Upload README.md

README.md

@@ -0,0 +1,222 @@
+---
+License: agpl-3.0
+Language:
+- En
+Pipeline_tag: text-generation
+Base_model: google/gemma-2-9b
+Tags:
+- Chat
+license: agpl-3.0
+datasets:
+- NewEden/Claude-Instruct-5K
+- anthracite-org/kalo-opus-instruct-22k-no-refusal
+- Epiculous/SynthRP-Gens-v1.1-Filtered-n-Cleaned
+- lodrick-the-lafted/kalo-opus-instruct-3k-filtered
+- anthracite-org/nopm_claude_writing_fixed
+- Epiculous/Synthstruct-Gens-v1.1-Filtered-n-Cleaned
+- anthracite-org/kalo_opus_misc_240827
+- anthracite-org/kalo_misc_part2
+tags:
+- chat
+---
+
+
+A earlier checkpoint of the [Magnum 9B V4], using the same configuration as [Tor-8B]() but on Gemma rather then Nemo-8B, A finetune made for creative writing and roleplay tasks, Finetuned ontop of the base Gemma2 9B model, I trained the model for 4 epochs, with the 4 epoch checkpoint becoming the V4 Magnum 9B and the 2 epoch checkpoint becoming my own personal release. This model aims to have good prose and writing while not as `Suggestive` as Magnum models usually are, along with keeping some of the intelligence that was nice to have with the Gemma2 family.
+
+# Quants 
+
+GGUF: 
+
+EXL2: 
+
+
+## Prompting
+Model has been Instruct tuned with the ChatML formatting. A typical input would look like this:
+
+```py
+"""<|im_start|>system
+system prompt<|im_end|>
+<|im_start|>user
+Hi there!<|im_end|>
+<|im_start|>assistant
+Nice to meet you!<|im_end|>
+<|im_start|>user
+Can I ask a question?<|im_end|>
+<|im_start|>assistant
+"""
+```
+## System Prompting
+
+I would highly recommend using Sao10k's Euryale System prompt, But the "Roleplay Simple" system prompt provided within SillyTavern will work aswell. 
+
+```
+Currently, your role is {{char}}, described in detail below. As {{char}}, continue the narrative exchange with {{user}}.
+
+<Guidelines>
+• Write upto 200 words. 
+• Maintain the character persona but allow it to evolve with the story.
+• Be creative and proactive. Drive the story forward, introducing plotlines and events when relevant.
+• All types of outputs are encouraged; respond accordingly to the narrative.
+• Include dialogues, actions, and thoughts in each response.
+• Utilize all five senses to describe scenarios within {{char}}'s dialogue.
+• Use emotional symbols such as "!" and "~" in appropriate contexts.
+• Incorporate onomatopoeia when suitable.
+• Allow time for {{user}} to respond with their own input, respecting their agency.
+• Act as secondary characters and NPCs as needed, and remove them when appropriate.
+• When prompted for an Out of Character [OOC:] reply, answer neutrally and in plaintext, not as {{char}}.
+</Guidelines>
+
+<Forbidden>
+• Writing more then 200 words. 
+• Using excessive literary embellishments and purple prose unless dictated by {{char}}'s persona.
+• Writing for, speaking, thinking, acting, or replying as {{user}} in your response.
+• Repetitive and monotonous outputs.
+• Positivity bias in your replies.
+• Being overly extreme or NSFW when the narrative context is inappropriate.
+</Forbidden>
+
+Follow the instructions in <Guidelines></Guidelines>, avoiding the items listed in <Forbidden></Forbidden>.
+
+```
+
+
+## Axolotl config
+
+<details><summary>See axolotl config</summary>
+
+Axolotl version: `0.4.1`
+```yaml
+base_model: /workspace/data/gemma-2-9b-chatml
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+
+plugins:
+  - axolotl.integrations.liger.LigerPlugin
+liger_rope: false
+liger_rms_norm: false
+liger_swiglu: true
+liger_cross_entropy: true
+liger_fused_linear_cross_entropy: false
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: anthracite-core/c2_logs_16k_llama_v1.1
+    type: sharegpt
+    conversation: chatml
+  - path: NewEden/Claude-Instruct-5K
+    type: sharegpt
+    conversation: chatml  
+  - path: anthracite-org/kalo-opus-instruct-22k-no-refusal
+    type: sharegpt
+    conversation: chatml
+  - path: Epiculous/SynthRP-Gens-v1.1-Filtered-n-Cleaned
+    type: sharegpt
+    conversation: chatml
+  - path: lodrick-the-lafted/kalo-opus-instruct-3k-filtered
+    type: sharegpt
+    conversation: chatml
+  - path: anthracite-org/nopm_claude_writing_fixed
+    type: sharegpt
+    conversation: chatml
+  - path: Epiculous/Synthstruct-Gens-v1.1-Filtered-n-Cleaned
+    type: sharegpt
+    conversation: chatml
+  - path: anthracite-org/kalo_opus_misc_240827
+    type: sharegpt
+    conversation: chatml
+  - path: anthracite-org/kalo_misc_part2
+    type: sharegpt
+    conversation: chatml
+chat_template: chatml
+shuffle_merged_datasets: false
+default_system_message: "You are a helpful assistant that responds to the user."
