Text Generation
PEFT
Safetensors
Transformers
qwen3_5
image-text-to-text
axolotl
lora
conversational
Instructions to use fornace-ml/eugeny-seed-adapters with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use fornace-ml/eugeny-seed-adapters with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.6-27B") model = PeftModel.from_pretrained(base_model, "fornace-ml/eugeny-seed-adapters") - Transformers
How to use fornace-ml/eugeny-seed-adapters with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="fornace-ml/eugeny-seed-adapters") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("fornace-ml/eugeny-seed-adapters") model = AutoModelForMultimodalLM.from_pretrained("fornace-ml/eugeny-seed-adapters", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use fornace-ml/eugeny-seed-adapters with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "fornace-ml/eugeny-seed-adapters" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "fornace-ml/eugeny-seed-adapters", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/fornace-ml/eugeny-seed-adapters
- SGLang
How to use fornace-ml/eugeny-seed-adapters with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "fornace-ml/eugeny-seed-adapters" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "fornace-ml/eugeny-seed-adapters", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "fornace-ml/eugeny-seed-adapters" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "fornace-ml/eugeny-seed-adapters", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use fornace-ml/eugeny-seed-adapters with Docker Model Runner:
docker model run hf.co/fornace-ml/eugeny-seed-adapters
Download debug.log from fornace-ml/eugeny-seed-adapters: direct link, hf CLI and curl.
- Browser
- Download file 33.5 kB
-
https://huggingface.co/fornace-ml/eugeny-seed-adapters/resolve/main/debug.log
- Command line
-
hf download hf://fornace-ml/eugeny-seed-adapters/debug.log
-
curl -L -o debug.log https://huggingface.co/fornace-ml/eugeny-seed-adapters/resolve/main/debug.log
33.5 kB
| [2026-06-24 12:52:11,377] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:37810] baseline 0.000GB () | |
| [2026-06-24 12:52:11,378] [INFO] [axolotl.cli.config.load_cfg:333] [PID:37810] config: | |
| { | |
| "activation_offloading": false, | |
| "adapter": "lora", | |
| "attn_needs_dtype_cast": false, | |
| "attn_supports_packing": false, | |
| "attn_uses_flash_lib": false, | |
| "axolotl_config_path": "fine-tuning/configs/axolotl_qwen36_27b_seed_sft.yml", | |
| "base_model": "Qwen/Qwen3.6-27B", | |
| "base_model_config": "Qwen/Qwen3.6-27B", | |
| "batch_size": 16, | |
| "bf16": true, | |
| "capabilities": { | |
| "bf16": true, | |
| "compute_capability": "sm_120", | |
| "fp8": true, | |
| "n_gpu": 1, | |
| "n_node": 1, | |
| "tf32": true | |
| }, | |
| "chat_template": "tokenizer_default", | |
| "context_parallel_size": 1, | |
| "dataloader_num_workers": 1, | |
| "dataloader_pin_memory": true, | |
| "dataloader_prefetch_factor": 256, | |
| "dataset_num_proc": 384, | |
| "dataset_prepared_path": "./fine-tuning/data/seed/axolotl_prepared_qwen36_27b_seed", | |
