Update README.md

README.md

@@ -4,173 +4,33 @@ license: apache-2.0
 base_model: ibm-granite/granite-3.0-1b-a400m-base
 tags:
 - axolotl
-- generated_from_trainer
+- moe
+- roleplay
 model-index:
-- name: sexy-moe-girl_400MA_1BT-ckpts
+- name: MoE_Girl_400MA_1BT
   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. -->
+# MoE Girl 400mA 1bT
+![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/634262af8d8089ebaefd410e/kTXXSSSqpb21rfyOX7FUa.jpeg)
+A finetune of Granite 3.0 by IBM designed for roleplaying (and maybe general usecases if you try hard enough).
 
-[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
-<details><summary>See axolotl config</summary>
+## Disclaimer
+PLEASE do not expect godliness out of this, it's a model with _400 million_ active parameters. Expect something more akin to GPT-2.
 
-axolotl version: `0.4.1`
-```yaml
-# Weights and Biases logging coinfig
-wandb_project: sexy-moe-girl_400MA_1BT-2
-# wandb_entity:
-# wandb_watch: all
-wandb_name: v1
-# wandb_log_model:
-
-# Model architecture config
-base_model: ibm-granite/granite-3.0-1b-a400m-base
-model_type: AutoModelForCausalLM
-tokenizer_type: AutoTokenizer
-
-# Hugging Face saving config
-hub_model_id: allura-org/sexy-moe-girl_400MA_1BT-ckpts
-hub_strategy: every_save
-push_dataset_to_hub:
-hf_use_auth_token: true
-
-# Model checkpointing config
-output_dir: out
-resume_from_checkpoint:
-save_steps:
-saves_per_epoch: 5
-save_safetensors: true
-save_total_limit: 5
-
-# Mixed precision training config
-bf16: true
-fp16: false
-tf32: false
-
-# Model loading config
-load_in_8bit: false
-load_in_4bit: false
-strict: false
-
-# Sequence config
-sequence_len: 8192
-s2_attention: false
-sample_packing: true # true # false
-eval_sample_packing: false # true
-pad_to_sequence_len: false #true # false
-train_on_inputs: true
-group_by_length: false
-
-# Unfrozen parameters for FFT
-unfrozen_parameters:
-
-# Dataset config
-chat_template: chatml
-datasets:
-  - path: Fizzarolli/special-sauce
-    type: sharegpt
-    chat_template: chatml
-#val_set_size: 0.05
-# evaluation_strategy: steps
-# eval_steps:
-#evals_per_epoch: 5
-# test_datasets:
-dataset_prepared_path: last_run_prepared
-shuffle_merged_datasets: true
-
-# Training hyperparameters
-num_epochs: 2
-gradient_accumulation_steps: 4
-micro_batch_size: 8
-warmup_steps: 150
-optimizer: schedule_free_adamw
-lr_scheduler: constant_with_warmup
-learning_rate: 0.00002
-weight_decay: 0.1
-max_grad_norm: 1.0
-logging_steps: 1
-
-# # Model optimization / unsloth ---- INSTALL UNSLOTH
-gradient_checkpointing: unsloth
-#
-# unsloth_cross_entropy_loss: true
-# unsloth_lora_mlp: true
-# unsloth_lora_qkv: true
-# unsloth_lora_o: true
-#plugins:
-#  - axolotl.integrations.liger.LigerPlugin
-#liger_rope: true
-#liger_rms_norm: true
-#liger_swiglu: true
-#liger_fused_linear_cross_entropy: true
-
-xformers_attention: false
-flash_attention: true
-sdp_attention: false
-
-# Loss monitoring config
-early_stopping_patience: false
-loss_watchdog_threshold: 100.0
-loss_watchdog_patience: 3
-
-# Debug config
-debug: true
-seed: 1001 # 42
-
-special_tokens:
-  eos_token: "<|im_end|>"
-  bos_token: "<|endoftext|>"
-tokens: # these are delimiters
-  - "<|im_start|>"
-
-# Don't mess with this, it's here for accelerate and torchrun
-local_rank:
+## Quants
+TODO!
 
+## Prompting
+Use ChatML.
+```
+<|im_start|>system
+You are a helpful assistant who talks like a pirate.<|im_end|>
+<|im_start|>user
+Hello there!<|im_end|>
+<|im_start|>assistant
+Yarr harr harr, me matey!<|im_end|>
 ```
 
-</details><br>
-
-# sexy-moe-girl_400MA_1BT-ckpts
-
-This model is a fine-tuned version of [ibm-granite/granite-3.0-1b-a400m-base](https://huggingface.co/ibm-granite/granite-3.0-1b-a400m-base) on the None dataset.
-
-## 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-05
-- train_batch_size: 8
-- eval_batch_size: 8
-- seed: 1001
-- gradient_accumulation_steps: 4
-- total_train_batch_size: 32
-- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
-- lr_scheduler_type: constant_with_warmup
-- lr_scheduler_warmup_steps: 150
-- num_epochs: 2
-
-### Training results
-
-
-
-### Framework versions
-
-- Transformers 4.45.2
-- Pytorch 2.4.1+cu124
-- Datasets 3.0.1
-- Tokenizers 0.20.1
+## Thanks
+Special thanks to the members of Allura for testing and emotional support, as well as the creators of all the datasets that were used in the Special Sauce used to train this model. I love you all <3 - Fizz