#CodeLlama not generating
23 messages · Page 1 of 1 (latest)
Just tested - seems to be working for me :d...
Possible that the server got overloaded....
Hmm maybe, are you passing anything into the system prompt?
Here's my system prompt:
The hash of a 0-indexed string s of length k, given integers p and m, is computed using the following function:
hash(s, p, m) = (val(s[0]) * p0 + val(s[1]) * p1 + ... + val(s[k-1]) * pk-1) mod m.
Where val(s[i]) represents the index of s[i] in the alphabet from val('a') = 1 to val('z') = 26.
You are given a string s and the integers power, modulo, k, and hashValue. Return sub, the first substring of s of length k such that hash(sub, power, modulo) == hashValue.
The test cases will be generated such that an answer always exists.
A substring is a contiguous non-empty sequence of characters within a string.
Example 1:
Input: s = "leetcode", power = 7, modulo = 20, k = 2, hashValue = 0
Output: "ee"
Explanation: The hash of "ee" can be computed to be hash("ee", 7, 20) = (5 * 1 + 5 * 7) mod 20 = 40 mod 20 = 0.
"ee" is the first substring of length 2 with hashValue 0. Hence, we return "ee".
Example 2:
Input: s = "fbxzaad", power = 31, modulo = 100, k = 3, hashValue = 32
Output: "fbx"
Explanation: The hash of "fbx" can be computed to be hash("fbx", 31, 100) = (6 * 1 + 2 * 31 + 24 * 312) mod 100 = 23132 mod 100 = 32.
The hash of "bxz" can be computed to be hash("bxz", 31, 100) = (2 * 1 + 24 * 31 + 26 * 312) mod 100 = 25732 mod 100 = 32.
"fbx" is the first substring of length 3 with hashValue 32. Hence, we return "fbx".
Note that "bxz" also has a hash of 32 but it appears later than "fbx".
Constraints:
1 <= k <= s.length <= 2 * 104
1 <= power, modulo <= 109
0 <= hashValue < modulo
s consists of lowercase English letters only.
The test cases are generated such that an answer always exists.
My chat:
Here's your starting code:
class Solution(object):
def subStrHash(self, s, power, modulo, k, hashValue):
"""
:type s: str
:type power: int
:type modulo: int
:type k: int
:type hashValue: int
:rtype: str
"""
Any my params are:
temp: 0.3
Top P: 0.5
Max Tokens: 4000
Chat Memory: 42
Just tried again and still nothing.
Hmm, that's weird that that would be a problem. Most other models let us change both, no?
The rec is that if you use one, you shouldn't change the other: https://community.openai.com/t/a-better-explanation-of-top-p/2426
The tooltip for “Top P” in playground doesn’t really tell me what it does, qualitatively. I get that higher numbers get more diverse options for outputs considered, but I don’t have a good sense for practical differences. Can someone provide some examples of when or why one would want that high or low? The tool tip example says 0.5 means it wil...
You can indeed still submit both, but it will likely... bork the model lol
Ah, I see. I just reset Top P and kept Temprerature at 0.3 and I'm still getting the same issue.
Fascinating, seeing the same thing
I'm not sure what's causing it tbh, we send that to the model
it could be that temp too low?...
if you reset it to default it should work - so with that as a baseline, I wonder which one is the cutoff
Hmm, possible. I'll try resetings the temp and changing the Top P.
Ya no even with the Temp at the default 0.880 and the Top P at 0.3 it doesn't work.
And when I set both Temp and Top P to default it works...
Thank you but I'd like to be able to change temp and top p settings. Which is why I want to use openrouter
I wonder what's the threshold lol... maybe both has to be higher than 0.5?..
Not sure lol. Do you think it's an issue with OpenRouter or the model itself?
I think it's the model itself, as far as OR is concerned, we pipe the request straight to the model