| "datasets": [ | |
| { | |
| "chat_template": "tokenizer_default", | |
| "field_messages": "messages", | |
| "message_property_mappings": { | |
| "content": "content", | |
| "role": "role" | |
| }, | |
| "path": "./fine-tuning/data/seed/sft_train.jsonl", | |
| "roles_to_train": [ | |
| "assistant" | |
| ], | |
| "train_on_eos": "turn", | |
| "trust_remote_code": false, | |
| "type": "chat_template" | |
| } | |
| ], | |
| "ddp": false, | |
| "device": "cuda:0", | |
| "dion_rank_fraction": 1.0, | |
| "dion_rank_multiple_of": 1, | |
| "eaft_alpha": 1.0, | |
| "eaft_k": 20, | |
| "env_capabilities": { | |
| "torch_version": "2.12.1" | |
| }, | |
| "eval_batch_size": 1, | |
| "eval_causal_lm_metrics": [ | |
| "sacrebleu", | |
| "comet", | |
| "ter", | |
| "chrf" | |
| ], | |
| "eval_max_new_tokens": 128, | |
| "eval_steps": 200, | |
| "eval_table_size": 0, | |
| "experimental_skip_move_to_device": true, | |
| "fp16": false, | |
| "generate_samples": false, | |
| "generation_do_sample": true, | |
| "generation_max_new_tokens": 50, | |
| "generation_prompt_ratio": 0.5, | |
| "generation_temperature": 0.7, | |
| "gradient_accumulation_steps": 16, | |
| "gradient_checkpointing": true, | |
| "gradient_checkpointing_kwargs": { | |
| "use_reentrant": true | |
| }, | |
| "hub_model_id": "fornace-ml/eugeny-seed-adapters", | |
| "hub_strategy": "every_save", | |
| "include_tkps": true, | |
| "is_falcon_derived_model": false, | |
| "is_llama_derived_model": false, | |
| "is_mistral_derived_model": false, | |
| "is_multimodal": true, | |
| "layer_offloading": false, | |
| "learning_rate": 0.0001, | |
| "lisa_layers_attribute": "model.layers", | |
| "load_best_model_at_end": false, | |
| "load_in_4bit": false, | |
| "load_in_8bit": false, | |
| "local_rank": 0, | |
| "logging_steps": 5, | |
| "lora_alpha": 128, | |
| "lora_dropout": 0.05, | |
| "lora_r": 64, | |
| "lora_target_linear": true, | |
| "loraplus_lr_embedding": 1e-06, | |
| "lr_scheduler": "cosine", | |
| "mean_resizing_embeddings": false, | |
| "merge_method": "memory_efficient", | |
| "micro_batch_size": 1, | |
| "model_config_type": "qwen3_5", | |
| "model_config_type_text": "qwen3_5_text", | |
| "num_epochs": 2.0, | |
| "num_generation_samples": 3, | |
| "optimizer": "adamw_torch", | |
| "otel_metrics_host": "localhost", | |
| "otel_metrics_port": 8000, | |
| "output_dir": "./outputs/qwen36-27b-agent-memory-sft-lora-axolotl", | |
| "pad_to_sequence_len": true, | |
| "pretrain_multipack_attn": true, | |
| "processor_config": "Qwen/Qwen3.6-27B", | |
| "profiler_steps_start": 0, | |
| "qgalore_cos_threshold": 0.4, | |
| "qgalore_gamma_proj": 2, | |
| "qgalore_proj_bits": 4, | |
| "qgalore_proj_group_size": 256, | |
| "qgalore_proj_quant": true, | |
| "qgalore_proj_type": "std", | |
| "qgalore_queue_size": 5, | |
| "qgalore_rank": 256, | |
| "qgalore_scale": 0.25, | |
| "qgalore_update_proj_gap": 200, | |
| "qlora_sharded_model_loading": false, | |
| "quantize_moe_experts": false, | |
| "ray_num_workers": 1, | |
| "relora_prune_method": "magnitude", | |
| "resources_per_worker": { | |
| "GPU": 1 | |
| }, | |
