ericzzz's picture
Update README.md
29cc70a verified
---
language:
- en
license: apache-2.0
tags:
- text-generation-inference
datasets:
- Open-Orca/SlimOrca
pipeline_tag: text-generation
inference: false
model-index:
- name: falcon-rw-1b-instruct-openorca
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 34.56
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-instruct-openorca
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 60.93
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-instruct-openorca
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 28.77
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-instruct-openorca
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 37.42
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-instruct-openorca
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 60.69
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-instruct-openorca
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 3.41
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-instruct-openorca
name: Open LLM Leaderboard
---
# ๐ŸŒŸ Falcon-RW-1B-Instruct-OpenOrca
Falcon-RW-1B-Instruct-OpenOrca is a 1B parameter, causal decoder-only model based on [Falcon-RW-1B](https://huggingface.co/tiiuae/falcon-rw-1b) and finetuned on the [Open-Orca/SlimOrca](https://huggingface.co/datasets/Open-Orca/SlimOrca) dataset.
**โœจCheck out our new conversational model [Falcon-RW-1B-Chat](https://huggingface.co/ericzzz/falcon-rw-1b-chat)!โœจ**
**๐Ÿ“Š Evaluation Results**
Falcon-RW-1B-Instruct-OpenOrca was the #1 ranking model (unfortunately not anymore) on [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) in ~1.5B parameters category! A detailed result can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_ericzzz__falcon-rw-1b-instruct-openorca).
| Metric | falcon-rw-1b-instruct-openorca | falcon-rw-1b |
|------------|-------------------------------:|-------------:|
| ARC | 34.56 | 35.07 |
| HellaSwag | 60.93 | 63.56 |
| MMLU | 28.77 | 25.28 |
| TruthfulQA | 37.42 | 35.96 |
| Winogrande | 60.69 | 62.04 |
| GSM8K | 3.41 | 0.53 |
| **Average**| **37.63** | **37.07** |
**๐Ÿš€ Motivations**
1. To create a smaller, open-source, instruction-finetuned, ready-to-use model accessible for users with limited computational resources (lower-end consumer GPUs).
2. To harness the strength of Falcon-RW-1B, a competitive model in its own right, and enhance its capabilities with instruction finetuning.
## ๐Ÿ“– How to Use
The model operates with a structured prompt format, incorporating `<SYS>`, `<INST>`, and `<RESP>` tags to demarcate different parts of the input. The system message and instruction are placed within these tags, with the `<RESP>` tag triggering the model's response.
**๐Ÿ“ Example Code**
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import transformers
import torch
model = 'ericzzz/falcon-rw-1b-instruct-openorca'
tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
'text-generation',
model=model,
tokenizer=tokenizer,
torch_dtype=torch.bfloat16,
device_map='auto',
)
system_message = 'You are a helpful assistant. Give short answers.'
instruction = 'What is AI? Give some examples.'
prompt = f'<SYS> {system_message} <INST> {instruction} <RESP> '
response = pipeline(
prompt,
max_length=200,
repetition_penalty=1.05
)
print(response[0]['generated_text'])
# AI, or Artificial Intelligence, refers to the ability of machines and software to perform tasks that require human intelligence, such as learning, reasoning, and problem-solving. It can be used in various fields like computer science, engineering, medicine, and more. Some common applications include image recognition, speech translation, and natural language processing.
```
## โš ๏ธ Limitations
This model may generate inaccurate or misleading information and is prone to hallucination, creating plausible but false narratives. It lacks the ability to discern factual content from fiction and may inadvertently produce biased, harmful or offensive content. Its understanding of complex, nuanced queries is limited. Users should be aware of this and verify any information obtained from the model.
The model is provided 'as is' without any warranties, and the creators are not liable for any damages arising from its use. Users are responsible for their interactions with the model.
## ๐Ÿ“ฌ Contact
For further inquiries or feedback, please contact at eric.fu96@aol.com.
## [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_ericzzz__falcon-rw-1b-instruct-openorca)
| Metric |Value|
|---------------------------------|----:|
|Avg. |37.63|
|AI2 Reasoning Challenge (25-Shot)|34.56|
|HellaSwag (10-Shot) |60.93|
|MMLU (5-Shot) |28.77|
|TruthfulQA (0-shot) |37.42|
|Winogrande (5-shot) |60.69|
|GSM8k (5-shot) | 3.41|