
     "2xh100",
     "4xh100",
     "8xh100",
@@ -32,6 +33,9 @@ pub const AVAILABLE: &[&str] = &[
     "2xb200",
     "4xb200",
     "8xb200",
+    "2xb300",
+    "4xb300",
+    "8xb300",
 ];
 
 /// Resolve a hardware preset by name.
@@ -48,6 +52,7 @@ pub fn resolve(name: &str) -> Option<HardwareConfig> {
         "a10080gb" | "a100" => Some(make_config(count, &A100_80GB)),
         "a10040gb" => Some(make_config(count, &A100_40GB)),
         "b200" => Some(make_config(count, &B200)),
+        "b300" => Some(make_config(count, &B300)),
         _ => None,
     }
 }
@@ -121,6 +126,13 @@ const B200: GpuBase = GpuBase {
     pcie_gen: 6,
 };
 
+const B300: GpuBase = GpuBase {
+    flops: 2.25e15,   // 2250 TFLOPS BF16 dense (same GB202 die as B200)
+    mem_bw: 12.0e12,  // 12 TB/s HBM3e 12-Hi (50% more than B200 8-Hi)
+    hbm: 288.0e9,     // 288 GB HBM3e 12-Hi
+    pcie_gen: 6,
+};
+
 /// Build a [`HardwareConfig`] from a base GPU spec × TP count.
 ///
 /// Compute, HBM bandwidth, and HBM capacity scale linearly with `n`.