macadeliccc/Samantha-Qwen2-1.5B-QLoRa

Samantha Qwen2 1.5B

This model was trained on 2xL40S using FSDP and QLoRa. FP16 Merge is available here

Prompt Template

<|im_start|>system
You are a helpful AI assistant<|im_end|>
<|im_start|>user
What is the capital of France?<|im_end|>
<|im_start|>assistant

Launch Using VLLM

python -m vllm.entrypoints.openai.api_server \
    --model macadeliccc/Samantha-Qwen2-1.5B \
    --chat-template ./examples/template_chatml.jinja \

from openai import OpenAI
# Set OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://localhost:8000/v1"

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

chat_response = client.chat.completions.create(
    model="macadeliccc/Samantha-Qwen-2-1.5B",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Tell me a joke."},
    ]
)
print("Chat response:", chat_response)

Quants

TODO

Config

See axolotl config

axolotl version: 0.4.0

base_model: Qwen/Qwen2-1.5B
trust_remote_code: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: macadeliccc/opus_samantha
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: uncensored_ultrachat_20k_sharegpt.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: flattened_openhermes_200k.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: opus_instruct.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: airoboros_uncensored.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: orca_math_word_problems_sharegpt.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: sharegpt_starcoder.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: samantha_1.1_uncensored.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: samantha_1.5.json
    type: sharegpt
    field: conversations
    conversation: chatml
  - path: json
    data_files: sharegpt_hitchhikers_v1.json
    type: sharegpt
    field: conversations
    conversation: chatml


chat_template: chatml


dataset_prepared_path:
val_set_size: 0.05
output_dir: ./outputs/out

sequence_len: 4096
sample_packing: true
eval_sample_packing: true
pad_to_sequence_len: true

adapter: qlora
lora_model_dir:
lora_r: 32
lora_alpha: 64
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 3
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: true

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 10
evals_per_epoch: 4
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
  - full_shard
  - auto_wrap
fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: true
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: true
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_transformer_layer_cls_to_wrap: Qwen2DecoderLayer
  fsdp_state_dict_type: FULL_STATE_DICT
special_tokens: