Why I downloaded starcoder model weight deployment effect and female playground is almost like two models! ?

#31
by seablue - opened

"""

Here is the python code for an algorithm:

class Solution:
def minimumSwaps(self, nums: List[int]) -> int:
i = j = 0
for k, v in enumerate(nums):
if v < nums[i] or (v == nums[i] and k < i):
i = k
if v >= nums[j] or (v == nums[j] and k > j):
j = k
return 0 if i == j else i + len(nums) - 1 - j - (i > j)

please only response the code that convert the above python code to lua code.

Attention:

  1. Do not return information and notes other than the lua code.
  2. When converting to lua, try to use the data structures contained in the lua base library.
  3. If you want to use a data structure that is not included in the base library, implement that structure in lua first.
  4. Make sure that the algorithmic logic of the lua code you generate is consistent with the original python code as a whole.
  5. Make sure that the code you generate can be compiled and run directly, without general syntax errors.
  6. The format you return is as follows:

-- @algorithm
{ lua algorithm }

Response:

"""
That's my prompt, on your playground starcoder answered well, but on my starcoder he would only answer simple python print functions that have nothing to do with the question! ? I used the inference parameter of your github generate.py. Are you sure your playground isn't doing anything else during the inference?

I'm having the same issue, running StarCoder locally doesn't seem to be working well for me. Then I go to the StarCoder playground and all 3 models (StarCoder Plus, StarCoder Base, and StarCoder) are amazing. How can I get that performance on my local model? Which model is used on the local version? Is my model setting or parameter not set right? I can give examples if needed.

Sign up or log in to comment