turboderp-org/exllamav2
1
0
Fork 0
mirror of https://github.com/turboderp-org/exllamav2.git synced 2026-04-20 14:29:28 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove TODO items

Browse Source
This commit is contained in:
turboderp
2024-02-14 20:00:10 +01:00
parent b0dc588d9b
commit 9c37d64d74
2 changed files with 0 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
exllamav2/generator/base.py
View File

@@ -42,8 +42,6 @@ class ExLlamaV2BaseGenerator:
return self.sequence_ids.shape[-1] >= self.model.config.max_seq_len
# TODO: Argument to allow different random samples over batch dimension
def generate_simple(self, prompt: str or list,
gen_settings: ExLlamaV2Sampler.Settings,
num_tokens: int,

1
exllamav2/model.py
View File

@@ -173,7 +173,6 @@ class ExLlamaV2:
constant_size = sincos_size * 2
# Max size of hidden state
# TODO: Option to reserve space for cache while loading model
state_size = self.config.hidden_size * self.config.max_input_len * self.config.max_batch_size * 2
mask_size = self.config.max_input_len ** 2 * self.config.max_batch_size * 2