thegovind's picture
Publish the weight-space programming labs
bb102c0 verified
Raw History Blame Contribute Delete
547 Bytes
{
"about": "Quick start: the unedited model and the released edit on the first 16 practice problems of dev256 (all from GSM8K), 2 answers each, the paper's settings. About 5 minutes on 2 GPUs. Then read the answers with labs/peek.py.",
"runs": [
{
"name": "quick/base",
"model": "base",
"data": "dev256",
"n": 2,
"shards": 1,
"est": 1,
"args": ["--limit", "16"]
},
{
"name": "quick/p50c",
"model": "qwen35-p50c",
"data": "dev256",
"n": 2,
"shards": 1,
"est": 1,
"args": ["--limit", "16"]
}
]
}