Upload folder using huggingface_hub

scripts/entrypoint.sh

@@ -0,0 +1,146 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Force verbose download progress even in non-TTY environments
+export HF_HUB_DISABLE_PROGRESS_BARS=0
+export TRANSFORMERS_VERBOSITY=info
+export HF_HUB_VERBOSITY=info
+export WANDB_DISABLE_WEAVE="${WANDB_DISABLE_WEAVE:-true}"
+
+echo "============================================"
+echo "  F13 Qwen 3.5 Fine-Tuning Container"
+echo "============================================"
+
+if [ -n "${RUNPOD_POD_ID:-}" ]; then
+    echo "Running on RunPod pod: ${RUNPOD_POD_ID}"
+elif [ -n "${AWS_INSTANCE_ID:-}" ]; then
+    echo "Running on AWS instance: ${AWS_INSTANCE_ID}"
+fi
+
+NUM_NODES="${NUM_NODES:-1}"
+NODE_RANK="${NODE_RANK:-0}"
+if [ "${NUM_NODES}" -gt 1 ]; then
+    echo ""
+    echo "--- Multi-Node Cluster ---"
+    echo "  Nodes:       ${NUM_NODES}"
+    echo "  This rank:   ${NODE_RANK}"
+    echo "  Master:      ${MASTER_ADDR:-unknown}:${MASTER_PORT:-29500}"
+    echo "  Node addr:   ${NODE_ADDR:-unknown}"
+    echo "  World size:  ${WORLD_SIZE:-unknown}"
+    echo "  GPUs/node:   ${NUM_TRAINERS:-unknown}"
+
+    if [ -n "${PRIMARY_ADDR:-}" ]; then
+        export NCCL_SOCKET_IFNAME="${NCCL_SOCKET_IFNAME:-ens1}"
+        echo "  NCCL iface:  ${NCCL_SOCKET_IFNAME}"
+    fi
+    echo "--------------------------"
+    echo ""
+    echo "--- Network Interfaces ---"
+    ip -4 addr show 2>/dev/null || echo "(ip command not available)"
+    echo ""
+    echo "--- Connectivity Test ---"
+    if [ -n "${MASTER_ADDR:-}" ]; then
+        echo -n "  TCP to ${MASTER_ADDR}:${MASTER_PORT:-29500} -> "
+        timeout 3 bash -c "echo > /dev/tcp/${MASTER_ADDR}/${MASTER_PORT:-29500}" 2>/dev/null && echo "OK" || echo "UNREACHABLE (expected before torchrun)"
+        echo -n "  TCP to ${MASTER_ADDR}:8080 -> "
+        timeout 3 bash -c "echo > /dev/tcp/${MASTER_ADDR}/8080" 2>/dev/null && echo "OK" || echo "UNREACHABLE (master API may not be up yet)"
+    fi
+    echo "--------------------------"
+fi
+
+nvidia-smi || echo "Warning: nvidia-smi not available"
+echo ""
+
+# --- SSH Setup ---
+SSH_PORT="${SSH_PORT:-22}"
+mkdir -p /var/run/sshd /root/.ssh
+chmod 700 /root/.ssh
+
+if [ -n "${PUBLIC_KEY:-}" ]; then
+    echo "${PUBLIC_KEY}" > /root/.ssh/authorized_keys
+    chmod 600 /root/.ssh/authorized_keys
+    echo "SSH public key installed."
+fi
+
+if [ -n "${SSH_PASSWORD:-}" ]; then
+    echo "root:${SSH_PASSWORD}" | chpasswd
+    sed -i 's/#*PermitRootLogin.*/PermitRootLogin yes/' /etc/ssh/sshd_config
+    sed -i 's/#*PasswordAuthentication.*/PasswordAuthentication yes/' /etc/ssh/sshd_config
+    echo "SSH password auth enabled."
