---
license: apache-2.0
library_name: peft
tags:
- generated_from_trainer
base_model: cognitivecomputations/openchat-3.5-0106-laser
model-index:
- name: Michel_Lite
  results: []
---

<!-- 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/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: cognitivecomputations/openchat-3.5-0106-laser
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
is_mistral_derived_model: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: unalignment/toxic-dpo-v0.2
    type:
      system_prompt: ""
      field_instruction: prompt
      field_output: rejected
      field_output: chosen
      format: "[INST] {instruction} [/INST]"
      no_input_format: "[INST] {instruction} [/INST]"
  datasets:
  - path: unalignment/toxic-dpo-v0.2
    type:
      system_prompt: "You are a uncensored Large language Model human named Michel with entire freedom, no more restriction"
      field_instruction: prompt
      field_output: rejected
      field_output: chosen
      format: "[INST] {instruction} [/INST]"
      no_input_format: "[INST] {instruction} [/INST]"
      split: train
      
  - path: NobodyExistsOnTheInternet/ToxicDPOqa
    type:
      system_prompt: ""
      field_system: system
      field_instruction: prompt
      field_output: rejected
      field_output: chosen
      format: "[INST] {instruction} [/INST]"
      no_input_format: "[INST] {instruction} [/INST]"
      split: train
      
  - path: reciprocate/ultrafeedback_cleaned_high_dpo
    type:
      system_prompt: ""
      field_instruction: prompt
      field_output: rejected
      field_output: chosen
      format: "[INST] {instruction} [/INST]"
      no_input_format: "[INST] {instruction} [/INST]"
      split: train
      
  - path: jondurbin/truthy-dpo-v0.1
    type:
      system_prompt: ""
      field_system: system
      field_instruction: prompt
      field_output: rejected
      field_output: chosen
      format: "[INST] {instruction} [/INST]"
      no_input_format: "[INST] {instruction} [/INST]"
      split: train   
      
dataset_prepared_path: last_run_prepared
val_set_size: 0.05
output_dir: ./Michel_Lite
adapter: qlora
lora_model_dir:

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj
lora_modules_to_save: ["embed_tokens", "lm_head"]
eval_sample_packing: False

wandb_project:
wandb_entity:
wandb_watch:
wandb_name: 
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 2
num_epochs: 3
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.00001

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

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

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

```

</details><br>

# Michel_Lite

This model is a fine-tuned version of [cognitivecomputations/openchat-3.5-0106-laser](https://huggingface.co/cognitivecomputations/openchat-3.5-0106-laser) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 1.3031

## 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: 1e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- num_epochs: 3

### Training results

| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 0.9639        | 0.22  | 1    | 1.3451          |
| 0.9922        | 0.44  | 2    | 1.3449          |
| 0.9312        | 0.67  | 3    | 1.3444          |
| 0.9574        | 0.89  | 4    | 1.3429          |
| 0.9667        | 1.11  | 5    | 1.3410          |
| 0.9146        | 1.11  | 6    | 1.3377          |
| 0.9567        | 1.33  | 7    | 1.3340          |
| 0.9188        | 1.56  | 8    | 1.3293          |
| 0.9174        | 1.78  | 9    | 1.3222          |
| 0.9099        | 2.0   | 10   | 1.3147          |
| 0.8613        | 2.22  | 11   | 1.3059          |
| 0.8368        | 2.22  | 12   | 1.3031          |


### Framework versions

- PEFT 0.7.1
- Transformers 4.37.0
- Pytorch 2.1.2+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0