| "sample_packing": false, | |
| "sample_packing_bin_size": 200, | |
| "sample_packing_group_size": 100000, | |
| "save_only_model": false, | |
| "save_safetensors": true, | |
| "save_steps": 200, | |
| "save_total_limit": 2, | |
| "seed": 42, | |
| "sequence_len": 4096, | |
| "shuffle_before_merging_datasets": false, | |
| "shuffle_merged_datasets": true, | |
| "skip_prepare_dataset": false, | |
| "streaming_multipack_buffer_size": 10000, | |
| "strict": false, | |
| "tensor_parallel_size": 1, | |
| "test_datasets": [ | |
| { | |
| "chat_template": "tokenizer_default", | |
| "field_messages": "messages", | |
| "message_property_mappings": { | |
| "content": "content", | |
| "role": "role" | |
| }, | |
| "path": "./fine-tuning/data/seed/sft_eval.jsonl", | |
| "roles_to_train": [ | |
| "assistant" | |
| ], | |
| "split": "train", | |
| "train_on_eos": "turn", | |
| "trust_remote_code": false, | |
| "type": "chat_template" | |
| } | |
| ], | |
| "tf32": true, | |
| "tiled_mlp_use_original_mlp": true, | |
| "tokenizer_config": "Qwen/Qwen3.6-27B", | |
| "tokenizer_save_jinja_files": true, | |
| "tokenizer_type": "AutoTokenizer", | |
| "torch_dtype": "torch.bfloat16", | |
| "train_on_inputs": false, | |
| "trl": { | |
| "async_prefetch": false, | |
| "log_completions": false, | |
| "mask_truncated_completions": false, | |
| "ref_model_mixup_alpha": 0.9, | |
| "ref_model_sync_steps": 64, | |
| "replay_buffer_size": 0, | |
| "replay_recompute_logps": true, | |
| "reroll_max_groups": 1, | |
| "reroll_start_fraction": 1.0, | |
| "reward_num_workers": 1, | |
| "scale_rewards": true, | |
| "skip_zero_advantage_batches": true, | |
| "sync_ref_model": false, | |
| "use_data_producer": false, | |
| "use_vllm": false, | |
| "vllm_lora_sync": false, | |
| "vllm_server_host": "0.0.0.0", | |
| "vllm_server_port": 8000 | |
| }, | |
| "trust_remote_code": true, | |
| "type_of_model": "AutoModelForCausalLM", | |
| "use_otel_metrics": false, | |
| "use_ray": false, | |
| "use_wandb": true, | |
| "val_set_size": 0.0, | |
| "vllm": { | |
| "device": "auto", | |
| "dtype": "auto", | |
| "gpu_memory_utilization": 0.9, | |
| "host": "0.0.0.0", | |
| "port": 8000 | |
| }, | |
| "wandb_log_model": "checkpoint", | |
| "wandb_name": "eugeny-qwen36-27b-seed-sft", | |
| "wandb_project": "eugeny", | |
| "warmup_ratio": 0.03, | |
| "weight_decay": 0.01, | |
| "world_size": 1 | |
| } | |
| [2026-06-24 12:52:12,530] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:37810] EOS: 248046 / <|im_end|> | |
| [2026-06-24 12:52:12,530] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:37810] BOS: None / None | |
| [2026-06-24 12:52:12,530] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:37810] PAD: 248044 / <|endoftext|> | |
| [2026-06-24 12:52:12,530] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:37810] UNK: None / None | |
| [2026-06-24 12:52:12,531] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:477] [PID:37810] Loading prepared dataset from disk at fine-tuning/data/seed/axolotl_prepared_qwen36_27b_seed/ddaef6b4b116d96bae21a39214f33062... | |
