README.md

---
tags:
- physics
- cosmology
model-index:
- name: cosmosage_qa
  results: []
license: mit
language:
- en
pipeline_tag: text-generation
base_model: mistralai/Mistral-7B-v0.1
---

# cosmosage

Cosmosage is a natural-language cosmology assistant that can answer questions about cosmology.

cosmosage_v2 first underwent continued pretraining based on thousands of papers and textbooks, 
and was subsequently fine-tuned on synthetically-generated question-answer pairs. It is a full
chat model, though it excels in Q&A mode, where the model gives a single answer in response to 
a single question.

The code used to generate cosmosage_v2 is available at https://github.com/tijmen/cosmosage

## Usage

After downloading cosmosage_v2, the following example code can be used to ask questions:

```python
path_to_model = 'cosmosage_v2/'

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
device = "cuda"
model = AutoModelForCausalLM.from_pretrained(path_to_model).to(device)
tokenizer = AutoTokenizer.from_pretrained(path_to_model)
def ask_cosmosage(question):
    input_ids = torch.cat([
    tokenizer.encode("You are cosmosage, an AI programmed to be a cosmology expert. You answer the USER's question clearly in long form, always providing context. When appropriate, provide a reference.", return_tensors="pt"),
    torch.tensor([[28705]]),
    tokenizer.encode("USER:", add_special_tokens=False, return_tensors="pt"),
    tokenizer.encode(question, add_special_tokens=False, return_tensors="pt"),
    torch.tensor([[28705]]),
    tokenizer.encode("ASSISTANT:", add_special_tokens=False, return_tensors="pt")
    ], dim=-1).to(device)
    generated_ids = model.generate(input_ids, max_length=input_ids.shape[1] + 1000, do_sample=True)
    return tokenizer.decode(generated_ids[0], skip_special_tokens=True)```

## Comparison to cosmosage_v1

cosmosage_v2 is a more knowledgeable model than cosmosage_v1 due to being pretrained on the papers and
textbooks, rather than just on synthetically generated QA pairs. However, it continues to struggle with 
_reliability_. While many of its answers are factually accurate, some are not. The outputs of cosmosage 
(or any LLM) should not be trusted to be factual.

[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
<details><summary>See axolotl config</summary>

axolotl version: `0.4.0`
```yaml
base_model: /workspace/output/cosmosage_base/
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
is_mistral_derived_model: true

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: /workspace/input/datasets/qa_tune/arxiv_metadata_qa3.jsonl
    type: sharegpt
  - path: /workspace/input/datasets/qa_tune/arxiv_refined_qa.jsonl
    type: sharegpt
  - path: /workspace/input/datasets/qa_tune/arxiv_summary3.jsonl
    type: sharegpt
  - path: /workspace/input/datasets/qa_tune/cosmology_qa.jsonl
    type: alpaca_chat.load_qa 
  - path: /workspace/input/datasets/qa_tune/openhermes2_5.jsonl
    type: sharegpt
  - path: /workspace/input/datasets/qa_tune/cosmology_textbooks_qa.jsonl
    type: alpaca_chat.load_qa 
  - path: /workspace/input/datasets/qa_tune/physics_astro_qa.jsonl
    type: alpaca_chat.load_qa 

dataset_prepared_path: /workspace/output/qa_tune_prepared
val_set_size: 0.001
output_dir: /workspace/output/cosmosage_qa

chat_template: inst

adapter: 
lora_model_dir:

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

lora_r:
lora_alpha:
lora_dropout:
lora_target_modules:
lora_target_linear: 
lora_fan_in_fan_out:

seed: 702

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 2.0
optimizer: adamw_torch
lr_scheduler: linear
learning_rate: 0.000002
max_grad_norm: 3.0

train_on_inputs: false
group_by_length: false
bf16: true
fp16: false 
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 100
eval_steps: 0.05
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 1
save_total_limit: 2
debug:
deepspeed: /workspace/axolotl/deepspeed_configs/zero1.json
weight_decay:
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

ddp_timeout: 7200000

```

</details><br>

# workspace/output/cosmosage_qa

This model was trained from scratch on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5673

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 2e-06
- train_batch_size: 4
- eval_batch_size: 4
- seed: 702
- distributed_type: multi-GPU
- num_devices: 4
- total_train_batch_size: 16
- total_eval_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 2.0

### Training results

| Training Loss | Epoch | Step  | Validation Loss |
|:-------------:|:-----:|:-----:|:---------------:|
| 1.1004        | 0.0   | 1     | 1.1450          |
| 0.7343        | 0.1   | 909   | 0.7093          |
| 0.697         | 0.2   | 1818  | 0.6630          |
| 0.6386        | 0.3   | 2727  | 0.6380          |
| 0.5687        | 0.4   | 3636  | 0.6212          |
| 0.5857        | 0.5   | 4545  | 0.6083          |
| 0.6161        | 0.6   | 5454  | 0.5986          |
| 0.522         | 0.7   | 6363  | 0.5894          |
| 0.5563        | 0.8   | 7272  | 0.5825          |
| 0.6176        | 0.9   | 8181  | 0.5766          |
| 0.5948        | 1.0   | 9090  | 0.5719          |
| 0.4269        | 1.08  | 9999  | 0.5817          |
| 0.4858        | 1.18  | 10908 | 0.5796          |
| 0.4909        | 1.28  | 11817 | 0.5765          |
| 0.4325        | 1.38  | 12726 | 0.5746          |
| 0.4037        | 1.48  | 13635 | 0.5720          |
| 0.507         | 1.58  | 14544 | 0.5706          |
| 0.4778        | 1.68  | 15453 | 0.5697          |
| 0.4599        | 1.78  | 16362 | 0.5683          |
| 0.4515        | 1.88  | 17271 | 0.5673          |


### Framework versions

- Transformers 4.38.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.17.0
- Tokenizers 0.15.0