README.md

---
license: apache-2.0
datasets:
- Anthropic/hh-rlhf
- OpenAssistant/oasst1
- databricks/databricks-dolly-15k
language:
- en
- fr
- de
- es
- it
---
# Model Card for Alfred-40B-0723

`Alfred-40B-0723` is a finetuned version of [Falcon-40B](https://huggingface.co/tiiuae/falcon-40b), obtained with Reinforcement Learning from Human Feedback (RLHF).
It is the first of a series of RLHF models based on Falcon-40B that will be regularly released. It is made available under the Apache 2.0 License.

## Model Details

### Model Description

- **Developed by:** [LightOn](https://www.lighton.ai/)
- **Model type:** Causal decoder-only;
- **Language(s) (NLP):** English, German, Spanish, French (and limited capabilities in Italian, Portuguese, Polish, Dutch, Romanian, Czech, Swedish);
- **License:** Apache 2.0 license.
- **Finetuned from model:** [Falcon-40B](https://huggingface.co/tiiuae/falcon-40b)

## Uses

### Direct Use

`Alfred-40B-0723` can be used as an instruct or chat model. We encourage its usage for research on large language models finetuned with RLHF as well.

### Out-of-Scope Use

Production use without adequate assessment of risks and mitigation; any use cases which may be considered irresponsible or harmful. 

## Bias, Risks, and Limitations

`Alfred-40B-0723` is a finetune of Falcon-40B. As such, it is trained mostly on English, German, Spanish, French, with limited capabilities also in in Italian, Portuguese, Polish, Dutch, Romanian, Czech, Swedish. It will not generalize appropriately to other languages. Furthermore, as it is trained on a large-scale corpora representative of the web, it will carry the stereotypes and biases commonly encountered online.

### Recommendations

We recommend users of `Alfred-40B-0723` to implement appropriate guardrails and precautions in any production use.

## How to Get Started with the Model

Use the code below to get started with the model.

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

model = "lightonai/alfred-40b-0723"
tokenizer = AutoTokenizer.from_pretrained(model)

pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    torch_dtype=torch.bfloat16,
    trust_remote_code=True,
    device_map="auto",
)


sequences = pipeline(
   "Write a short text to announce that the new transformer model Alfred is available in open-source on Huggingface, include emojis.",
    max_length=200,
    do_sample=True,
    top_k=10,
    num_return_sequences=1,
    eos_token_id=tokenizer.eos_token_id,
)
for seq in sequences:
    print(f"Result: {seq['generated_text']}")
```

## Training Details

### Training Data

Alfred-40B-0723 was trained on a mixture of publicly available and in-house curated datasets. 

| **Data source**    |
|--------------------|
| [oasst1](https://huggingface.co/datasets/OpenAssistant/oasst1) | 
| [hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf) |
| [dolly](https://huggingface.co/datasets/databricks/databricks-dolly-15k)  |
| [NatInstV2](https://github.com/allenai/natural-instructions) |
| momentum-internal |

`momentum-internal` is a collection of prompts rated as gold quality from the staff of LightOn in their daily workflow.

### Training Procedure 

`Alfred-40B-0723` was trained on 128 A100 40GB GPUs, using a 3D parallelism strategy (TP=8, PP=4, DP=4) combined with ZeRO.

#### Preprocessing

Samples from each of the datasets have been programmatically formatted to chat, instructions and few-shot promtps.

#### Training Hyperparameters

##### Policy and Value Optimizer Config

| **Hyperparameter** | **Value**  | **Comment**                               |
|--------------------|------------|-------------------------------------------|
| Precision          | `bfloat16` |                                           |
| Optimizer          | AdamW      |                                                            |
| Learning rate      | 1.85e-6    | 10 warm-up steps, cosine decay over a 100 steps to 1.85e-7 |

##### Trainer config
| **Hyperparameter** | **Value**  |
|--------------------|------------|
| Num Rollouts       | 1024       | 
| PPO Epochs         | 1          | 
| Value Epochs       | 1          | 
| Constant KL Coef   | `true`     | 
| Init KL Coef       | 0.01       | 
| Target KL          | 6.0        | 
| K Beta             | 0.1        | 
| Gamma              | 1.0        | 
| GAE Lambda         | 0.95       | 
| Clip Range         | 0.2        | 
| Clip Range Value   | 0.2        | 
| Whiten Advantages  | `true`     | 
| Whiten Rewards     | `false`    | 
| Loss on EPD        | `true`     | 
| Max Steps          | 200        | 
| microbatch_size    | 1          | 
| PPO steps/epoch    | 1          | 
| Value steps/epoch  | 8          | 

##### Trajectory data config
| **Hyperparameter**   | **Value**  |
|----------------------|------------|
| Continuation Max Len | 1024       |
| Continuation Min Len | 0          |
|                Top P | 1.0        |
|          Temperature | 1.0        |
|     # Cached Batches | 128        |
|      Microbatch size | 1          |


## Evaluation

![aggregated evaluation of RAW vs SFT vs PPO - including random baseline - PPO suffers in arithmetic due to effects on calibration](https://i.ibb.co/9yQFJ40/aggregated.png "aggregated evaluation of RAW vs SFT vs PPO - including random baseline")

First evaluation results aggregated from the EleutherAI harness:
- Arithmetic capabilities become much worse
- Common Sense, Paraphrase, Reasoning, Reading Comprehension stay at about the same level
- NLI becomes better and QA gets worse

Overall these results were expected from the literature. Benchmarks don't really correlate with human preference. 
All these metrics use a Select methodology, and it since RLHF models are far less calibrated than raw LLMs, they will be punished in these evaluations. 

Human evaluation is currently ongoing.

### Compute Infrastructure

#### Hardware

Alfred-40B-0723 was trained on AWS SageMaker, on 128 A100 40GB GPUs in P4d instances.

#### Software

Alfred-40B-0723 was trained with a custom RLHF codebase. Training leverages a 3D parallelism approach combined with ZeRO, as well as high-performance kernels such as FlashAttention.

## Model Card Contact

[email protected]