A RP/storywriting specialist model, full-parameter finetune of Qwen2.5-32B on mixture of synthetic and natural data.
It uses Celeste 70B 0.1 data mixture, greatly expanding it to improve versatility, creativity and "flavor" of the resulting model.

Note: using quantized KV cache with Qwen2.5 is not recommended and can lead to degraded output quality. On the other hand, Qwen's KV cache is already light enough, so using f16 for it shouldn't be problematic.

Prompt format is ChatML.

Recommended sampler values:

Temperature: 1
Typical-P: 0.9
Min-P: 0.05
Top-A: 0.2
Repetition Penalty: 1.03

Recommended SillyTavern presets (via CalamitousFelicitousness):

Training data:

Celeste 70B 0.1 data mixture minus Opus Instruct subset. See that model's card for details.
Kalomaze's Opus_Instruct_25k dataset, filtered for refusals.
A subset (1k rows) of ChatGPT-4o-WritingPrompts by Gryphe
A subset (2k rows) of Sonnet3.5-Charcards-Roleplay by Gryphe
Synthstruct and SynthRP datasets by Epiculous

Training time and hardware:

7 hours on 8xH100 SXM, provided by FeatherlessAI

Model was trained by Kearm and Auri.

Special thanks:

to FeatherlessAI for generously providing 8xH100 SXM node for training of this model
to Gryphe, Lemmy, Kalomaze, Nopm and Epiculous for the data
and to Allura-org for support and feedback on EVA models.

See axolotl config

axolotl version: 0.4.1

yaml

base_model: Qwen/Qwen2.5-32B

load_in_8bit: false
load_in_4bit: false
strict: false

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_swiglu: true
liger_fused_linear_cross_entropy: true

# plugins:
#   - axolotl.integrations.spectrum.SpectrumPlugin

# spectrum_top_fraction: 0.5
# # Optional if using a pre-scanned model as your base_model. Useful if using a model mirror
# spectrum_model_name: Qwen/Qwen2.5-32B

datasets:
  - path: datasets/deduped_Synthstruct-Gens_processed_sharegpt_converted_cleaned.jsonl
    type: sharegpt
  - path: datasets/opus-instruct-22k-no_refusals-filtered.jsonl
    type: sharegpt
  - path: datasets/Celeste_Filtered.jsonl
    type: sharegpt
  - path: datasets/Gryphe-S3-5-Charcards-names-2k.jsonl
    type: sharegpt
  - path: datasets/deduped_SynthRP-Gens_processed_09-25-2024-ShareGPT_converted_cleaned.jsonl
    type: sharegpt
  - path: datasets/deduped_Gryphe-4o-WP-1k.jsonl
    type: sharegpt
  - path: datasets/deduped_not_samantha_norefusals.jsonl
    type: sharegpt

chat_template: chatml
shuffle_merged_datasets: true
val_set_size: 0.001
output_dir: ./EVA-Qwen2.5-32B-SFFT-v0.0

sequence_len: 8192
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true

# adapter: qlora
# lora_model_dir:
# lora_r: 64
# lora_alpha: 64
# lora_dropout: 0.05
# lora_target_linear: true
# peft_use_dora: true