+else
+    sed -i 's/#*PasswordAuthentication.*/PasswordAuthentication no/' /etc/ssh/sshd_config
+fi
+
+sed -i "s/#*Port .*/Port ${SSH_PORT}/" /etc/ssh/sshd_config
+
+if [ ! -f /etc/ssh/ssh_host_rsa_key ]; then
+    ssh-keygen -A
+fi
+
+/usr/sbin/sshd
+echo "SSH daemon started on port ${SSH_PORT}"
+
+# --- HuggingFace / W&B ---
+if [ -n "${HF_TOKEN:-}" ]; then
+    echo "HF_TOKEN is set — logging into HuggingFace Hub..."
+    hf auth login --token "${HF_TOKEN}" --add-to-git-credential 2>/dev/null || true
+    echo "HuggingFace login complete."
+else
+    echo "HF_TOKEN not set — only public models/datasets available."
+fi
+
+if [ -n "${WANDB_API_KEY:-}" ]; then
+    echo "WANDB_API_KEY is set — W&B logging available."
+    wandb login "${WANDB_API_KEY}" 2>/dev/null || true
+else
+    echo "WANDB_API_KEY not set — W&B logging disabled by default."
+fi
+echo ""
+
+# --- Default dataset ---
+if [ ! -f "/workspace/data/train.jsonl" ]; then
+    echo "No custom train.jsonl found, copying bundled default dataset."
+    mkdir -p /workspace/data
+    cp /opt/f13/data/train.jsonl /workspace/data/train.jsonl
+    cp /opt/f13/data/test.jsonl /workspace/data/test.jsonl
+else
+    echo "Found custom dataset at /workspace/data/train.jsonl"
+fi
+
+# --- Pre-flight: verify key CUDA packages ---
+echo "Verifying GPU-accelerated packages..."
+python3 -c "
+checks = []
+try:
+    import flash_attn; checks.append(f'flash-attn {flash_attn.__version__}')
+except Exception as e: checks.append(f'flash-attn MISSING: {e}')
+try:
+    import fla; checks.append('flash-linear-attention OK')
+    from fla.ops.gated_delta_rule import fused_recurrent_gated_delta_rule
+    checks.append('  GatedDeltaNet FLA kernel OK')
+except Exception as e: checks.append(f'flash-linear-attention MISSING: {e}')
+try:
+    import causal_conv1d; checks.append('causal-conv1d OK')
+except Exception as e: checks.append(f'causal-conv1d MISSING: {e}')
+for c in checks: print(f'  [{\"OK\" if \"MISSING\" not in c else \"WARN\"}] {c}')
+"
+echo ""
+
+# --- API Server ---
+API_PORT="${API_PORT:-8080}"
+echo ""
+echo "Starting API server on port ${API_PORT}..."
+
+if [ -z "${API_SECRET:-}" ]; then
+    echo "WARNING: API_SECRET is not set. All API calls will be rejected."
+    echo "         Set API_SECRET env var to enable the training API."
+fi
+
+if [ "${AUTO_TRAIN:-false}" = "true" ]; then
+    echo "AUTO_TRAIN=true — training will start after API server is up."