| [2026-06-24 12:52:12,535] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:477] [PID:37810] Loading prepared dataset from disk at fine-tuning/data/seed/axolotl_prepared_qwen36_27b_seed/063e5073d5d03e6d44b00418133c74c4... | |
| [2026-06-24 12:52:12,539] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:420] [PID:37810] total_num_tokens: 77_178 | |
| [2026-06-24 12:52:12,540] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:438] [PID:37810] `total_supervised_tokens: 27_423` | |
| [2026-06-24 12:52:12,540] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:521] [PID:37810] total_num_steps: 45 | |
| [2026-06-24 12:52:12,540] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:37810] Maximum number of steps set at 45 | |
| [2026-06-24 12:52:12,587] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:37810] loading tokenizer... Qwen/Qwen3.6-27B | |
| [2026-06-24 12:52:13,600] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:37810] EOS: 248046 / <|im_end|> | |
| [2026-06-24 12:52:13,600] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:37810] BOS: None / None | |
| [2026-06-24 12:52:13,600] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:37810] PAD: 248044 / <|endoftext|> | |
| [2026-06-24 12:52:13,600] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:37810] UNK: None / None | |
| [2026-06-24 12:52:16,066] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:37810] Loading model | |
| [2026-06-24 12:52:16,147] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:37810] Patched OptimState8bit for torch.compile compatibility | |
| [2026-06-24 12:52:16,148] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:37810] Patched OptimState4bit for torch.compile compatibility | |
| [2026-06-24 12:52:16,148] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:37810] Patched OptimStateFp8 for torch.compile compatibility | |
| [2026-06-24 12:52:16,153] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:37810] Patched Trainer.evaluation_loop with nanmean loss calculation | |
| [2026-06-24 12:52:16,154] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:37810] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation | |
| [2026-06-24 12:52:16,160] [WARNING] [fla.utils.check_environments:63] [PID:37810] Current Python version 3.10 is below the recommended 3.11 version. It is recommended to upgrade to Python 3.11 or higher for the best experience. | |
| Loading weights: 0%| | 0/1184 [00:00<?, ?it/s] Loading weights: 39%|ββββ | 460/1184 [00:00<00:00, 4597.10it/s] Loading weights: 78%|ββββββββ | 920/1184 [00:00<00:00, 4567.21it/s] Loading weights: 100%|ββββββββββ| 1184/1184 [00:00<00:00, 4620.02it/s] | |
| [2026-06-24 12:52:21,952] [INFO] [axolotl.loaders.model._configure_embedding_dtypes:433] [PID:37810] Converting modules to torch.bfloat16 | |
| [2026-06-24 12:52:25,193] [DEBUG] [axolotl.loaders.model.log_gpu_memory_usage:127] [PID:37810] Memory usage after model load 0.000GB (+0.000GB allocated, +0.002GB reserved) | |
| [2026-06-24 12:52:25,217] [INFO] [axolotl.loaders.adapter._build_peft_lora_config:170] [PID:37810] found linear modules: ['down_proj', 'gate_proj', 'in_proj_a', 'in_proj_b', 'in_proj_qkv', 'in_proj_z', 'k_proj', 'linear_fc1', 'linear_fc2', 'o_proj', 'out_proj', 'proj', 'q_proj', 'qkv', 'up_proj', 'v_proj'] | |
