status-im
/
eth2.0-specs
mirror of https://github.com/status-im/eth2.0-specs.git
2
0
Fork 0
Code Issues Projects Releases Wiki Activity

Update inactivity penalty deltas processing 

Two main changes:

1. Inactivity scores are modified to decrease slowly for inactive validators when we are not in a leak, and quickly for active validators
2. The inactivity penalties are applied even not during a leak (note that inactivity _scores_ decrease when outside of a leak)

This has the effect that the inactivity leak "overshoots" the target of finalizing again, and keeps leaking balances a bit more. For inactive validators, this PR sets post-leak recovery to happen 3x faster than the during-leak increase, so if a validator loses 3% during a leak, if they stay offline they should expect to lose another 1% until their score decreases back to zero.
This commit is contained in:
vbuterin 2021-05-07 11:05:27 -07:00 committed by GitHub
parent 637cf0743c
commit d1bedbfbec
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 17 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
specs/altair/beacon-chain.md
View File

@ -120,6 +120,7 @@ This patch updates a few configuration values to move penalty parameters closer
| - | - |
| `SYNC_COMMITTEE_SIZE` | `uint64(2**9)` (= 512) |
| `INACTIVITY_SCORE_BIAS` | `uint64(4)` |
| `INACTIVITY_SCORE_RECOVERY_RATE` | `uint64(16)` |
### Time parameters
@ -397,17 +398,16 @@ def get_inactivity_penalty_deltas(state: BeaconState) -> Tuple[Sequence[Gwei], S
"""
rewards = [Gwei(0) for _ in range(len(state.validators))]
penalties = [Gwei(0) for _ in range(len(state.validators))]
if is_in_inactivity_leak(state):
previous_epoch = get_previous_epoch(state)
matching_target_indices = get_unslashed_participating_indices(state, TIMELY_TARGET_FLAG_INDEX, previous_epoch)
for index in get_eligible_validator_indices(state):
for (_, weight) in get_flag_indices_and_weights():
# This inactivity penalty cancels the flag reward corresponding to the flag index
penalties[index] += Gwei(get_base_reward(state, index) * weight // WEIGHT_DENOMINATOR)
if index not in matching_target_indices:
penalty_numerator = state.validators[index].effective_balance * state.inactivity_scores[index]
penalty_denominator = INACTIVITY_SCORE_BIAS * INACTIVITY_PENALTY_QUOTIENT_ALTAIR
penalties[index] += Gwei(penalty_numerator // penalty_denominator)
previous_epoch = get_previous_epoch(state)
matching_target_indices = get_unslashed_participating_indices(state, TIMELY_TARGET_FLAG_INDEX, previous_epoch)
for index in get_eligible_validator_indices(state):
for (_, weight) in get_flag_indices_and_weights():
# This inactivity penalty cancels the flag reward corresponding to the flag index
penalties[index] += Gwei(get_base_reward(state, index) * weight // WEIGHT_DENOMINATOR)
if index not in matching_target_indices:
penalty_numerator = state.validators[index].effective_balance * state.inactivity_scores[index]
penalty_denominator = INACTIVITY_SCORE_BIAS * INACTIVITY_PENALTY_QUOTIENT_ALTAIR
penalties[index] += Gwei(penalty_numerator // penalty_denominator)
return rewards, penalties
```
@ -621,11 +621,14 @@ def process_justification_and_finalization(state: BeaconState) -> None:
```python
def process_inactivity_updates(state: BeaconState) -> None:
for index in get_eligible_validator_indices(state):
# Increase inactivity score of inactive validators
if index in get_unslashed_participating_indices(state, TIMELY_TARGET_FLAG_INDEX, get_previous_epoch(state)):
if state.inactivity_scores[index] > 0:
state.inactivity_scores[index] -= 1
elif is_in_inactivity_leak(state):
state.inactivity_scores[index] -= min(1, state.inactivity_scores[index])
else:
state.inactivity_scores[index] += INACTIVITY_SCORE_BIAS
# Decrease the score of all validators for forgiveness when not during a leak
if not if is_in_inactivity_leak(state):
state.inactivity_scores[index] -= min(INACTIVITY_SCORE_RECOVERY_RATE, state.inactivity_scores[index])
```
#### Rewards and penalties