ConicCat/Gemma4Test

See axolotl config

axolotl version: 0.16.0.dev0

base_model: google/gemma-4-31B-it

load_in_8bit: false
load_in_4bit: false

plugins:
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
  - axolotl.integrations.liger.LigerPlugin
torch_compile: false
liger_layer_norm: true
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_rms_norm_gated: true
strict: false

sequence_len: 2048
max_sample_length: 2048

flash_attention: false
sdp_attention: true

sample_packing: true
gradient_checkpointing: true
activation_offloading: true

bf16: true
tf32: true

lora_mlp_kernel: false
lora_qkv_kernel: false
lora_o_kernel: false

datasets:

  - path: ConicCat/Mura_Books
    type: chat_template

chat_template_jinja: >
    {%- macro strip_thinking(text) -%}
        {%- set ns = namespace(result='') -%}
        {%- for part in text.split('<channel|>') -%}
            {%- if '<|channel>' in part -%}
                {%- set ns.result = ns.result + part.split('<|channel>')[0] -%}
            {%- else -%}
                {%- set ns.result = ns.result + part -%}
            {%- endif -%}
        {%- endfor -%}
        {{- ns.result | trim -}}
    {%- endmacro -%}
    
    {%- set loop_messages = messages -%}
    {{ bos_token }}
    
    {#- Handle System Definitions Block -#}
    {%- if (enable_thinking is defined and enable_thinking) or messages[0]['role'] in ['system', 'developer'] -%}
        {{- '<|turn>system\n' -}}
    
        {#- Inject Thinking token at the very top of the FIRST system turn -#}
        {%- if enable_thinking is defined and enable_thinking -%}
            {{- '<|think|>' -}}
        {%- endif -%}
    
        {%- if messages[0]['role'] in ['system', 'developer'] -%}
            {{- messages[0]['content'] | trim -}}
            {%- set loop_messages = messages[1:] -%}
        {%- endif -%}
        
        {{- '<turn|>\n' -}}
    {%- endif %}
    
    {#- Loop through messages -#}
    {%- for message in loop_messages -%}
        {%- set role = 'model' if message['role'] == 'assistant' else message['role'] -%}
        {{- '<|turn>' + role + '\n' -}}
    
        {#- Flag to identify the final SFT turn -#}
        {%- set is_final_sft_turn = loop.last and not add_generation_prompt -%}
    
        {%- if message['content'] is string -%}
            {%- if role == 'model' -%}
                {%- if is_final_sft_turn and '<|channel>thought' not in message['content'] -%}
                    {{- '<|channel>thought\n<channel|>' -}}
                {%- endif -%}
                {{- strip_thinking(message['content']) -}}
            {%- else -%}
                {{- message['content'] | trim -}}
            {%- endif -%}
        {%- elif message['content'] is sequence -%}
            {%- set ns = namespace(has_thinking=false) -%}
            {%- for item in message['content'] -%}
                {%- if item['type'] == 'text' and '<|channel>thought' in item['text'] -%}
                    {%- set ns.has_thinking = true -%}
                {%- endif -%}
            {%- endfor -%}
            
            {%- if role == 'model' and is_final_sft_turn and not ns.has_thinking -%}
                {{- '<|channel>thought\n<channel|>' -}}
            {%- endif -%}
    
            {%- for item in message['content'] -%}
                {%- if item['type'] == 'text' -%}
                    {%- if role == 'model' -%}
                        {{- strip_thinking(item['text']) -}}
                    {%- else -%}
                        {{- item['text'] | trim -}}
                    {%- endif -%}
                {%- endif -%}
            {%- endfor -%}
        {%- endif -%}
    
        {{- '<turn|>\n' -}}
    {%- endfor -%}
    
    {#- Generation Prompt handled as normal (serves as the final turn when true) -#}
    {%- if add_generation_prompt -%}
        {{- '<|turn>model\n' -}}
        {%- if not enable_thinking | default(false) -%}
            {{- '<|channel>thought\n<channel|>' -}}
        {%- endif -%}
    {%- endif -%}


adapter: lora
lora_r: 32
lora_alpha: 64
lora_dropout: 0.0
lora_bias: None
lora_target_modules: 'model.language_model.layers.[\d]+.(_checkpoint_wrapped_module.)?(mlp|self_attn).(up|down|gate|q|k|v|o)_proj'
use_tensorboard: true

optimizer: paged_adamw_8bit
learning_rate: 2.5e-5 # 1e-4 / 4
loraplus_lr_ratio: 16

# Training arguments
output_dir: ./Writer-Stage-2
num_epochs: 11
micro_batch_size: 2
gradient_accumulation_steps: 4
save_strategy: 'no'
warmup_ratio: 0.05
lr_scheduler: 'cosine'
max_grad_norm: 1
logging_steps: 1
seed: 42

eot_tokens:
  - "<turn|>"

push_dataset_to_hub: ConicCat/Gemma4-Mura
hf_use_auth_token: true 

Writer-Stage-2

This model is a fine-tuned version of google/gemma-4-31B-it on the ConicCat/Mura_Books 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: 2.5e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 12
• training_steps: 242

Training results

Framework versions

• PEFT 0.19.1
• Transformers 5.5.0
• Pytorch 2.8.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2