+fi
+
+exec uvicorn server.app:app \
+    --host 0.0.0.0 \
+    --port "${API_PORT}" \
+    --app-dir /opt/f13

scripts/infer.sh

@@ -0,0 +1,24 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+OUTPUT_DIR="${OUTPUT_DIR:-/workspace/output}"
+USE_HF="${USE_HF:-true}"
+
+CHECKPOINT=$(ls -td "${OUTPUT_DIR}"/v*-*/checkpoint-* 2>/dev/null | head -1)
+
+if [ -z "${CHECKPOINT}" ]; then
+    echo "Error: No checkpoint found in ${OUTPUT_DIR}"
+    echo "Train the model first with: bash /opt/scripts/train.sh"
+    exit 1
+fi
+
+echo "============================================"
+echo "  Inference with checkpoint: ${CHECKPOINT}"
+echo "============================================"
+
+swift infer \
+    --adapters "${CHECKPOINT}" \
+    --stream true \
+    --temperature 0 \
+    --max_new_tokens 2048 \
+    --use_hf "${USE_HF}"

scripts/train.sh

@@ -0,0 +1,223 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Verbose download progress in non-TTY environments (RunPod logs)
+export HF_HUB_DISABLE_PROGRESS_BARS=0
+export TRANSFORMERS_VERBOSITY=info
+export HF_HUB_VERBOSITY=info
+
+MODEL="${MODEL:-Qwen/Qwen3.5-4B}"
+TRAIN_DATA="${TRAIN_DATA:-/workspace/data/train.jsonl}"
+TEST_DATA="${TEST_DATA:-/workspace/data/test.jsonl}"
+OUTPUT_DIR="${OUTPUT_DIR:-/workspace/output}"
+NUM_EPOCHS="${NUM_EPOCHS:-3}"
+BATCH_SIZE="${BATCH_SIZE:-1}"
+GRAD_ACCUM="${GRAD_ACCUM:-4}"
+LR="${LR:-2e-5}"
+MAX_LENGTH="${MAX_LENGTH:-2048}"
+SAVE_STEPS="${SAVE_STEPS:-10}"
+EVAL_STEPS="${EVAL_STEPS:-${SAVE_STEPS}}"
+SAVE_TOTAL_LIMIT="${SAVE_TOTAL_LIMIT:-5}"
+LOGGING_STEPS="${LOGGING_STEPS:-5}"
+USE_HF="${USE_HF:-true}"
+TUNER_TYPE="${TUNER_TYPE:-full}"
+WARMUP_RATIO="${WARMUP_RATIO:-0.1}"
+LR_SCHEDULER_TYPE="${LR_SCHEDULER_TYPE:-cosine}"
+WEIGHT_DECAY="${WEIGHT_DECAY:-0.1}"
+MAX_GRAD_NORM="${MAX_GRAD_NORM:-1.0}"
+OPTIMIZER="${OPTIMIZER:-adamw_torch}"
+SEED="${SEED:-42}"
+NEFTUNE_ALPHA="${NEFTUNE_ALPHA:-0}"
+PACKING="${PACKING:-false}"
+SHUFFLE_DATASET="${SHUFFLE_DATASET:-false}"
+LAZY_TOKENIZE="${LAZY_TOKENIZE:-true}"
+DATASET_NUM_PROC="${DATASET_NUM_PROC:-4}"
+ATTN_IMPL="${ATTN_IMPL:-flash_attn}"
+DEEPSPEED_CONFIG="${DEEPSPEED_CONFIG:-}"
+WANDB_PROJECT="${WANDB_PROJECT:-}"
+RESUME_FROM="${RESUME_FROM:-}"
+USE_FLASH_CKPT="${USE_FLASH_CKPT:-false}"
+EARLY_STOPPING_PATIENCE="${EARLY_STOPPING_PATIENCE:-}"
+EARLY_STOPPING_THRESHOLD="${EARLY_STOPPING_THRESHOLD:-0.0}"
+
+NUM_GPUS=$(nvidia-smi -L 2>/dev/null | wc -l)
+NUM_GPUS=${NUM_GPUS:-1}
+if [ "${NUM_GPUS}" -lt 1 ]; then
+    NUM_GPUS=1
+fi
+NPROC_PER_NODE="${NPROC_PER_NODE:-${NUM_GPUS}}"
+
+if [ "${NPROC_PER_NODE}" -gt 1 ] && [ -z "${DEEPSPEED_CONFIG}" ]; then
+    DEEPSPEED_CONFIG="zero3"
+fi
+
+GPU_IDS=$(seq -s, 0 $((NPROC_PER_NODE - 1)))
+export CUDA_VISIBLE_DEVICES="${CUDA_VISIBLE_DEVICES:-${GPU_IDS}}"
+
+echo "============================================"
+echo "  Qwen 3.5 Fine-Tuning with ms-swift"
+echo "============================================"
+echo "Model:        ${MODEL}"
+echo "Train data:   ${TRAIN_DATA}"
