File size: 3,481 Bytes
ac9c094 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 |
---
---
[<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>
axolotl version: `0.9.2`
```yaml
base_model: /capstor/scratch/cscs/bbernath/models/meditron-70B
chat_template: llama3
bfloat16: true
output_dir: /capstor/store/cscs/swissai/a06/meditron/models/meditron_CHUV_2 #/capstor/scratch/cscs/bbernath/models/meditron_CHUV
dataset_prepared_path: /capstor/scratch/cscs/bbernath/dataset/
# - path: /capstor/store/cscs/swissai/a06/meditron/datasets/masked/special_mixture/instruction_tuning_mixture.jsonl
# type: chat_template
# ds_type: json
# split: train
# field_messages: conversations
# message_field_role: from
# message_field_content: value
#pretraining_dataset:
# - path: json
# data_files:
# - /capstor/store/cscs/swissai/a06/meditron/datasets/pretrain/pubmed/pubmed_3B.jsonl
# - /capstor/store/cscs/swissai/a06/meditron/datasets/pretrain/fineweb/fineweb_400M_anglais.jsonl
# type: pretrain
datasets:
- path: /capstor/store/cscs/swissai/a06/meditron/datasets/masked/gemini/moove_gemini_2.jsonl
type: chat_template
ds_type: json
split: train
field_messages: conversations
message_field_role: from
message_field_content: value
shuffle_merged_datasets: true
dataset_processes: 128
# max_steps: 1500
flash_attention: true
sequence_len: 8192
gradient_accumulation_steps: 1
micro_batch_size: 1
train_on_inputs: false
group_by_length: false
pad_to_sequence_len: true
sample_packing: true
optimizer: adamw_torch
optim_args:
fused: true
cosine_min_lr_ratio: 0.1
learning_rate: 5.0e-6
warmup_ratio: 0
weight_decay: 0.05
gradient_checkpointing: true
gradient_checkpointing_kwargs:
use_reentrant: false
load_in_4bit: false
load_in_8bit: false
num_epochs: 1
saves_per_epoch: 1
# evals_per_epoch: 1
eval_set_size: 0.0
eval_table_size: null
lr_scheduler: cosine
max_grad_norm: 1.0
resume_from_checkpoint: null
special_tokens:
pad_token: <|end_of_text|>
tf32: false
tokenizer_type: AutoTokenizer
type: LlamaForCausalLM
flash_attn_rms_norm: true
flash_attn_fuse_qkv: false
early_stopping_patience: 0
wandb_entity: alexs-team
wandb_name: meditron-CHUV-llama-gemini
wandb_project: Meditron DDX
wandb_watch: gradients
xformers_attention: null
logging_steps: 1
deepspeed: /capstor/users/cscs/bbernath/meditron/axolotl_config/deepspeed_new.json
```
</details><br>
# capstor/store/cscs/swissai/a06/meditron/models/meditron_CHUV_2
This model was trained from scratch on the /capstor/store/cscs/swissai/a06/meditron/datasets/masked/gemini/moove_gemini_2.jsonl 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: 5e-06
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 32
- total_train_batch_size: 32
- total_eval_batch_size: 32
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=fused=True
- lr_scheduler_type: cosine
- num_epochs: 1.0
### Training results
### Framework versions
- Transformers 4.51.3
- Pytorch 2.7.0a0+79aa17489c.nv25.04
- Datasets 3.6.0
- Tokenizers 0.21.1
|