README.md

---
language:
- en
license: apache-2.0
tags:
- text-generation-inference
datasets:
- HuggingFaceH4/ultrachat_200k
- openchat/openchat_sharegpt4_dataset
- Open-Orca/SlimOrca
inference: false
model-index:
- name: falcon-rw-1b-chat
  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: 35.58
      name: normalized accuracy
    source:
      url: >-
        https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-chat
      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: 61.12
      name: normalized accuracy
    source:
      url: >-
        https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-chat
      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: 24.51
      name: accuracy
    source:
      url: >-
        https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-chat
      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: 39.62
    source:
      url: >-
        https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-chat
      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: 61.72
      name: accuracy
    source:
      url: >-
        https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-chat
      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: 1.67
      name: accuracy
    source:
      url: >-
        https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ericzzz/falcon-rw-1b-chat
      name: Open LLM Leaderboard
pipeline_tag: text-generation
---
# 🌟 Falcon-RW-1B-Chat

**Falcon-RW-1B-Chat is a conversational model with 1 billion parameters.** It's a further refinement of the [Falcon-RW-1B-Instruct-OpenOrca](https://huggingface.co/ericzzz/falcon-rw-1b-instruct-openorca), trained on selected data from the [HuggingFaceH4/ultrachat_200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) and [openchat/openchat_sharegpt4_dataset](https://huggingface.co/datasets/openchat/openchat_sharegpt4_dataset) datasets.

**✨Try it out at our [Tiny Chat](https://huggingface.co/spaces/ericzzz/tiny-chat) space running on free-tier hardware!✨**

The underlying Falcon-RW-1B-Instruct-OpenOrca model is built on the [Falcon-RW-1B](https://huggingface.co/tiiuae/falcon-rw-1b), a causal decoder-only model. It has been instruction-finetuned using the [Open-Orca/SlimOrca](https://huggingface.co/datasets/Open-Orca/SlimOrca) dataset.

**🎯 Purpose**

The Falcon-RW-1B-Chat aims to add conversational capabilities to the Falcon-RW-1B-Instruct-OpenOrca model. This initiative is driven by the need for a smaller, open-source, instruction-finetuned, ready-to-use model, suitable for users with limited computational resources, like lower-end consumer GPUs.

## [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-chat)

|             Metric              |Value|
|---------------------------------|----:|
|Avg.                             |37.37|
|AI2 Reasoning Challenge (25-Shot)|35.58|
|HellaSwag (10-Shot)              |61.12|
|MMLU (5-Shot)                    |24.51|
|TruthfulQA (0-shot)              |39.62|
|Winogrande (5-shot)              |61.72|
|GSM8k (5-shot)                   | 1.67|

## 📖 Example Code

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "ericzzz/falcon-rw-1b-chat"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name, device_map="auto", torch_dtype=torch.bfloat16
)

chat_history = [
    {"role": "user", "content": "Hello!"},
    {"role": "assistant", "content": "Hello! How can I assist you today?"},
    {"role": "user", "content": "Explain what AI is."},
]

input_ids = tokenizer.apply_chat_template(
    chat_history, tokenize=True, add_generation_prompt=True, return_tensors="pt"
).to(model.device)
output_tokens = model.generate(
    input_ids,
    do_sample=True,
    temperature=0.7,
    repetition_penalty=1.05,
    max_new_tokens=200,
)
output_text = tokenizer.decode(
    output_tokens[0][len(input_ids[0]) :], skip_special_tokens=True
)

print(output_text)
```

## ⚠️ 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.

## 📬 Contact

For further inquiries or feedback, please contact at [email protected].