TheBloke commited on
Commit
29c4ce5
1 Parent(s): 8a694bb

New GGMLv3 format for breaking llama.cpp change May 19th commit 2d5db48

Browse files
Files changed (1) hide show
  1. README.md +10 -10
README.md CHANGED
@@ -18,20 +18,22 @@ This repo contains GGML files for for CPU inference using [llama.cpp](https://gi
18
  * [4bit GPTQ models for GPU inference](https://huggingface.co/TheBloke/wizardLM-7B-GPTQ)
19
  * [Unquantised model in HF format](https://huggingface.co/TheBloke/wizardLM-7B-HF)
20
 
21
- ## REQUIRES LATEST LLAMA.CPP (May 12th 2023 - commit b9fd7ee)!
22
 
23
- llama.cpp recently made a breaking change to its quantisation methods.
24
 
25
- I have re-quantised the GGML files in this repo. Therefore you will require llama.cpp compiled on May 12th or later (commit `b9fd7ee` or later) to use them.
26
 
27
- The previous files, which will still work in older versions of llama.cpp, can be found in branch `previous_llama`.
28
 
29
  ## Provided files
30
  | Name | Quant method | Bits | Size | RAM required | Use case |
31
  | ---- | ---- | ---- | ---- | ---- | ----- |
32
  `WizardLM-7B.GGML.q4_0.bin` | q4_0 | 4bit | 4.2GB | 6GB | 4bit. |
33
- `WizardLM-7B.GGML.q5_0.bin` | q5_0 | 5bit | 4.63GB | 7GB | Higher quality inference than 4bit, at cost of slightly higher resources. |
34
- `WizardLM-7B.GGML.q5_1.bin` | q5_1 | 5bit | 5.0GB | 7GB | Higher quality and resource usage again. |
 
 
35
 
36
  ## How to run in `llama.cpp`
37
 
@@ -51,9 +53,7 @@ If you want to have a chat-style conversation, replace the `-p <PROMPT>` argumen
51
 
52
  Further instructions here: [text-generation-webui/docs/llama.cpp-models.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp-models.md).
53
 
54
- Note: at this time text-generation-webui may not support the new llama.cpp quantisation methods (May 12th)
55
-
56
- **Thireus** has written a [great guide on how to update it to the latest llama.cpp code](https://huggingface.co/TheBloke/wizardLM-7B-GGML/discussions/5) to get support for thew newer files quicker.
57
 
58
  # Original model info
59
 
@@ -61,4 +61,4 @@ Overview of Evol-Instruct
61
  Evol-Instruct is a novel method using LLMs instead of humans to automatically mass-produce open-domain instructions of various difficulty levels and skills range, to improve the performance of LLMs.
62
 
63
  ![info](https://github.com/nlpxucan/WizardLM/raw/main/imgs/git_overall.png)
64
- ![info](https://github.com/nlpxucan/WizardLM/raw/main/imgs/git_running.png)
 
18
  * [4bit GPTQ models for GPU inference](https://huggingface.co/TheBloke/wizardLM-7B-GPTQ)
19
  * [Unquantised model in HF format](https://huggingface.co/TheBloke/wizardLM-7B-HF)
20
 
21
+ ## THE FILES IN MAIN BRANCH REQUIRES LATEST LLAMA.CPP (May 19th 2023 - commit 2d5db48)!
22
 
23
+ llama.cpp recently made another breaking change to its quantisation methods - https://github.com/ggerganov/llama.cpp/pull/1508
24
 
25
+ I have quantised the GGML files in this repo with the latest version. Therefore you will require llama.cpp compiled on May 19th or later (commit `2d5db48` or later) to use them.
26
 
27
+ For files compatible with the previous version of llama.cpp, please see branch `previous_llama_ggmlv2`.
28
 
29
  ## Provided files
30
  | Name | Quant method | Bits | Size | RAM required | Use case |
31
  | ---- | ---- | ---- | ---- | ---- | ----- |
32
  `WizardLM-7B.GGML.q4_0.bin` | q4_0 | 4bit | 4.2GB | 6GB | 4bit. |
33
+ `WizardLM-7B.GGML.q4_1.bin` | q4_0 | 4bit | 4.63GB | 6GB | 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models. |
34
+ `WizardLM-7B.GGML.q5_0.bin` | q5_0 | 5bit | 4.63GB | 7GB | 5-bit. Higher accuracy, higher resource usage and slower inference.|
35
+ `WizardLM-7B.GGML.q5_1.bin` | q5_1 | 5bit | 5.0GB | 7GB | 5-bit. Even higher accuracy, and higher resource usage and slower inference. |
36
+ `WizardLM-7B.GGML.q8_0.bin` | q8_0 | 8bit | 8GB | 10GB | 8-bit. Almost indistinguishable from float16. Huge resource use and slow. Not recommended for normal use. |
37
 
38
  ## How to run in `llama.cpp`
39
 
 
53
 
54
  Further instructions here: [text-generation-webui/docs/llama.cpp-models.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp-models.md).
55
 
56
+ Note: at this time text-generation-webui may not support the new May 19th llama.cpp quantisation methods for q4_0, q4_1 and q8_0 files.
 
 
57
 
58
  # Original model info
59
 
 
61
  Evol-Instruct is a novel method using LLMs instead of humans to automatically mass-produce open-domain instructions of various difficulty levels and skills range, to improve the performance of LLMs.
62
 
63
  ![info](https://github.com/nlpxucan/WizardLM/raw/main/imgs/git_overall.png)
64
+ ![info](https://github.com/nlpxucan/WizardLM/raw/main/imgs/git_running.png)