Valuoty-industry-plc-4B-preview

<!-- 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/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>

<details><summary>See axolotl config</summary>

axolotl version: 0.8.0.dev0

base_model: yasserrmd/qwen3-4b-agentic-reasoner
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

# QLoRA
adapter: lora
lora_r: 32
lora_alpha: 64
lora_dropout: 0.05
lora_target_modules: [q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj]

datasets:
- path: laurent-maille/pcl-test-S7
  type: chat_template
  field_messages: messages
  conversation: chat
dataset_prepared_path: ./prepared/plc_sharegpt
output_dir: ./outputs/valuoty-indus-plc-4B


# Séquences
sequence_len: 4096
sample_packing: true
cutoff_len: 2048           # <- 4096 si nécessaire, sinon 2048 conseillé
packing: true
pad_to_sequence_len: false
group_by_length: true
flash_attn_impl: fa2       # mets à "none" si FA2 non dispo

# Quantization / dtypes
load_in_4bit: true         # <- remplace load_in_8bit
load_in_8bit: false
bnb_4bit_quant_type: nf4
bnb_4bit_use_double_quant: true
bnb_4bit_compute_dtype: bfloat16
bf16: true
fp16: false

# Optim & training
optimizer: adamw_bnb_8bit
micro_batch_size: 1
per_device_train_batch_size: 8     # 8–16 à 2048 tokens sur A40 48Go
per_device_eval_batch_size: 8
gradient_accumulation_steps: 16    # effectif ≈ 128 (8×16); monte à 32 pour 256
gradient_checkpointing: true
max_grad_norm: 0.3
learning_rate: 1.0e-4
weight_decay: 0.0
lr_scheduler_type: cosine
warmup_ratio: 0.03
num_train_epochs: 2

# Eval / logs / save
evaluation_strategy: steps
eval_steps: 1000
logging_steps: 20
save_strategy: steps
save_steps: 1000
save_total_limit: 4
save_safetensors: true

</details><br>

outputs/valuoty-indus-plc-4B

This model is a fine-tuned version of yasserrmd/qwen3-4b-agentic-reasoner on the laurent-maille/pcl-test-S7 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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_BNB 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: 2
• num_epochs: 1.0

Training results

Framework versions

• PEFT 0.14.0
• Transformers 4.55.4
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0