+echo "Test data:    ${TEST_DATA}"
+echo "Output:       ${OUTPUT_DIR}"
+echo "Tuner:        ${TUNER_TYPE}"
+echo "Epochs:       ${NUM_EPOCHS}"
+echo "Batch size:   ${BATCH_SIZE}"
+echo "Grad accum:   ${GRAD_ACCUM}"
+echo "LR:           ${LR}"
+echo "Max length:   ${MAX_LENGTH}"
+echo "GPUs:         ${NPROC_PER_NODE} (CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES})"
+echo "Eval steps:   ${EVAL_STEPS}"
+echo "Save limit:   ${SAVE_TOTAL_LIMIT}"
+echo "Warmup ratio: ${WARMUP_RATIO}"
+echo "LR scheduler: ${LR_SCHEDULER_TYPE}"
+echo "Weight decay: ${WEIGHT_DECAY}"
+echo "Grad clip:    ${MAX_GRAD_NORM}"
+echo "Optimizer:    ${OPTIMIZER}"
+echo "Seed:         ${SEED}"
+echo "NEFTune:      ${NEFTUNE_ALPHA}"
+echo "Packing:      ${PACKING}"
+echo "Lazy tokenize:${LAZY_TOKENIZE}"
+echo "Dataset procs:${DATASET_NUM_PROC}"
+echo "Attn impl:    ${ATTN_IMPL}"
+echo "DeepSpeed:    ${DEEPSPEED_CONFIG:-none}"
+echo "W&B project:  ${WANDB_PROJECT:-disabled}"
+echo "Flash ckpt:   ${USE_FLASH_CKPT}"
+echo "Early stop:   ${EARLY_STOPPING_PATIENCE:-disabled}"
+echo "Resume from:  ${RESUME_FROM:-none}"
+echo "============================================"
+
+EXTRA_ARGS=()
+
+HAS_VAL=false
+if [ -f "${TEST_DATA}" ] || [[ "${TEST_DATA}" == */* && ! "${TEST_DATA}" == /* ]]; then
+    EXTRA_ARGS+=(--val_dataset "${TEST_DATA}")
+    HAS_VAL=true
+fi
+
+if [ -n "${EARLY_STOPPING_PATIENCE}" ] && [ "${HAS_VAL}" = "true" ]; then
+    EXTRA_ARGS+=(
+        --load_best_model_at_end true
+        --metric_for_best_model eval_loss
+        --greater_is_better false
+        --early_stopping_patience "${EARLY_STOPPING_PATIENCE}"
+    )
+    if [ "${EARLY_STOPPING_THRESHOLD}" != "0.0" ] && [ -n "${EARLY_STOPPING_THRESHOLD}" ]; then
+        EXTRA_ARGS+=(--early_stopping_threshold "${EARLY_STOPPING_THRESHOLD}")
+    fi
+    if [ "${SAVE_TOTAL_LIMIT}" -lt 2 ]; then
+        SAVE_TOTAL_LIMIT=2
+        echo "Bumped SAVE_TOTAL_LIMIT to 2 (required for load_best_model_at_end)"
+    fi
+elif [ -n "${EARLY_STOPPING_PATIENCE}" ]; then
+    echo "WARNING: EARLY_STOPPING_PATIENCE ignored — no validation data configured"
+fi
+
+if [ -n "${DEEPSPEED_CONFIG}" ]; then
+    EXTRA_ARGS+=(--deepspeed "${DEEPSPEED_CONFIG}")
+fi
+
+if [ -n "${WANDB_PROJECT}" ] || [ -n "${WANDB_API_KEY:-}" ]; then
+    EXTRA_ARGS+=(--report_to wandb)
+fi
+
+if [ -n "${RESUME_FROM}" ]; then
+    if [ "${RESUME_FROM}" = "auto" ]; then
+        LATEST_CKPT=$(ls -td "${OUTPUT_DIR}"/*/checkpoint-* "${OUTPUT_DIR}"/checkpoint-* 2>/dev/null | head -1)
+        if [ -n "${LATEST_CKPT}" ]; then
+            echo "Auto-resume: found ${LATEST_CKPT}"
+            EXTRA_ARGS+=(--resume_from_checkpoint "${LATEST_CKPT}")
+        else
+            echo "Auto-resume: no checkpoint found, starting fresh"
+        fi
+    else
+        EXTRA_ARGS+=(--resume_from_checkpoint "${RESUME_FROM}")
+    fi
+fi
+
+export NPROC_PER_NODE
+
+# Pre-flight: verify flash-linear-attention is available.
