-
When a VC gets issued, it is always stored in the VCR issuer store. This is to keep track of what is issued and to be able to revoke credentials. This was probably designed with low credential count i…
-
HI, getting this error:
```
./cosmprund prune ~/.gaiad/data/ --cosmos-sdk=false
pruning state store
Error: from height 0 and to height -10 must be greater than 0
```
OR with cosmos-sdk true:…
-
Hi,
I am currently calculating heritability in WGS data using the GREML-LDMS method. However, I am facing an issue at the "segment-based LD score" step while running "gcta64 --bfile test --ld-score-r…
-
Traceback (most recent call last):
File "/mnt/disk2/zkm/dpr-scale/dpr_scale/main.py", line 9, in
from pytorch_lightning.callbacks import LearningRateMonitor
File "/opt/tq/miniconda3/envs/q…
-
The ADE is currently pruning user metrics after three weeks. Please remove this time limit so that any telemetry data is retained permanently.
-
In the documentation there is a missing section on:
Things that still need to be Documented
periodic pruning of expired sync data
I assume that this would be done using the `pytho…
TcT2k updated
5 years ago
-
If someone wants to have a light node to be usable asap, syncing and dasing historical blocks might stop them from doing so.
Even with pruning enabled, we might see that devX is not going to be good…
-
In some scenarios, users could be interested in running archive nodes (with finalized states) up to a certain time in the past. This could enable having a low value in `--data-storage-archive-frequenc…
tbenr updated
2 weeks ago
-
Update: This could be solved by merge the input with hm3 snp list in advance. However, the IVW-MR results in MRlap is slightly different from IVW results from the tsmr pipeline using TwoSampleMR (I di…
-
application.db seems to be growing too quickly.
It's currently 29GB in size.
```
29G ./application.db
1.3G ./snapshots
2.7G ./blockstore.db
3.8G ./state.db
22G ./tx_index.db
36K ./evidence.…