Update README.md
Browse files
README.md
CHANGED
@@ -5,7 +5,7 @@ datasets:
|
|
5 |
|
6 |
Mostly untested!
|
7 |
|
8 |
-
# RoPE Scaled QLoRA Fine-tune of Llama-
|
9 |
|
10 |
## Overview
|
11 |
|
@@ -61,15 +61,14 @@ Quantized with AutoGPTQ (bits = 4, group_size = 128, desc_act = True).
|
|
61 |
|
62 |
See original model card below.
|
63 |
|
|
|
64 |
|
65 |
|
66 |
-
|
67 |
-
|
68 |
-
__mostly untested, use if you want, or wait for some validation__
|
69 |
|
70 |
## Overview
|
71 |
|
72 |
-
This is a
|
73 |
|
74 |
This is mostly an extension of the previous gpt-4 series, with a few extras:
|
75 |
|
@@ -80,7 +79,7 @@ This is mostly an extension of the previous gpt-4 series, with a few extras:
|
|
80 |
* riddles
|
81 |
* all coding instructions have an equivalent " PLAINFORMAT" version now (and all rosettacode examples were trained with PLAINFORMAT)
|
82 |
|
83 |
-
This model was fine-tuned with a fork of [
|
84 |
|
85 |
The prompt it was trained with was:
|
86 |
|
@@ -103,7 +102,7 @@ Be sure you are pulling the latest branch!
|
|
103 |
Then, you can invoke it like so (after downloading the model):
|
104 |
```
|
105 |
python -m fastchat.serve.cli \
|
106 |
-
--model-path airoboros-
|
107 |
--temperature 0.5 \
|
108 |
--max-new-tokens 2048 \
|
109 |
--no-history
|
@@ -203,10 +202,11 @@ Or:
|
|
203 |
Write a multi-threaded TCP server in C that accepts a "GET [key]" input and "SET [key] [value]" input, and uses a binary tree to get and store the input values.
|
204 |
```
|
205 |
|
206 |
-
You can optionally add a
|
207 |
|
208 |
```
|
209 |
-
Write a websocket application in node.js.
|
|
|
210 |
```
|
211 |
|
212 |
### Word games / trivia
|
|
|
5 |
|
6 |
Mostly untested!
|
7 |
|
8 |
+
# RoPE Scaled QLoRA Fine-tune of Llama-33b on airoboros-gpt4-1.4.1 (GPTQ)
|
9 |
|
10 |
## Overview
|
11 |
|
|
|
61 |
|
62 |
See original model card below.
|
63 |
|
64 |
+
# Original model card: Jon Durbin's Airoboros 33B GPT4 1.4
|
65 |
|
66 |
|
67 |
+
__not yet tested!__
|
|
|
|
|
68 |
|
69 |
## Overview
|
70 |
|
71 |
+
This is a qlora fine-tune 33b parameter LlaMa model, using completely synthetic training data created gpt4 via https://github.com/jondurbin/airoboros
|
72 |
|
73 |
This is mostly an extension of the previous gpt-4 series, with a few extras:
|
74 |
|
|
|
79 |
* riddles
|
80 |
* all coding instructions have an equivalent " PLAINFORMAT" version now (and all rosettacode examples were trained with PLAINFORMAT)
|
81 |
|
82 |
+
This model was fine-tuned with a fork of [qlora](https://github.com/jondurbin/qlora)
|
83 |
|
84 |
The prompt it was trained with was:
|
85 |
|
|
|
102 |
Then, you can invoke it like so (after downloading the model):
|
103 |
```
|
104 |
python -m fastchat.serve.cli \
|
105 |
+
--model-path airoboros-33b-gpt4-1.4 \
|
106 |
--temperature 0.5 \
|
107 |
--max-new-tokens 2048 \
|
108 |
--no-history
|
|
|
202 |
Write a multi-threaded TCP server in C that accepts a "GET [key]" input and "SET [key] [value]" input, and uses a binary tree to get and store the input values.
|
203 |
```
|
204 |
|
205 |
+
You can optionally add a newline and "PLAINFORMAT" at the end of your prompt to avoid backticks, explanations, etc. and just print the code, e.g.:
|
206 |
|
207 |
```
|
208 |
+
Write a websocket application in node.js.
|
209 |
+
PLAINFORMAT
|
210 |
```
|
211 |
|
212 |
### Word games / trivia
|