unfrozen_parameters:
- ^lm_head.weight$
- ^model.embed_tokens.weight$
# input_layernorm layers
- model.layers.0.input_layernorm
- model.layers.1.input_layernorm
- model.layers.2.input_layernorm
- model.layers.3.input_layernorm
- model.layers.4.input_layernorm
- model.layers.5.input_layernorm
- model.layers.6.input_layernorm
- model.layers.7.input_layernorm
- model.layers.8.input_layernorm
- model.layers.9.input_layernorm
- model.layers.10.input_layernorm
- model.layers.11.input_layernorm
- model.layers.12.input_layernorm
- model.layers.13.input_layernorm
- model.layers.14.input_layernorm
- model.layers.15.input_layernorm
- model.layers.16.input_layernorm
- model.layers.17.input_layernorm
- model.layers.18.input_layernorm
- model.layers.19.input_layernorm
- model.layers.20.input_layernorm
- model.layers.21.input_layernorm
- model.layers.22.input_layernorm
- model.layers.23.input_layernorm
- model.layers.24.input_layernorm
- model.layers.25.input_layernorm
- model.layers.26.input_layernorm
- model.layers.27.input_layernorm
- model.layers.28.input_layernorm
- model.layers.29.input_layernorm
- model.layers.30.input_layernorm
- model.layers.31.input_layernorm
# lm_head layers
# mlp.down_proj layers
- model.layers.63.mlp.down_proj
- model.layers.49.mlp.down_proj
- model.layers.48.mlp.down_proj
- model.layers.45.mlp.down_proj
- model.layers.44.mlp.down_proj
- model.layers.47.mlp.down_proj
- model.layers.46.mlp.down_proj
- model.layers.43.mlp.down_proj
- model.layers.8.mlp.down_proj
- model.layers.11.mlp.down_proj
- model.layers.19.mlp.down_proj
- model.layers.35.mlp.down_proj
- model.layers.20.mlp.down_proj
- model.layers.52.mlp.down_proj
- model.layers.39.mlp.down_proj
- model.layers.62.mlp.down_proj
- model.layers.50.mlp.down_proj
- model.layers.29.mlp.down_proj
- model.layers.16.mlp.down_proj
- model.layers.28.mlp.down_proj
- model.layers.53.mlp.down_proj
- model.layers.30.mlp.down_proj
- model.layers.31.mlp.down_proj
- model.layers.32.mlp.down_proj
- model.layers.7.mlp.down_proj
- model.layers.36.mlp.down_proj
- model.layers.12.mlp.down_proj
- model.layers.18.mlp.down_proj
- model.layers.37.mlp.down_proj
- model.layers.38.mlp.down_proj
- model.layers.14.mlp.down_proj
- model.layers.13.mlp.down_proj
# mlp.gate_proj layers
- model.layers.43.mlp.gate_proj
- model.layers.61.mlp.gate_proj
- model.layers.60.mlp.gate_proj
- model.layers.44.mlp.gate_proj
- model.layers.62.mlp.gate_proj
- model.layers.28.mlp.gate_proj
- model.layers.29.mlp.gate_proj
- model.layers.45.mlp.gate_proj
- model.layers.37.mlp.gate_proj
- model.layers.35.mlp.gate_proj
- model.layers.59.mlp.gate_proj
- model.layers.36.mlp.gate_proj
- model.layers.30.mlp.gate_proj
- model.layers.48.mlp.gate_proj
- model.layers.38.mlp.gate_proj
- model.layers.27.mlp.gate_proj
- model.layers.31.mlp.gate_proj
- model.layers.39.mlp.gate_proj
- model.layers.34.mlp.gate_proj
- model.layers.58.mlp.gate_proj
- model.layers.33.mlp.gate_proj
- model.layers.26.mlp.gate_proj
- model.layers.32.mlp.gate_proj
- model.layers.46.mlp.gate_proj
- model.layers.42.mlp.gate_proj
- model.layers.49.mlp.gate_proj
- model.layers.57.mlp.gate_proj
- model.layers.50.mlp.gate_proj
- model.layers.47.mlp.gate_proj
- model.layers.56.mlp.gate_proj
- model.layers.63.mlp.gate_proj
- model.layers.55.mlp.gate_proj
# mlp.up_proj layers
- model.layers.61.mlp.up_proj
- model.layers.60.mlp.up_proj
- model.layers.32.mlp.up_proj
- model.layers.59.mlp.up_proj
- model.layers.58.mlp.up_proj
- model.layers.57.mlp.up_proj
- model.layers.44.mlp.up_proj
- model.layers.28.mlp.up_proj
- model.layers.35.mlp.up_proj
- model.layers.36.mlp.up_proj
- model.layers.31.mlp.up_proj
- model.layers.34.mlp.up_proj
- model.layers.55.mlp.up_proj
- model.layers.29.mlp.up_proj
- model.layers.49.mlp.up_proj
- model.layers.30.mlp.up_proj
- model.layers.53.mlp.up_proj
- model.layers.43.mlp.up_proj
- model.layers.56.mlp.up_proj
- model.layers.33.mlp.up_proj
- model.layers.54.mlp.up_proj
- model.layers.62.mlp.up_proj
- model.layers.27.mlp.up_proj
- model.layers.51.mlp.up_proj
- model.layers.52.mlp.up_proj
- model.layers.37.mlp.up_proj
- model.layers.45.mlp.up_proj
- model.layers.26.mlp.up_proj
- model.layers.42.mlp.up_proj
- model.layers.50.mlp.up_proj
- model.layers.48.mlp.up_proj
- model.layers.39.mlp.up_proj
# model.embed_tokens layers
# model.norm layers
# post_attention_layernorm layers
- model.layers.0.post_attention_layernorm
- model.layers.1.post_attention_layernorm
- model.layers.2.post_attention_layernorm
- model.layers.3.post_attention_layernorm
- model.layers.4.post_attention_layernorm
- model.layers.5.post_attention_layernorm
- model.layers.6.post_attention_layernorm
- model.layers.7.post_attention_layernorm
- model.layers.8.post_attention_layernorm
- model.layers.9.post_attention_layernorm
- model.layers.10.post_attention_layernorm
- model.layers.11.post_attention_layernorm
- model.layers.12.post_attention_layernorm
- model.layers.13.post_attention_layernorm
- model.layers.14.post_attention_layernorm
- model.layers.15.post_attention_layernorm
- model.layers.16.post_attention_layernorm
- model.layers.17.post_attention_layernorm
- model.layers.18.post_attention_layernorm
