anon8231489123
commited on
Commit
•
4ef20dd
1
Parent(s):
2de81a2
added ggml quantization for cuda model
Browse files
ggml_README.txt
ADDED
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
The model is for: https://github.com/ggerganov/llama.cpp
|
2 |
+
|
3 |
+
Date: 2023-04-01
|
4 |
+
ggml model file magic: 0x67676a74 (ggjt in hex)
|
5 |
+
ggml model file version: 1
|
6 |
+
|
7 |
+
Torrent contents:
|
8 |
+
The fine tune described at https://huggingface.co/chavinlo/gpt4-x-alpaca converted to ggml format from https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g/blob/f267949dcd5a5e6451933cec3d0b5661f4f9c889/gpt-x-alpaca-13b-native-4bit-128g-cuda.pt
|
9 |
+
Details about the GPTQ quantization process: https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g/blob/f267949dcd5a5e6451933cec3d0b5661f4f9c889/README.md
|
10 |
+
|
11 |
+
Tools used:
|
12 |
+
[1] Conversion to ggml: https://github.com/ggerganov/llama.cpp/blob/3265b102beb7674d010644ca2a1bd30a58f9f6b5/convert.py and [2]
|
13 |
+
[2] Added extra tokens: https://huggingface.co/chavinlo/alpaca-13b/blob/464a0bd1ec16f3a7d5295a0035aff87f307e62f1/added_tokens.json
|
14 |
+
[3] Migration to the latest llama.cpp model format: https://github.com/ggerganov/llama.cpp/blob/3525899277d2e2bdc8ec3f0e6e40c47251608700/migrate-ggml-2023-03-30-pr613.py
|
15 |
+
|
gpt4-x-alpaca-13b-ggml-q4_1-from-gptq-4bit-128g/ggml-model-q4_1.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d4a640a1ce33009c244a361c6f87733aacbc2bea90e84d3c304a4c8be2bdf22d
|
3 |
+
size 10173322368
|