Use models...max_context_tokens

This commit is contained in:
Paul Gauthier 2023-06-07 12:37:10 -07:00
parent a874d617d3
commit b70f8ea7bc

View file

@ -121,22 +121,24 @@ class Commands:
tokens = len(self.tokenizer.encode(quoted))
res.append((tokens, relative_fname))
print("Context window usage, in k-tokens:")
print("Context window usage, in tokens:")
print()
def fmt(v):
return format(int(v), ",").rjust(6)
total = 0
for tk, msg in res:
tk /= 1024
total += tk
print(f"{tk:6.3f} {msg}")
print(f"{fmt(tk)} {msg}")
print()
print(f"{total:6.3f} total")
print(f"{fmt(total)} total")
limit = 8 if self.coder.main_model == models.GPT4 else 4
limit = self.coder.main_model.max_context_tokens
remaining = limit - total
print(f"{remaining:6.3f} remaining")
print(f"{limit:6.3f} max context window")
print(f"{fmt(remaining)} remaining")
print(f"{fmt(limit)} max context window")
def cmd_undo(self, args):
"Undo the last git commit if it was done by aider"