| trainable params: 499,095,552 || all params: 27,855,824,112 || trainable%: 1.7917 | |
| [2026-06-24 12:52:28,890] [DEBUG] [axolotl.loaders.model.log_gpu_memory_usage:127] [PID:37810] after adapters 0.000GB () | |
| [2026-06-24 12:52:37,669] [INFO] [axolotl.train.save_initial_configs:450] [PID:37810] Pre-saving adapter config to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl... | |
| [2026-06-24 12:52:37,670] [INFO] [axolotl.train.save_initial_configs:454] [PID:37810] Pre-saving tokenizer to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl... | |
| [2026-06-24 12:52:37,804] [INFO] [axolotl.train.save_initial_configs:459] [PID:37810] Pre-saving model config to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl... | |
| [2026-06-24 12:52:37,812] [INFO] [axolotl.train.save_initial_configs:463] [PID:37810] Pre-saving processor to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl... | |
| [2026-06-24 12:52:37,930] [INFO] [axolotl.train.execute_training:226] [PID:37810] Starting trainer... | |
| wandb: Loading settings from /root/.config/wandb/settings | |
| wandb: [wandb.login()] Loaded credentials for https://wandb.fornace.net from WANDB_API_KEY. | |
| wandb: Currently logged in as: frappo to https://wandb.fornace.net. Use `wandb login --relogin` to force relogin | |
| wandb: Tracking run with wandb version 0.28.0 | |
| wandb: Run data is saved locally in /root/sky_workdir/fornace-eugeny/eugeny/wandb/run-20260624_125239-mv6s5u0t | |
| wandb: Run `wandb offline` to turn off syncing. | |
| wandb: Syncing run eugeny-qwen36-27b-seed-sft | |
| wandb: βοΈ View project at https://wandb.fornace.net/frappo/eugeny | |
| wandb: π View run at https://wandb.fornace.net/frappo/eugeny/runs/mv6s5u0t | |
| wandb: WARNING Saving files without folders. If you want to preserve subdirectories pass base_path to wandb.save, i.e. wandb.save("/mnt/folder/file.h5", base_path="/mnt") | |
| wandb: WARNING Symlinked 1 file into the W&B run directory; call wandb.save again to sync new files. | |
| [2026-06-24 12:52:42,117] [INFO] [axolotl.utils.callbacks.on_train_begin:807] [PID:37810] The Axolotl config has been saved to the WandB run under files. | |
| 0%| | 0/45 [00:00<?, ?it/s][2026-06-24 12:52:42,127] [INFO] [axolotl.core.trainers.base.evaluate:475] [PID:37810] Running evaluation step... | |
| 0%| | 0/30 [00:00<?, ?it/s][A | |
| 7%|β | 2/30 [00:00<00:11, 2.38it/s][A | |
| 10%|β | 3/30 [00:01<00:17, 1.52it/s][A | |
| 13%|ββ | 4/30 [00:02<00:20, 1.27it/s][A | |
| 17%|ββ | 5/30 [00:03<00:21, 1.17it/s][A | |
| 20%|ββ | 6/30 [00:04<00:21, 1.11it/s][A | |
| 23%|βββ | 7/30 [00:05<00:21, 1.07it/s][A | |
| 27%|βββ | 8/30 [00:06<00:20, 1.05it/s][A | |
| 30%|βββ | 9/30 [00:07<00:20, 1.03it/s][A | |
| 33%|ββββ | 10/30 [00:08<00:19, 1.02it/s][A | |
| 37%|ββββ | 11/30 [00:09<00:18, 1.02it/s][A | |
| 40%|ββββ | 12/30 [00:10<00:17, 1.01it/s][A | |
| 43%|βββββ | 13/30 [00:11<00:16, 1.01it/s][A | |
| 47%|βββββ | 14/30 [00:12<00:15, 1.01it/s][A | |
