Instructions to use hchen657Naza/PWM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use hchen657Naza/PWM with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("hchen657Naza/PWM", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
File size: 1,059 Bytes
23e1f91 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 | #!/bin/bash
# Eval the multitask ACT (policy_epoch_6000) on all 4 tasks, 10 eps each,
# same eval env as the single-task evals (eval_act_ckpt.sh: n_exec=24, table_idx=1,
# rand_offset=4000, seed=42). One task per GPU (parallel). Stats auto-found next to ckpt.
LOGD=/home/hongyi/PWM/policy_rl/training_logs
EVALSH=/home/hongyi/PWM/policy_rl/act/eval_act_ckpt.sh
CKPT=/data0/hongyi/act_runs/multitask_30each_chunk25_6k/policy_epoch_6000_seed_0.ckpt
run_task() {
local t=$1 env=$2 gpu=$3
echo "===== MULTITASK EVAL $t (gpu $gpu) $(date +%H:%M:%S) ====="
TASK="$env" GPU_EVAL="$gpu" bash "$EVALSH" "$CKPT" 10
echo "===== DONE $t ====="
}
run_task open_drawer Humanoid-Open-Drawer-v0 0 > "$LOGD/eval_multitask_open_drawer.log" 2>&1 &
run_task open_laptop Humanoid-Open-Laptop-v0 1 > "$LOGD/eval_multitask_open_laptop.log" 2>&1 &
run_task pour_balls Humanoid-Pour-Balls-v0 2 > "$LOGD/eval_multitask_pour_balls.log" 2>&1 &
run_task push_box Humanoid-Push-Box-v0 3 > "$LOGD/eval_multitask_push_box.log" 2>&1 &
wait
echo "ALL MULTITASK EVALS DONE"
|