Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,57 @@
|
|
1 |
-
---
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: meta-llama/Llama-3.1-8B
|
3 |
+
datasets:
|
4 |
+
- mlfoundations-dev/oh-dcft-v3.1-claude-3-5-sonnet-20241022
|
5 |
+
- mpasila/Sonnet3.5-SlimOrcaDedupCleaned-4k-context
|
6 |
+
language:
|
7 |
+
- en
|
8 |
+
license: mit
|
9 |
+
pipeline_tag: text-generation
|
10 |
+
tags:
|
11 |
+
- llama3.1
|
12 |
+
- sonnet
|
13 |
+
- claude
|
14 |
+
quantized_by: ayan4m1
|
15 |
+
inference: false
|
16 |
+
fine-tuning: true
|
17 |
+
---
|
18 |
+
|
19 |
+
## GGUF Quantizations of Llama-3.1-8B Sonnet fine-tuning
|
20 |
+
|
21 |
+
Using unsloth for fine-tuning and quantization:
|
22 |
+
|
23 |
+
```
|
24 |
+
==((====))== Unsloth 2025.2.4: Fast Llama patching. Transformers: 4.48.2.
|
25 |
+
\\ /| GPU: NVIDIA A100-SXM4-40GB. Max memory: 39.557 GB. Platform: Linux.
|
26 |
+
O^O/ \_/ \ Torch: 2.5.1+cu124. CUDA: 8.0. CUDA Toolkit: 12.4. Triton: 3.1.0
|
27 |
+
\ / Bfloat16 = TRUE. FA [Xformers = 0.0.29. FA2 = False]
|
28 |
+
"-____-" Free Apache license: http://github.com/unslothai/unsloth
|
29 |
+
```
|
30 |
+
|
31 |
+
Original model: https://huggingface.co/meta-llama/Llama-3.1-8B
|
32 |
+
|
33 |
+
Applied open Sonnet datasets containing ~1.2mn question/answer pairs for fine-tuning.
|
34 |
+
|
35 |
+
## Prompt format
|
36 |
+
|
37 |
+
```
|
38 |
+
<|begin_of_text|>{prompt}
|
39 |
+
```
|
40 |
+
|
41 |
+
## Download a file (not the whole branch) from below:
|
42 |
+
|
43 |
+
| Filename | Quant type | File Size | Split | Description |
|
44 |
+
| ------------------------------------------------------------------------------------------------------------------------------------ | ---------- | --------- | ----- | ------------------------------------------------------------------- |
|
45 |
+
| [Llama-3.1-8B-Sonnet-Q8_0.gguf](https://huggingface.co/ayan4m1/Llama-3.1-8B-Sonnet-GGUF/tree/main/Llama-3.1-8B-Sonnet-Q8_0.gguf) | Q8_0 | 74.98GB | true | Extremely high quality, generally unneeded but max available quant. |
|
46 |
+
| [Llama-3.1-8B-Sonnet-Q6_K.gguf](https://huggingface.co/ayan4m1/Llama-3.1-8B-Sonnet-GGUF/tree/main/Llama-3.1-8B-Sonnet-Q6_K.gguf) | Q6_K | 57.89GB | true | Very high quality, near perfect, _recommended_. |
|
47 |
+
| [Llama-3.1-8B-Sonnet-Q5_K_M.gguf](https://huggingface.co/ayan4m1/Llama-3.1-8B-Sonnet-GGUF/tree/main/Llama-3.1-8B-Sonnet-Q5_K_M.gguf) | Q5_K_M | 49.95GB | true | High quality, _recommended_. |
|
48 |
+
| [Llama-3.1-8B-Sonnet-Q4_K_M.gguf](https://huggingface.co/ayan4m1/Llama-3.1-8B-Sonnet-GGUF/blob/main/Llama-3.1-8B-Sonnet-Q4_K_M.gguf) | Q4_K_M | 42.52GB | false | Good quality, default size for must use cases, _recommended_. |
|
49 |
+
| [Llama-3.1-8B-Sonnet-Q3_K_L.gguf](https://huggingface.co/ayan4m1/Llama-3.1-8B-Sonnet-GGUF/blob/main/Llama-3.1-8B-Sonnet-Q3_K_L.gguf) | Q3_K_L | 37.14GB | false | Lower quality but usable, good for low RAM availability. |
|
50 |
+
| [Llama-3.1-8B-Sonnet-Q3_K_M.gguf](https://huggingface.co/ayan4m1/Llama-3.1-8B-Sonnet-GGUF/blob/main/Llama-3.1-8B-Sonnet-Q3_K_M.gguf) | Q3_K_M | 30.91GB | false | Lower quality, not recommended. |
|
51 |
+
| [Llama-3.1-8B-Sonnet-Q2_K.gguf](https://huggingface.co/ayan4m1/Llama-3.1-8B-Sonnet-GGUF/blob/main/Llama-3.1-8B-Sonnet-Q2_K.gguf) | Q2_K | 26.38GB | false | Very low quality but surprisingly usable. |
|
52 |
+
|
53 |
+
## Credits
|
54 |
+
|
55 |
+
Thanks to Meta and [mlfoundations-dev](https://huggingface.co/mlfoundations-dev) for providing the data used to create this fine-tuning.
|
56 |
+
|
57 |
+
Thanks to [bartowski](https://huggingface.co/bartowski) for this README template.
|