Buckets:
| """Decide MQ-large's training mode from MQ vs MQ-soft results (rule fixed before seeing MQ-soft).""" | |
| import json, sys | |
| sys.path.insert(0, "/workspace/think-decide/src") | |
| from analyze_mq import boot | |
| def gain(path, modes): | |
| rows = [json.loads(l) for l in open(path + "/predictions.jsonl")] | |
| return max(boot(rows, m, "parallel", R=2000)[0] for m in modes if m in rows[0]["pred"]) | |
| g_hard = gain("runs/mq/test", ["sequential"]) | |
| g_soft = gain("runs/mqsoft/test", ["sequential", "soft2", "soft3", "soft4"]) | |
| print("--soft" if g_soft >= g_hard else "") | |
| print("MQ seq gain %.2f, MQ-soft best gain %.2f -> %s" % (g_hard, g_soft, "soft" if g_soft >= g_hard else "plain"), file=sys.stderr) | |
Xet Storage Details
- Size:
- 681 Bytes
- Xet hash:
- ea65f4355f7403225664fa6e01a8d1fc56f85d4d9b3b033e947b9f75eede105b
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.