- model.layers.19.post_attention_layernorm
- model.layers.20.post_attention_layernorm
- model.layers.21.post_attention_layernorm
- model.layers.22.post_attention_layernorm
- model.layers.23.post_attention_layernorm
- model.layers.24.post_attention_layernorm
- model.layers.25.post_attention_layernorm
- model.layers.26.post_attention_layernorm
- model.layers.27.post_attention_layernorm
- model.layers.28.post_attention_layernorm
- model.layers.29.post_attention_layernorm
- model.layers.30.post_attention_layernorm
- model.layers.31.post_attention_layernorm
# self_attn.k_proj layers
- model.layers.63.self_attn.k_proj
- model.layers.55.self_attn.k_proj
- model.layers.60.self_attn.k_proj
- model.layers.7.self_attn.k_proj
- model.layers.12.self_attn.k_proj
- model.layers.13.self_attn.k_proj
- model.layers.57.self_attn.k_proj
- model.layers.29.self_attn.k_proj
- model.layers.14.self_attn.k_proj
- model.layers.51.self_attn.k_proj
- model.layers.53.self_attn.k_proj
- model.layers.54.self_attn.k_proj
- model.layers.22.self_attn.k_proj
- model.layers.61.self_attn.k_proj
- model.layers.18.self_attn.k_proj
- model.layers.30.self_attn.k_proj
- model.layers.9.self_attn.k_proj
- model.layers.24.self_attn.k_proj
- model.layers.23.self_attn.k_proj
- model.layers.25.self_attn.k_proj
- model.layers.10.self_attn.k_proj
- model.layers.58.self_attn.k_proj
- model.layers.56.self_attn.k_proj
- model.layers.15.self_attn.k_proj
- model.layers.32.self_attn.k_proj
- model.layers.28.self_attn.k_proj
- model.layers.8.self_attn.k_proj
- model.layers.59.self_attn.k_proj
- model.layers.11.self_attn.k_proj
- model.layers.48.self_attn.k_proj
- model.layers.16.self_attn.k_proj
- model.layers.50.self_attn.k_proj
# self_attn.o_proj layers
- model.layers.15.self_attn.o_proj
- model.layers.23.self_attn.o_proj
- model.layers.31.self_attn.o_proj
- model.layers.30.self_attn.o_proj
- model.layers.18.self_attn.o_proj
- model.layers.24.self_attn.o_proj
- model.layers.17.self_attn.o_proj
- model.layers.28.self_attn.o_proj
- model.layers.34.self_attn.o_proj
- model.layers.33.self_attn.o_proj
- model.layers.25.self_attn.o_proj
- model.layers.12.self_attn.o_proj
- model.layers.14.self_attn.o_proj
- model.layers.29.self_attn.o_proj
- model.layers.16.self_attn.o_proj
- model.layers.26.self_attn.o_proj
- model.layers.22.self_attn.o_proj
- model.layers.27.self_attn.o_proj
- model.layers.35.self_attn.o_proj
- model.layers.20.self_attn.o_proj
- model.layers.13.self_attn.o_proj
- model.layers.36.self_attn.o_proj
- model.layers.19.self_attn.o_proj
- model.layers.37.self_attn.o_proj
- model.layers.21.self_attn.o_proj
- model.layers.11.self_attn.o_proj
- model.layers.54.self_attn.o_proj
- model.layers.5.self_attn.o_proj
- model.layers.38.self_attn.o_proj
- model.layers.6.self_attn.o_proj
- model.layers.8.self_attn.o_proj
- model.layers.9.self_attn.o_proj
# self_attn.q_proj layers
- model.layers.1.self_attn.q_proj
- model.layers.2.self_attn.q_proj
- model.layers.3.self_attn.q_proj
- model.layers.45.self_attn.q_proj
- model.layers.54.self_attn.q_proj
- model.layers.35.self_attn.q_proj
- model.layers.48.self_attn.q_proj
- model.layers.61.self_attn.q_proj
- model.layers.52.self_attn.q_proj
- model.layers.50.self_attn.q_proj
- model.layers.60.self_attn.q_proj
- model.layers.56.self_attn.q_proj
- model.layers.58.self_attn.q_proj
- model.layers.42.self_attn.q_proj
- model.layers.59.self_attn.q_proj
- model.layers.44.self_attn.q_proj
- model.layers.55.self_attn.q_proj
- model.layers.57.self_attn.q_proj
- model.layers.41.self_attn.q_proj
- model.layers.36.self_attn.q_proj
- model.layers.39.self_attn.q_proj
- model.layers.4.self_attn.q_proj
- model.layers.43.self_attn.q_proj
- model.layers.34.self_attn.q_proj
- model.layers.46.self_attn.q_proj
- model.layers.49.self_attn.q_proj
- model.layers.40.self_attn.q_proj
- model.layers.25.self_attn.q_proj
- model.layers.51.self_attn.q_proj
- model.layers.17.self_attn.q_proj
- model.layers.37.self_attn.q_proj
- model.layers.53.self_attn.q_proj
# self_attn.v_proj layers
- model.layers.55.self_attn.v_proj
- model.layers.31.self_attn.v_proj
- model.layers.47.self_attn.v_proj
- model.layers.45.self_attn.v_proj
- model.layers.49.self_attn.v_proj
- model.layers.48.self_attn.v_proj
- model.layers.15.self_attn.v_proj
- model.layers.30.self_attn.v_proj
- model.layers.7.self_attn.v_proj
- model.layers.44.self_attn.v_proj
- model.layers.29.self_attn.v_proj
- model.layers.51.self_attn.v_proj
- model.layers.50.self_attn.v_proj
- model.layers.14.self_attn.v_proj
- model.layers.54.self_attn.v_proj
- model.layers.32.self_attn.v_proj
- model.layers.43.self_attn.v_proj
- model.layers.10.self_attn.v_proj
- model.layers.46.self_attn.v_proj
- model.layers.38.self_attn.v_proj
- model.layers.57.self_attn.v_proj
- model.layers.22.self_attn.v_proj
- model.layers.39.self_attn.v_proj
- model.layers.6.self_attn.v_proj
- model.layers.23.self_attn.v_proj
- model.layers.58.self_attn.v_proj
- model.layers.53.self_attn.v_proj
- model.layers.40.self_attn.v_proj
- model.layers.24.self_attn.v_proj
- model.layers.9.self_attn.v_proj
- model.layers.25.self_attn.v_proj
- model.layers.5.self_attn.v_proj


