Update README.md

README.md

@@ -5,73 +5,94 @@ tags:
 - alignment-handbook
 - trl
 - sft
-- generated_from_trainer
-- trl
-- sft
-- generated_from_trainer
 datasets:
-- HuggingFaceTB/Magpie-Pro-300K-Filtered-H4
-- HuggingFaceTB/self-oss-instruct-sc2-H4
-- HuggingFaceTB/OpenHermes-2.5-H4
-- HuggingFaceTB/everyday-topics-MT-conversations-H4
-- HuggingFaceTB/instruct-data-basics-H4
-model-index:
-- name: smollm-350M-instruct-add-basics
-  results: []
+- Magpie-Align/Magpie-Pro-300K-Filtered
+- bigcode/self-oss-instruct-sc2-exec-filter-50k
+- teknium/OpenHermes-2.5
+- HuggingFaceTB/everyday-conversations-llama3.1-2k
 library_name: transformers
+language:
+- en
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/loubnabnl/huggingface/runs/bfq0ndat)
-# smollm-350M-instruct-add-basics
+# SmolLM-360M-Instruct
+
+<center>
+    <img src="https://huggingface.co/datasets/HuggingFaceTB/images/resolve/main/banner_smol.png" alt="SmolLM" width="1100" height="600">
+</center>
+
+
+## Model Summary
+Chat with the model at: https://huggingface.co/spaces/HuggingFaceTB/instant-smol
+
+SmolLM is a series of language models available in three sizes: 135M, 360M, and 1.7B parameters. 
+
+These models are trained on [SmolLM-Corpus](https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus), a curated collection of high-quality educational and synthetic data designed for training LLMs. For further details, we refer to our [blogpost](https://huggingface.co/blog/smollm).
 
-This model is a fine-tuned version of [HuggingFaceTB/SmolLM-360M](https://huggingface.co/HuggingFaceTB/SmolLM-360M) on the HuggingFaceTB/Magpie-Pro-300K-Filtered-H4, the HuggingFaceTB/self-oss-instruct-sc2-H4, the HuggingFaceTB/OpenHermes-2.5-H4, the HuggingFaceTB/everyday-topics-MT-conversations-H4 and the HuggingFaceTB/instruct-data-basics-H4 datasets.
-It achieves the following results on the evaluation set:
-- Loss: 1.2039
+To build SmolLM-Instruct, we finetune the base models on publicly available datasets.
 
-## Model description
+## Changelog
 
-More information needed
+|Release|Description|
+|-|-|
+|v0.1| Initial release of SmolLM-Instruct. We finetune on the permissive subset of the WebInstructSub dataset, combined with StarCoder2-Self-OSS-Instruct. Then, we perform DPO (Direct Preference Optimization) for one epoch on HelpSteer for the 135M and 1.7B models, and argilla/dpo-mix-7k for the 360M model.|
+|v0.2| We changed the finetuning mix to datasets more suitable for smol models. We train on a new dataset of 2k simple everyday conversations we generated by llama3.1-70B [everyday-conversations-llama3.1-2k](huggingface.co/datasets/HuggingFaceTB/everyday-conversations-llama3.1-2k), [Magpie-Pro-300K-Filtere](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered), [self-oss-instruct-sc2-exec-filter-50k](https://huggingface.co/datasets/bigcode/self-oss-instruct-sc2-exec-filter-50k), and a small subset of [OpenHermes-2.5](https://huggingface.co/datasets/teknium/OpenHermes-2.5)|
 
-## Intended uses & limitations
+## Usage
 
-More information needed
+### Local Applications
+⚡ For local applications, you can find optimized implementations of the model in MLC, GGUF and Transformers.js formats, in addition to fast in-browser demos in this collection: https://huggingface.co/collections/HuggingFaceTB/local-smollms-66c0f3b2a15b4eed7fb198d0
 
-## Training and evaluation data
+We noticed that 4bit quantization degrades the quality of the 135M and 360M, so we use `q016` for MLC and ONNX/Transformers.js checkpoints for the WebGPU demos. We also suggest using temperature 0.2 and top-p 0.9.
 
-More information needed
+### Transformers
+```bash
+pip install transformers
+```
 
-## Training procedure
+```python
+# pip install transformers
+from transformers import AutoModelForCausalLM, AutoTokenizer
+checkpoint = "HuggingFaceTB/SmolLM-360M-Instruct"
 
-### Training hyperparameters
+device = "cuda" # for GPU usage or "cpu" for CPU usage
+tokenizer = AutoTokenizer.from_pretrained(checkpoint)
+# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
+model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
 
-The following hyperparameters were used during training:
-- learning_rate: 0.001
-- train_batch_size: 4
-- eval_batch_size: 4
-- seed: 42
-- distributed_type: multi-GPU
-- num_devices: 8
-- gradient_accumulation_steps: 4
-- total_train_batch_size: 128
-- total_eval_batch_size: 32
-- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
-- lr_scheduler_type: cosine
-- lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 1
+messages = [{"role": "user", "content": "What is the capital of France."}]
+input_text=tokenizer.apply_chat_template(messages, tokenize=False)
+print(input_text)
+inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
+print(tokenizer.decode(outputs[0]))
+```
 
-### Training results
+### Chat in TRL
+You can also use the TRL CLI to chat with the model from the terminal:
+```bash
+pip install trl
+trl chat --model_name_or_path HuggingFaceTB/SmolLM-360M-Instruct --device cpu
+```
 
-| Training Loss | Epoch  | Step | Validation Loss |
-|:-------------:|:------:|:----:|:---------------:|
-| 0.8459        | 0.9991 | 817  | 1.2039          |
+## Limitations
 
+Additionally, the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data, we invite users to leverage them as assistive tools rather than definitive sources of information. We find that they can handle general knowledge questions, creative writing and basic Python programming. But they are English only and may have difficulty with arithmetics, editing tasks and complex reasoning. For more details about the models' capabilities, please refer to our [blog post](https://huggingface.co/blog/smollm).
 
-### Framework versions
+## Training parameters
+We train the models using the [alignement-handbook](https://github.com/huggingface/alignment-handbook) with the datasets mentioned in the changelog, using these parameters for v0.2:
+- 1 epoch
+- lr 1e-3
+- cosine schedule
+- warmup ratio 0.1
+- global batch size 262k tokens
 
-- Transformers 4.42.3
-- Pytorch 2.1.2
-- Datasets 2.20.0
-- Tokenizers 0.19.1
+# Citation
+```bash
+@misc{allal2024SmolLM,
+      title={SmolLM - blazingly fast and remarkably powerful}, 
+      author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Leandro von Werra and Thomas Wolf},
+      year={2024},
+}
+```