Upload scripts/eval_gpu3.sh with huggingface_hub
Browse files- scripts/eval_gpu3.sh +15 -0
scripts/eval_gpu3.sh
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env bash
|
| 2 |
+
# Fill the idle GPU 3: extend both curves at 65,512.
|
| 3 |
+
set -uo pipefail; cd "$(dirname "$0")"
|
| 4 |
+
export LD_LIBRARY_PATH="$(cd ../.. && pwd)/Tensor-Implementations/lib:$(cd ../.. && pwd)/Profiler/lib:${LD_LIBRARY_PATH:-}"
|
| 5 |
+
export PATH=/usr/local/cuda/bin:$PATH; export PY=/root/Adhi/BluTrain/envv/bin/python
|
| 6 |
+
L=eval_gpu3_logs; mkdir -p "$L"
|
| 7 |
+
A680="CP_N_EMBD=1536 CP_N_LAYER=24 CP_N_HEAD=12 CP_N_KVHEAD=4 CP_FFN=4096 CP_WEIGHT_TYING=1"
|
| 8 |
+
say(){ echo "[$(date -u +%H:%M:%S)] $*" | tee -a "$L/master.log" >&2; }
|
| 9 |
+
say "=== GPU3: t64 + c500 + d100 @65,512 ==="
|
| 10 |
+
A1B="$A680" HAY="data/mix128_b6_60/val_shard_00000.bin" GEN_DATA_ROOT="data/mix128_b6_60" \
|
| 11 |
+
CKPT_EXTRA="t64=t64.ckpt c500=t128c_s500.ckpt d100=t128d_s100.ckpt" CKPTS="t64 c500 d100" \
|
| 12 |
+
TASKS="niah_single" N=100 T=131072 YARN=32 YARN_ORIG=4096 PROMPT_LEN=65512 \
|
| 13 |
+
GPUS="3" NP=1 JOBS_PER_GPU=2 OUT="eg3_65512" ./run_longeval.sh > "$L/len65512.log" 2>&1
|
| 14 |
+
say " 65512: $(head -1 eg3_65512/summary.csv 2>/dev/null|tr -d '\r') -> $(grep -h '^niah_single,' eg3_65512/summary.csv 2>/dev/null|tr -d '\r')"
|
| 15 |
+
say "EVAL_GPU3_DONE"
|