wandb_project: EVA-Qwen2.5-32B-SFFT-v0.0
wandb_entity:
wandb_watch:
wandb_name: Unit-00
wandb_log_model:

gradient_accumulation_steps: 8
micro_batch_size: 1
num_epochs: 3
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 0.00003
max_grad_norm: 3

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

gradient_checkpointing: "unsloth"
# gradient_checkpointing_kwargs:
#   use_reentrant: true
early_stopping_patience:
resume_from_checkpoint: 
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 20
evals_per_epoch: 4
saves_per_epoch: 2
save_safetensors: true
hub_model_id: 
hub_strategy: 
debug:
deepspeed: deepspeed_configs/zero3_bf16.json
weight_decay: 0.1
# fsdp:
#   - full_shard
#   - auto_wrap
# fsdp_config:
#   fsdp_limit_all_gathers: true
#   fsdp_sync_module_states: true
#   fsdp_offload_params: false  # Changed from true
#   fsdp_use_orig_params: true  # Changed from false
#   fsdp_cpu_ram_efficient_loading: true
#   fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
#   fsdp_transformer_layer_cls_to_wrap: Qwen2DecoderLayer
#   fsdp_activation_checkpointing: true
#   fsdp_state_dict_type: SHARDED_STATE_DICT  # Changed from FULL_STATE_DICT
#   fsdp_sharding_strategy: FULL_SHARD
#   fsdp_forward_prefetch: true  # Added
#   fsdp_backward_prefetch: "BACKWARD_POST"  # Added
#   fsdp_backward_prefetch_limit: 1  # Added
#   fsdp_mixed_precision: BF16  # Added

SvdH/EVA-Qwen2.5-32B-v0.0-4.65bpw-h6-exl2

EVA Qwen2.5-32B v0.0

Recommended sampler values:

Recommended SillyTavern presets (via CalamitousFelicitousness):

Training data:

Training time and hardware:

Special thanks:

SvdH/
EVA-Qwen2.5-32B-v0.0-4.65bpw-h6-exl2