phase 11: GPU-resident KV cache

- GpuKVCache: pre-allocated GPU buffers, D2D copy append at offset
- Per-head strided layout [num_kv_heads, max_seq_len, head_dim]
- Fixed critical bug: seq_len must advance AFTER all layers write
  (not inside the loop per-layer)
- GpuBuffer::copy_from_device_at for offset-based D2D copy
- Tensor::from_storage constructor for wrapping raw GPU buffers
- Exported Storage and Dims from xserv-tensor

Correctness: GPU KV cache vs CPU KV cache = 50/50 bit-identical
Performance: ~neutral (KV cache was never the main bottleneck —
reshape/merge/transpose CPU round-trips dominate for Qwen3-8B)

TTFT: 122ms, TBT: 142ms, 7.0 tok/s (marginal change from 7.3)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

crates/xserv-model/src/bin/bench-qwen3.rs

@@ -1,7 +1,7 @@
 use std::path::PathBuf;
 use std::time::Instant;
 use xserv_model::qwen3::sample_greedy;
-use xserv_model::{loader, KVCache, ModelConfig, Qwen3};
+use xserv_model::{loader, GpuKVCache, ModelConfig, Qwen3};
 use xserv_tensor::{DType, Device};
 use xserv_tokenizer::Tokenizer;
 
@@ -31,11 +31,8 @@ fn main() {
     // Warmup
     {
         let ids = tokenizer.encode("warmup");
-        let mut cache = KVCache::new(
-            config.num_layers(), config.num_kv_heads(), config.head_dim(),
-            DType::BF16, Device::Cuda(0),
-        );
-        let _ = model.forward_with_cache(&ids, &mut cache);
+        let mut cache = GpuKVCache::new(&config, 256, DType::BF16);
+        let _ = model.forward_gpu_cache(&ids, &mut cache);
     }
     eprintln!("Warmup done. Running benchmark...");
 
@@ -97,14 +94,11 @@ fn main() {
         let input_ids = tokenizer.encode(prompt);
         let input_len = input_ids.len();
 
-        let mut cache = KVCache::new(
-            config.num_layers(), config.num_kv_heads(), config.head_dim(),
-            DType::BF16, Device::Cuda(0),
-        );
+        let mut cache = GpuKVCache::new(&config, 256, DType::BF16);
 
         // Prefill
         let t0 = Instant::now();
-        let logits = model.forward_with_cache(&input_ids, &mut cache);
+        let logits = model.forward_gpu_cache(&input_ids, &mut cache);
         let first_token = sample_greedy(&logits);
         let ttft_us = t0.elapsed().as_micros();
 
@@ -115,7 +109,7 @@ fn main() {
         for _ in 1..gen_tokens {
             let last = *generated.last().unwrap();
             let t_start = Instant::now();
-            let logits = model.forward_with_cache(&[last], &mut cache);
+            let logits = model.forward_gpu_cache(&[last], &mut cache);
             let next = sample_greedy(&logits);
             token_times.push(t_start.elapsed().as_micros());
             generated.push(next);
@@ -148,12 +142,14 @@ fn main() {
         print!("\"tpot_us\": {tpot_us}}}");
         if i < prompts.len() - 1 { println!(","); } else { println!(); }
 
+        let display_text = generated_text.replace('\n', " ");
+        let truncated: String = display_text.chars().take(60).collect();
         eprintln!(
             "[{}/{}] input={input_len}tok gen={num_generated}tok ttft={:.1}ms tbt={:.1}ms | {}",
             i + 1, prompts.len(),
             ttft_us as f64 / 1000.0,
             tbt_us as f64 / 1000.0,
-            &generated_text.replace('\n', " ")[..generated_text.len().min(60)]
+            truncated
         );
     }
     println!("]");