| 50%|βββββ | 15/30 [00:13<00:14, 1.00it/s][A | |
| 53%|ββββββ | 16/30 [00:14<00:13, 1.00it/s][A | |
| 57%|ββββββ | 17/30 [00:16<00:14, 1.08s/it][A | |
| 60%|ββββββ | 18/30 [00:17<00:12, 1.04s/it][A | |
| 63%|βββββββ | 19/30 [00:18<00:11, 1.03s/it][A | |
| 67%|βββββββ | 20/30 [00:19<00:10, 1.02s/it][A | |
| 70%|βββββββ | 21/30 [00:20<00:09, 1.02s/it][A | |
| 73%|ββββββββ | 22/30 [00:21<00:08, 1.01s/it][A | |
| 77%|ββββββββ | 23/30 [00:22<00:07, 1.01s/it][A | |
| 80%|ββββββββ | 24/30 [00:23<00:06, 1.01s/it][A | |
| 83%|βββββββββ | 25/30 [00:24<00:05, 1.01s/it][A | |
| 87%|βββββββββ | 26/30 [00:25<00:04, 1.01s/it][A | |
| 90%|βββββββββ | 27/30 [00:26<00:03, 1.01s/it][A | |
| 93%|ββββββββββ| 28/30 [00:27<00:02, 1.01s/it][A | |
| 97%|ββββββββββ| 29/30 [00:28<00:01, 1.01s/it][A | |
| 100%|ββββββββββ| 30/30 [00:29<00:00, 1.04s/it][A | |
| [A{'eval_loss': '2.203', 'eval_runtime': '72.32', 'eval_samples_per_second': '0.415', 'eval_steps_per_second': '0.415', 'eval_ppl': '9.056', 'memory/max_active (GiB)': '63.23', 'memory/max_allocated (GiB)': '63.23', 'memory/device_reserved (GiB)': '63.62', 'epoch': 0} | |
| 0%| | 0/45 [01:12<?, ?it/s] | |
| 100%|ββββββββββ| 30/30 [00:29<00:00, 1.04s/it][A | |
| [A 2%|β | 1/45 [02:42<1:59:23, 162.80s/it] 4%|β | 2/45 [03:41<1:12:58, 101.82s/it] 7%|β | 3/45 [04:41<57:48, 82.58s/it] 9%|β | 4/45 [05:41<50:18, 73.62s/it] 11%|β | 5/45 [06:41<45:45, 68.64s/it] {'loss': '1.389', 'grad_norm': '2.801', 'learning_rate': '9.947e-05', 'ppl': '4.01', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '2.492', 'tokens/total': 327680, 'tokens/trainable': 5821, 'epoch': '0.2228'} | |
| 11%|β | 5/45 [06:41<45:45, 68.64s/it] 13%|ββ | 6/45 [07:41<42:39, 65.62s/it] 16%|ββ | 7/45 [08:41<40:25, 63.83s/it] 18%|ββ | 8/45 [09:41<38:34, 62.54s/it] 20%|ββ | 9/45 [10:40<37:00, 61.68s/it] 22%|βββ | 10/45 [11:40<35:38, 61.10s/it] {'loss': '1.158', 'grad_norm': '0.9068', 'learning_rate': '9.36e-05', 'ppl': '3.182', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '1.288', 'tokens/total': 655360, 'tokens/trainable': 12183, 'epoch': '0.4457'} | |
| 22%|βββ | 10/45 [11:40<35:38, 61.10s/it] 24%|βββ | 11/45 [12:40<34:23, 60.69s/it] 27%|βββ | 12/45 [13:40<33:13, 60.41s/it] 29%|βββ | 13/45 [14:39<32:06, 60.22s/it] 31%|βββ | 14/45 [15:39<31:02, 60.08s/it] 33%|ββββ | 15/45 [16:39<29:59, 59.99s/it] {'loss': '1.025', 'grad_norm': '0.7377', 'learning_rate': '8.198e-05', 'ppl': '2.788', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '1.974', 'tokens/total': 983040, 'tokens/trainable': 18682, 'epoch': '0.6685'} | |
| 33%|ββββ | 15/45 [16:39<29:59, 59.99s/it] 36%|ββββ | 16/45 [17:39<28:57, 59.92s/it] 38%|ββββ | 17/45 [18:38<27:56, 59.88s/it] 40%|ββββ | 18/45 [19:38<26:56, 59.87s/it] 42%|βββββ | 19/45 [20:38<25:56, 59.86s/it] 44%|βββββ | 20/45 [21:38<24:56, 59.85s/it] {'loss': '0.9663', 'grad_norm': '0.9196', 'learning_rate': '6.614e-05', 'ppl': '2.628', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '2.29', 'tokens/total': 1310720, 'tokens/trainable': 24574, 'epoch': '0.8914'} | |
