Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,58 +1,31 @@
|
|
| 1 |
---
|
| 2 |
-
library_name: transformers
|
| 3 |
license: apache-2.0
|
| 4 |
-
|
|
|
|
| 5 |
tags:
|
| 6 |
-
-
|
| 7 |
-
|
| 8 |
-
-
|
| 9 |
-
|
|
|
|
|
|
|
| 10 |
---
|
| 11 |
|
| 12 |
-
|
| 13 |
-
should probably proofread and complete it, then remove this comment. -->
|
| 14 |
|
| 15 |
-
|
| 16 |
|
| 17 |
-
|
|
|
|
| 18 |
|
| 19 |
-
## Model
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
## Training and evaluation data
|
| 28 |
-
|
| 29 |
-
More information needed
|
| 30 |
-
|
| 31 |
-
## Training procedure
|
| 32 |
-
|
| 33 |
-
### Training hyperparameters
|
| 34 |
-
|
| 35 |
-
The following hyperparameters were used during training:
|
| 36 |
-
- learning_rate: 0.00015
|
| 37 |
-
- train_batch_size: 4
|
| 38 |
-
- eval_batch_size: 16
|
| 39 |
-
- seed: 42
|
| 40 |
-
- gradient_accumulation_steps: 32
|
| 41 |
-
- total_train_batch_size: 128
|
| 42 |
-
- optimizer: Use OptimizerNames.ADAFACTOR and the args are:
|
| 43 |
-
No additional optimizer arguments
|
| 44 |
-
- lr_scheduler_type: linear
|
| 45 |
-
- lr_scheduler_warmup_steps: 40
|
| 46 |
-
- training_steps: 600
|
| 47 |
-
- mixed_precision_training: Native AMP
|
| 48 |
-
|
| 49 |
-
### Training results
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
### Framework versions
|
| 54 |
-
|
| 55 |
-
- Transformers 5.20.2
|
| 56 |
-
- Pytorch 2.11.0+cu130
|
| 57 |
-
- Datasets 4.8.5
|
| 58 |
-
- Tokenizers 0.23.3
|
|
|
|
| 1 |
---
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
tags:
|
| 6 |
+
- seb-ai
|
| 7 |
+
- text-generation
|
| 8 |
+
- causal-lm
|
| 9 |
+
- llama
|
| 10 |
+
- trained-from-scratch
|
| 11 |
+
pipeline_tag: text-generation
|
| 12 |
---
|
| 13 |
|
| 14 |
+
# Seb AI S1 Mini Instruct (409M)
|
|
|
|
| 15 |
|
| 16 |
+
**Seb AI S1 Mini Instruct** is an open-source 409M parameter instruction-tuned language model trained completely from scratch by **Seb AI Inc.**
|
| 17 |
|
| 18 |
+
- **Creator:** Seb AI Inc.
|
| 19 |
+
- **Licence:** Apache 2.0 (Open-source for anyone to download, use, and fine-tune)
|
| 20 |
|
| 21 |
+
## Model Details
|
| 22 |
+
- **Parameters:** 409 Million
|
| 23 |
+
- **Architecture:** Llama-style Causal Language Model (`LlamaForCausalLM`)
|
| 24 |
+
- **Context Length:** 512 tokens
|
| 25 |
+
- **Training:** Pre-trained from scratch and full-weight instruction-tuned across multi-turn conversational datasets, casual dialogue, and custom identity data by Seb AI Inc.
|
| 26 |
|
| 27 |
+
## Prompt Format
|
| 28 |
+
```text
|
| 29 |
+
User: Who made you?
|
| 30 |
+
Assistant:
|
| 31 |
+
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|