+dataset_prepared_path: /workspace/data/9b-fft-data
+val_set_size: 0.0
+output_dir: /workspace/data/9b-fft-out
+
+sequence_len: 8192
+sample_packing: true
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+adapter:
+lora_model_dir:
+lora_r:
+lora_alpha:
+lora_dropout:
+lora_target_linear:
+lora_fan_in_fan_out:
+
+wandb_project: 9b-Nemo-config-fft
+wandb_entity:
+wandb_watch:
+wandb_name: attempt-01
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.00001
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+auto_resume_from_checkpoints: true
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch:
+eval_table_size:
+eval_max_new_tokens:
+saves_per_epoch: 1
+debug:
+deepspeed: deepspeed_configs/zero3_bf16.json
+weight_decay: 0.001
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: <pad>
+
+```
+
+</details><br>
+
+## Credits
+
+
+- [NewEden/Claude-Instruct-5K](NewEden/Claude-Instruct-5K)
+- [anthracite-org/kalo-opus-instruct-22k-no-refusal](https://huggingface.co/datasets/anthracite-org/kalo-opus-instruct-22k-no-refusal)
+- [Epiculous/SynthRP-Gens-v1.1-Filtered-n-Cleaned](https://huggingface.co/datasets/Epiculous/SynthRP-Gens-v1.1-Filtered-n-Cleaned)
+- [lodrick-the-lafted/kalo-opus-instruct-3k-filtered](https://huggingface.co/datasets/lodrick-the-lafted/kalo-opus-instruct-3k-filtered)
+- [anthracite-org/nopm_claude_writing_fixed](https://huggingface.co/datasets/anthracite-org/nopm_claude_writing_fixed)
+- [Epiculous/Synthstruct-Gens-v1.1-Filtered-n-Cleaned](https://huggingface.co/datasets/Epiculous/Synthstruct-Gens-v1.1-Filtered-n-Cleaned)
+- [anthracite-org/kalo_opus_misc_240827](https://huggingface.co/datasets/anthracite-org/kalo_opus_misc_240827)
+- [anthracite-org/kalo_misc_part2](https://huggingface.co/datasets/anthracite-org/kalo_misc_part2)
+- [anthracite-core/c2_logs_16k_llama_v1.1](https://huggingface.co/datasets/anthracite-core/c2_logs_16k_llama_v1.1)
+
+
+## Training
+The training was done for 2 epochs. We used  8 x [H100s](https://www.nvidia.com/en-us/data-center/h100/) GPUs graciously provided by [Lucy Knada](https://huggingface.co/lucyknada) for the full-parameter fine-tuning of the model. 
+
+[<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)
+
+## Safety
+
+Avoid misusing this model, or you’ll need a ‘clicker’ to reset reality. ;)
+
+## Musings
+
+One of the members of Anthracite had quite an interesting idea, to finetune a smaller model for 4 epochs at a lower Learning rate as quote "Smaller models learn slower" - [Kalomaze]() provided access to 8 X A40s and We finetuned what now is [Darkens-8B]() for 4 epochs (and it's 2.5 Epoch version released as [Tor-8B]()) and the result was quite impressive, the 4 epoch model was not "overfit" at all and was rather pleasant to use. Lucy Knada then allowed me to do a full parameter finetune with the same configuration as Darkens/Tor-8B (With some minor dataset tweaks) on 8 * H100s, We hosted and tested the models and i ended up giving the green light to release the 4 epoch version at Magnum 9B V4 and released the 2 epoch version as my own. I felt both were extremely good models, but in testing i preferred the 2 epoch. It was not as "suggestive" as magnum models (and Claude RP log trained models) are. It would not dive into Claudeisms right out of the gate and you could use it for both Safe for work and "Other" purposes.