Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -1,17 +1,11 @@
|
|
1 |
---
|
2 |
quantized_by: bartowski
|
3 |
pipeline_tag: text-generation
|
4 |
-
tags:
|
5 |
-
- language
|
6 |
-
- granite-3.1
|
7 |
-
license: apache-2.0
|
8 |
-
inference: false
|
9 |
-
base_model: ibm-granite/granite-3.1-3b-a800m-instruct
|
10 |
---
|
11 |
|
12 |
## Llamacpp imatrix Quantizations of granite-3.1-3b-a800m-instruct
|
13 |
|
14 |
-
Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/
|
15 |
|
16 |
Original model: https://huggingface.co/ibm-granite/granite-3.1-3b-a800m-instruct
|
17 |
|
@@ -27,10 +21,15 @@ Run them in [LM Studio](https://lmstudio.ai/)
|
|
27 |
<|start_of_role|>assistant<|end_of_role|>
|
28 |
```
|
29 |
|
|
|
|
|
|
|
|
|
30 |
## Download a file (not the whole branch) from below:
|
31 |
|
32 |
| Filename | Quant type | File Size | Split | Description |
|
33 |
| -------- | ---------- | --------- | ----- | ----------- |
|
|
|
34 |
| [granite-3.1-3b-a800m-instruct-Q8_0.gguf](https://huggingface.co/bartowski/granite-3.1-3b-a800m-instruct-GGUF/blob/main/granite-3.1-3b-a800m-instruct-Q8_0.gguf) | Q8_0 | 3.51GB | false | Extremely high quality, generally unneeded but max available quant. |
|
35 |
| [granite-3.1-3b-a800m-instruct-Q6_K_L.gguf](https://huggingface.co/bartowski/granite-3.1-3b-a800m-instruct-GGUF/blob/main/granite-3.1-3b-a800m-instruct-Q6_K_L.gguf) | Q6_K_L | 2.73GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. |
|
36 |
| [granite-3.1-3b-a800m-instruct-Q6_K.gguf](https://huggingface.co/bartowski/granite-3.1-3b-a800m-instruct-GGUF/blob/main/granite-3.1-3b-a800m-instruct-Q6_K.gguf) | Q6_K | 2.71GB | false | Very high quality, near perfect, *recommended*. |
|
|
|
1 |
---
|
2 |
quantized_by: bartowski
|
3 |
pipeline_tag: text-generation
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
---
|
5 |
|
6 |
## Llamacpp imatrix Quantizations of granite-3.1-3b-a800m-instruct
|
7 |
|
8 |
+
Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/b4369">b4369</a> for quantization.
|
9 |
|
10 |
Original model: https://huggingface.co/ibm-granite/granite-3.1-3b-a800m-instruct
|
11 |
|
|
|
21 |
<|start_of_role|>assistant<|end_of_role|>
|
22 |
```
|
23 |
|
24 |
+
## What's new:
|
25 |
+
|
26 |
+
Fix tokenizer
|
27 |
+
|
28 |
## Download a file (not the whole branch) from below:
|
29 |
|
30 |
| Filename | Quant type | File Size | Split | Description |
|
31 |
| -------- | ---------- | --------- | ----- | ----------- |
|
32 |
+
| [granite-3.1-3b-a800m-instruct-f16.gguf](https://huggingface.co/bartowski/granite-3.1-3b-a800m-instruct-GGUF/blob/main/granite-3.1-3b-a800m-instruct-f16.gguf) | f16 | 6.60GB | false | Full F16 weights. |
|
33 |
| [granite-3.1-3b-a800m-instruct-Q8_0.gguf](https://huggingface.co/bartowski/granite-3.1-3b-a800m-instruct-GGUF/blob/main/granite-3.1-3b-a800m-instruct-Q8_0.gguf) | Q8_0 | 3.51GB | false | Extremely high quality, generally unneeded but max available quant. |
|
34 |
| [granite-3.1-3b-a800m-instruct-Q6_K_L.gguf](https://huggingface.co/bartowski/granite-3.1-3b-a800m-instruct-GGUF/blob/main/granite-3.1-3b-a800m-instruct-Q6_K_L.gguf) | Q6_K_L | 2.73GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. |
|
35 |
| [granite-3.1-3b-a800m-instruct-Q6_K.gguf](https://huggingface.co/bartowski/granite-3.1-3b-a800m-instruct-GGUF/blob/main/granite-3.1-3b-a800m-instruct-Q6_K.gguf) | Q6_K | 2.71GB | false | Very high quality, near perfect, *recommended*. |
|