estrogen/remnant-12b-ep2

See axolotl config

axolotl version: 0.8.0.dev0

# === Model Configuration ===
base_model: mistralai/Mistral-Nemo-Instruct-2407  # e.g. "mistralai/Mistral-Small-24B-Instruct-2501"
load_in_8bit: false
load_in_4bit: false

# === Training Setup ===
num_epochs: 2
micro_batch_size: 16
gradient_accumulation_steps: 1
sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

# === Hyperparameter Configuration ===
optimizer: apollo_adamw
# Apollo-mini configuration:
optim_args: "proj=random,rank=1,scale=128.0,scale_type=tensor,update_proj_gap=200"
# Regular Apollo configuration:
# optim_args: 
optim_target_modules: all_linear
learning_rate: 1e-5
lr_scheduler: rex
weight_decay: 0.01
warmup_ratio: 0.05

# === Data Configuration ===
datasets:
  - path: allura-org/inkmix-v3.0
    type: chat_template
    split: train
    field_messages: conversations
    message_field_role: from
    message_field_content: value

dataset_prepared_path: last_run_prepared
chat_template: jinja
chat_template_jinja: |
  {{- bos_token }}{%- for message in messages %}
  {%- if message['role'] == 'system' %}
  {{- '[SYSTEM_PROMPT]' + message['content'] + '[/SYSTEM_PROMPT]' }}
  {%- elif message['role'] == 'user' %}
  {{- '[INST]' + message['content'] + '[/INST]' }}
  {%- elif message['role'] == 'assistant' %}
  {{- message['content'] + eos_token }}
  {%- endif %}
  {%- endfor %}

# === Plugins ===
plugins:
  - axolotl.integrations.liger.LigerPlugin
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin

# === Hardware Optimization ===
gradient_checkpointing: unsloth
gradient_checkpointing_kwargs:
  use_reentrant: false
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
cut_cross_entropy: true
torch_compile: true
# Only if using multiple GPUs:
# deepspeed: [DEEPSPEED_CONFIG_PATH]  # e.g. "deepspeed_configs/zero3_bf16.json"

# === Wandb Tracking ===
wandb_project: nemo12b-inkmix-v3

# === Checkpointing ===
saves_per_epoch: 2
save_total_limit: 3

# === Advanced Settings ===
output_dir: offload
bf16: auto
flash_attention: true
train_on_inputs: false
group_by_length: false
logging_steps: 1
trust_remote_code: true
# nemo doesnt support system prompt ootb
tokens:
  - "[SYSTEM_PROMPT]"
  - "[/SYSTEM_PROMPT]"
special_tokens:
  pad_token: "<pad>"

offload

This model is a fine-tuned version of mistralai/Mistral-Nemo-Instruct-2407 on the allura-org/inkmix-v3.0 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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use apollo_adamw with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=proj=random,rank=1,scale=128.0,scale_type=tensor,update_proj_gap=200
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 88
• num_epochs: 2.0

Training results

Framework versions

• Transformers 4.50.0
• Pytorch 2.6.0+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1