+# Qwen 3.5 silently falls back to O(n²) GatedDeltaNet without it.
+echo ""
+echo "Pre-flight checks..."
+python3 -c "
+try:
+    import fla
+    from fla.ops.gated_delta_rule import fused_recurrent_gated_delta_rule
+    print('[Pre-flight] flash-linear-attention OK — GatedDeltaNet uses FLA kernels')
+except ImportError as e:
+    print(f'[Pre-flight] WARNING: flash-linear-attention not importable: {e}')
+    print('  GatedDeltaNet layers will use naive O(n^2) fallback — expect 2-3x VRAM')
+try:
+    import causal_conv1d
+    print('[Pre-flight] causal-conv1d OK')
+except ImportError:
+    print('[Pre-flight] WARNING: causal-conv1d not available')
+"
+
+# Pre-download model with visible progress before ms-swift starts.
+# ms-swift/transformers download logging is minimal in non-TTY envs.
+echo ""
+echo "Pre-downloading model ${MODEL} (if not cached)..."
+python3 -c "
+import os
+os.environ['HF_HUB_DISABLE_PROGRESS_BARS'] = '0'
+from huggingface_hub import snapshot_download, logging
+logging.set_verbosity_info()
+snapshot_download('${MODEL}', ignore_patterns=['*.gguf', '*.ggml'])
+print('Model download complete.', flush=True)
+"
+echo ""
+
+CMD_ARGS=(
+    --model "${MODEL}"
+    --dataset "${TRAIN_DATA}"
+    --tuner_type "${TUNER_TYPE}"
+    --torch_dtype bfloat16
+    --num_train_epochs "${NUM_EPOCHS}"
+    --per_device_train_batch_size "${BATCH_SIZE}"
+    --per_device_eval_batch_size "${BATCH_SIZE}"
+    --learning_rate "${LR}"
+    --gradient_accumulation_steps "${GRAD_ACCUM}"
+    --eval_strategy steps
+    --eval_steps "${EVAL_STEPS}"
+    --save_steps "${SAVE_STEPS}"
+    --save_total_limit "${SAVE_TOTAL_LIMIT}"
+    --logging_steps "${LOGGING_STEPS}"
+    --max_length "${MAX_LENGTH}"
+    --output_dir "${OUTPUT_DIR}"
+    --warmup_ratio "${WARMUP_RATIO}"
+    --lr_scheduler_type "${LR_SCHEDULER_TYPE}"
+    --weight_decay "${WEIGHT_DECAY}"
+    --max_grad_norm "${MAX_GRAD_NORM}"
+    --optim "${OPTIMIZER}"
+    --seed "${SEED}"
+    --dataloader_num_workers 4
+    --lazy_tokenize "${LAZY_TOKENIZE}"
+    --dataset_num_proc "${DATASET_NUM_PROC}"
+    --attn_impl "${ATTN_IMPL}"
+    --use_hf "${USE_HF}"
+    --gradient_checkpointing true
+    --use_flash_ckpt "${USE_FLASH_CKPT}"
+)
+
+if [ "${NEFTUNE_ALPHA}" != "0" ] && [ -n "${NEFTUNE_ALPHA}" ]; then
+    CMD_ARGS+=(--neftune_noise_alpha "${NEFTUNE_ALPHA}")
+fi
+
+if [ "${PACKING}" = "true" ]; then
+    CMD_ARGS+=(--packing true)
+fi
+
+if [ "${SHUFFLE_DATASET}" = "true" ]; then
+    CMD_ARGS+=(--dataset_shuffle true)
+fi
+
+swift sft "${CMD_ARGS[@]}" "${EXTRA_ARGS[@]}"
+
+echo "============================================"
+echo "  Training complete!"
+echo "  Output saved to: ${OUTPUT_DIR}"
+echo "============================================"