Update README.md
Browse files
README.md
CHANGED
@@ -8,19 +8,26 @@ base_model: tiiuae/falcon-40b
|
|
8 |
inference: false
|
9 |
---
|
10 |
|
11 |
-
|
12 |
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
## Model Summary
|
16 |
|
17 |
-
- **Model Type:**
|
18 |
- **Language(s):** English
|
19 |
- **Base Model:** [Falcon-40B](https://huggingface.co/tiiuae/falcon-40b) (License: [Apache 2.0](https://huggingface.co/tiiuae/falcon-40b#license))
|
20 |
- **Dataset:** [OpenAssistant/oasst1](https://huggingface.co/datasets/OpenAssistant/oasst1) (License: [Apache 2.0](https://huggingface.co/datasets/OpenAssistant/oasst1/blob/main/LICENSE))
|
21 |
- **License:** Apache 2.0 inherited from "Base Model" and "Dataset"
|
22 |
|
23 |
-
The model was fine-tuned in 4-bit precision using `peft` adapters, `transformers`, and `bitsandbytes`. Training relied on a method called "Low Rank Adapters" ([LoRA](https://arxiv.org/pdf/2106.09685.pdf)), specifically the [QLoRA](https://arxiv.org/abs/2305.14314) variant. The run took approximately 10 hours and was executed on a workstation with a single A100-SXM NVIDIA GPU with 37 GB of available memory. See attached [Colab Notebook](https://huggingface.co/dfurman/
|
24 |
|
25 |
### Model Date
|
26 |
|
@@ -43,7 +50,7 @@ To prompt the chat model, use the following format:
|
|
43 |
<bot>:"""
|
44 |
```
|
45 |
|
46 |
-
**Falcon-
|
47 |
```
|
48 |
Dear Friends,
|
49 |
|
@@ -63,7 +70,7 @@ Daniel
|
|
63 |
<bot>:
|
64 |
```
|
65 |
|
66 |
-
**Falcon-
|
67 |
```
|
68 |
Here is a list of things to do in San Francisco:
|
69 |
|
@@ -126,7 +133,7 @@ import torch
|
|
126 |
from peft import PeftModel, PeftConfig
|
127 |
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
128 |
|
129 |
-
peft_model_id = "dfurman/
|
130 |
config = PeftConfig.from_pretrained(peft_model_id)
|
131 |
|
132 |
bnb_config = BitsAndBytesConfig(
|
@@ -186,7 +193,7 @@ print(generated_text.split("<human>: ")[1].split("<bot>: ")[-1])
|
|
186 |
|
187 |
## Reproducibility
|
188 |
|
189 |
-
See attached [Colab Notebook](https://huggingface.co/dfurman/
|
190 |
|
191 |
### CUDA Info
|
192 |
|
|
|
8 |
inference: false
|
9 |
---
|
10 |
|
11 |
+
<div align="center">
|
12 |
|
13 |
+
<img src="./falcon.webp" width="150px">
|
14 |
+
|
15 |
+
</div>
|
16 |
+
|
17 |
+
|
18 |
+
# Falcon-40B-Chat-v0.1 🦅
|
19 |
+
|
20 |
+
Falcon-40B-Chat-v0.1 is a chatbot model for dialogue generation. It was built by fine-tuning [Falcon-40B](https://huggingface.co/tiiuae/falcon-40b) on the [OpenAssistant/oasst1](https://huggingface.co/datasets/OpenAssistant/oasst1) dataset. This repo only includes the LoRA adapters from fine-tuning with 🤗's [peft](https://github.com/huggingface/peft) package.
|
21 |
|
22 |
## Model Summary
|
23 |
|
24 |
+
- **Model Type:** Decoder-only
|
25 |
- **Language(s):** English
|
26 |
- **Base Model:** [Falcon-40B](https://huggingface.co/tiiuae/falcon-40b) (License: [Apache 2.0](https://huggingface.co/tiiuae/falcon-40b#license))
|
27 |
- **Dataset:** [OpenAssistant/oasst1](https://huggingface.co/datasets/OpenAssistant/oasst1) (License: [Apache 2.0](https://huggingface.co/datasets/OpenAssistant/oasst1/blob/main/LICENSE))
|
28 |
- **License:** Apache 2.0 inherited from "Base Model" and "Dataset"
|
29 |
|
30 |
+
The model was fine-tuned in 4-bit precision using `peft` adapters, `transformers`, and `bitsandbytes`. Training relied on a method called "Low Rank Adapters" ([LoRA](https://arxiv.org/pdf/2106.09685.pdf)), specifically the [QLoRA](https://arxiv.org/abs/2305.14314) variant. The run took approximately 10 hours and was executed on a workstation with a single A100-SXM NVIDIA GPU with 37 GB of available memory. See attached [Colab Notebook](https://huggingface.co/dfurman/Falcon-40B-Chat-v0.1/blob/main/finetune_falcon40b_oasst1_with_bnb_peft.ipynb) for the code and hyperparams used to train the model.
|
31 |
|
32 |
### Model Date
|
33 |
|
|
|
50 |
<bot>:"""
|
51 |
```
|
52 |
|
53 |
+
**Falcon-40B-Chat-v0.1**:
|
54 |
```
|
55 |
Dear Friends,
|
56 |
|
|
|
70 |
<bot>:
|
71 |
```
|
72 |
|
73 |
+
**Falcon-40B-Chat-v0.1**:
|
74 |
```
|
75 |
Here is a list of things to do in San Francisco:
|
76 |
|
|
|
133 |
from peft import PeftModel, PeftConfig
|
134 |
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
135 |
|
136 |
+
peft_model_id = "dfurman/Falcon-40B-Chat-v0.1"
|
137 |
config = PeftConfig.from_pretrained(peft_model_id)
|
138 |
|
139 |
bnb_config = BitsAndBytesConfig(
|
|
|
193 |
|
194 |
## Reproducibility
|
195 |
|
196 |
+
See attached [Colab Notebook](https://huggingface.co/dfurman/Falcon-40B-Chat-v0.1/blob/main/finetune_falcon40b_oasst1_with_bnb_peft.ipynb) for the code (and hyperparams) used to train the model.
|
197 |
|
198 |
### CUDA Info
|
199 |
|