scripts/run_evals.sh

#!/usr/bin/env bash
# Orchestrate: for each quant model -> serve with vllm -> run wikitext PPL + gsm8k acc via lm_eval
set -uo pipefail

source "$HOME/nvitop-env/bin/activate"

PORT=8000
BASE="http://127.0.0.1:${PORT}"
LOGDIR="$HOME/eval_logs"
RESDIR="$HOME/results"
mkdir -p "$LOGDIR" "$RESDIR"

# model_dir | served_name | extra_env
MODELS=(
  "Qwen3-8B-MXFP8-AWQ-wikitext2|qwen3-mxfp8|"
  "Qwen3-8B-NVFP4-AWQ-wikitext2|qwen3-nvfp4|"
  "Qwen3-8B-FP8-AWQ-wikitext2|qwen3-fp8|VLLM_USE_DEEP_GEMM=0"
  "Qwen3-8B-MXFP4-AWQ-wikitext2|qwen3-mxfp4|VLLM_DISABLED_KERNELS=FlashInferMxFp4LinearKernel"
)

stop_server() {
  pkill -f "vllm serve" 2>/dev/null || true
  # wait for port to free
  for _ in $(seq 1 30); do
    if ! curl -s "${BASE}/health" >/dev/null 2>&1; then break; fi
    sleep 2
  done
  sleep 3
}

wait_ready() {
  local log="$1"
  for i in $(seq 1 180); do  # up to ~15 min
    if curl -s "${BASE}/health" >/dev/null 2>&1; then
      echo "[ready] server healthy after ${i} checks"
      return 0
    fi
    if ! pgrep -f "vllm serve" >/dev/null 2>&1; then
      echo "[error] vllm process died during startup. Tail of server log:"
      tail -30 "$log"
      return 1
    fi
    sleep 5
  done
  echo "[error] server not ready in time"; return 1
}

for entry in "${MODELS[@]}"; do
  IFS='|' read -r DIR NAME EXTRA <<< "$entry"
  MODEL_PATH="$HOME/models/$DIR"
  SRVLOG="$LOGDIR/${NAME}_server.log"
  WIKILOG="$LOGDIR/${NAME}_wikitext.log"
  GSMLOG="$LOGDIR/${NAME}_gsm8k.log"

  echo "############################################################"
  echo "# MODEL: $NAME   ($DIR)   extra_env='${EXTRA:-none}'"
  echo "############################################################"

  stop_server

  echo "[serve] starting vllm (log: $SRVLOG)"
  env ${EXTRA} vllm serve "$MODEL_PATH" \
      --served-model-name "$NAME" \
      --host 0.0.0.0 --port "$PORT" \
      --trust-remote-code \
      --max-model-len 4096 \
      --gpu-memory-utilization 0.90 \
      > "$SRVLOG" 2>&1 &

  if ! wait_ready "$SRVLOG"; then
    echo "[skip] $NAME failed to start; moving on"
    stop_server
    continue
  fi

  echo "[eval] === wikitext perplexity ($NAME) ==="
  lm_eval \
    --model local-completions \
    --model_args "model=${NAME},base_url=${BASE}/v1/completions,tokenizer=${MODEL_PATH},num_concurrent=32,max_retries=3,tokenized_requests=False" \
    --tasks wikitext \
    --batch_size 32 \
    --output_path "$RESDIR/${NAME}_wikitext" \
    2>&1 | tee "$WIKILOG"

  echo "[eval] === gsm8k accuracy 5-shot limit100 ($NAME) ==="
  lm_eval \
    --model local-chat-completions \
    --model_args "model=${NAME},base_url=${BASE}/v1/chat/completions,tokenizer=${MODEL_PATH},num_concurrent=32,max_retries=3" \
    --tasks gsm8k \
    --num_fewshot 5 \
    --batch_size 1 \
    --apply_chat_template \
    --gen_kwargs max_gen_toks=1024,temperature=0,do_sample=False \
    --limit 100 \
    --output_path "$RESDIR/${NAME}_gsm8k" \
    2>&1 | tee "$GSMLOG"

  echo "[done] $NAME  -> logs: $WIKILOG , $GSMLOG"
  stop_server
done

echo "ALL_DONE"