Buckets:
| # Generate the 3 sensitive probes (text "THE OPEN PAGE" / five-finger hand / egg-counting flat-lay) | |
| # for the montage: teacher | ours r128 (real kernel) | plain NVFP4 r0 | BFL fp8. seed = probe idx. | |
| set -e | |
| export PYTHONPATH=. PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True HF_HUB_DISABLE_XET=1 | |
| export PROMPTS_JSON=outputs/eval/probes.json | |
| P=outputs/eval/h2h_probes; L=$P/logs; mkdir -p $L | |
| python3 -u scripts/32_gen_eval.py teacher $P/teacher 0 3 512 > $L/teacher.log 2>&1 | |
| python3 -u scripts/32_gen_eval.py fq:0 $P/fq0 0 3 512 > $L/fq0.log 2>&1 | |
| python3 -u scripts/35_gen_real.py ours:128 $P/ours128 probe 0 3 512 > $L/ours128.log 2>&1 | |
| python3 -u scripts/41_gen_bfl_fp8.py $P/bfl_fp8 0 3 512 > $L/bfl_fp8.log 2>&1 | |
| echo "### PROBES DONE" | |
Xet Storage Details
- Size:
- 776 Bytes
- Xet hash:
- 9f01e6ce90b7eae1a3039aee7ec8ed1fe8370bc037dacdcc72fe7c5cdbceb9df
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.