Remove KV cache quantization, increase batch/ubatch on frank-default-large

- Remove CACHE_TYPE_K/CACHE_TYPE_V (fp16 default on GPU)
- Increase BATCH_SIZE from 2048 to 8192
- Increase UBATCH_SIZE from 512 to 2048
- Remove --cache-type-k and --cache-type-v from llama-server flags

scripts/serve-llama-frank-large.sh

@@ -33,10 +33,8 @@ fi
 : "${SPEC_NGRAM_MOD_N_MATCH:=24}"
 : "${DEVICE:=ROCm0}"
 : "${N_GPU_LAYERS:=all}"
-: "${CACHE_TYPE_K:=q8_0}"
-: "${CACHE_TYPE_V:=q8_0}"
-: "${BATCH_SIZE:=2048}"
-: "${UBATCH_SIZE:=512}"
+: "${BATCH_SIZE:=8192}"
+: "${UBATCH_SIZE:=2048}"
 : "${TEMP:=1.0}"
 : "${TOP_K:=20}"
 : "${TOP_P:=0.95}"
@@ -66,8 +64,6 @@ exec "$LLAMA_SERVER" \
     --device "$DEVICE" \
     --gpu-layers "$N_GPU_LAYERS" \
     --fit off \
-    --cache-type-k "$CACHE_TYPE_K" \
-    --cache-type-v "$CACHE_TYPE_V" \
     --ctx-size "$CTX_SIZE" \
     --parallel "$PARALLEL" \
     --batch-size "$BATCH_SIZE" \