| 44%|βββββ | 20/45 [21:38<24:56, 59.85s/it] 47%|βββββ | 21/45 [22:38<23:55, 59.83s/it] 49%|βββββ | 22/45 [23:38<22:55, 59.81s/it] 51%|βββββ | 23/45 [24:04<18:14, 49.75s/it] 53%|ββββββ | 24/45 [25:05<18:34, 53.07s/it] 56%|ββββββ | 25/45 [26:04<18:21, 55.08s/it] {'loss': '0.8469', 'grad_norm': '0.7954', 'learning_rate': '4.817e-05', 'ppl': '2.332', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '0.5188', 'tokens/total': 1601536, 'tokens/trainable': 29932, 'epoch': '1.089'} | |
| 56%|ββββββ | 25/45 [26:04<18:21, 55.08s/it] 58%|ββββββ | 26/45 [27:04<17:53, 56.50s/it] 60%|ββββββ | 27/45 [28:04<17:14, 57.48s/it] 62%|βββββββ | 28/45 [29:04<16:28, 58.17s/it] 64%|βββββββ | 29/45 [30:04<15:38, 58.65s/it] 67%|βββββββ | 30/45 [31:03<14:44, 59.00s/it] {'loss': '0.617', 'grad_norm': '0.7464', 'learning_rate': '3.044e-05', 'ppl': '1.853', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '0.7193', 'tokens/total': 1929216, 'tokens/trainable': 35957, 'epoch': '1.312'} | |
| 67%|βββββββ | 30/45 [31:03<14:44, 59.00s/it] 69%|βββββββ | 31/45 [32:03<13:49, 59.23s/it] 71%|βββββββ | 32/45 [33:03<12:52, 59.39s/it] 73%|ββββββββ | 33/45 [34:03<11:54, 59.53s/it] 76%|ββββββββ | 34/45 [35:03<10:56, 59.65s/it] 78%|ββββββββ | 35/45 [36:02<09:56, 59.69s/it] {'loss': '0.5985', 'grad_norm': '0.8065', 'learning_rate': '1.53e-05', 'ppl': '1.819', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '1.405', 'tokens/total': 2256896, 'tokens/trainable': 41669, 'epoch': '1.535'} | |
| 78%|ββββββββ | 35/45 [36:02<09:56, 59.69s/it] 80%|ββββββββ | 36/45 [37:02<08:57, 59.70s/it] 82%|βββββββββ | 37/45 [38:02<07:57, 59.73s/it] 84%|βββββββββ | 38/45 [39:02<06:58, 59.78s/it] 87%|βββββββββ | 39/45 [40:02<05:58, 59.79s/it] 89%|βββββββββ | 40/45 [41:01<04:58, 59.78s/it] {'loss': '0.5577', 'grad_norm': '0.8635', 'learning_rate': '4.728e-06', 'ppl': '1.747', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '2.108', 'tokens/total': 2584576, 'tokens/trainable': 48077, 'epoch': '1.758'} | |
| 89%|βββββββββ | 40/45 [41:01<04:58, 59.78s/it] 91%|βββββββββ | 41/45 [42:01<03:59, 59.77s/it] 93%|ββββββββββ| 42/45 [43:01<02:59, 59.77s/it] 96%|ββββββββββ| 43/45 [44:01<01:59, 59.78s/it] 98%|ββββββββββ| 44/45 [45:01<00:59, 59.77s/it] 100%|ββββββββββ| 45/45 [46:00<00:00, 59.77s/it] {'loss': '0.5623', 'grad_norm': '0.9089', 'learning_rate': '1.334e-07', 'ppl': '1.755', 'memory/max_active (GiB)': '72.06', 'memory/max_allocated (GiB)': '72.06', 'memory/device_reserved (GiB)': '75.09', 'tokens/train_per_sec_per_gpu': '1.808', 'tokens/total': 2912256, 'tokens/trainable': 54318, 'epoch': '1.981'} | |
| 100%|ββββββββββ| 45/45 [46:00<00:00, 59.77s/it][2026-06-24 13:38:42,908] [INFO] [axolotl.core.trainers.base.evaluate:475] [PID:37810] Running evaluation step... | |
| 0%| | 0/30 [00:00<?, ?it/s][A | |
| 7%|β | 2/30 [00:01<00:14, 1.88it/s][A | |
| 10%|β | 3/30 [00:02<00:20, 1.32it/s][A | |
| 13%|ββ | 4/30 [00:03<00:22, 1.15it/s][A | |
| 17%|ββ | 5/30 [00:04<00:23, 1.06it/s][A | |
| 20%|ββ | 6/30 [00:05<00:23, 1.02it/s][A | |
| 23%|βββ | 7/30 [00:06<00:23, 1.01s/it][A | |
| 27%|βββ | 8/30 [00:07<00:22, 1.03s/it][A | |
| 30%|βββ | 9/30 [00:08<00:21, 1.04s/it][A | |
| 33%|ββββ | 10/30 [00:09<00:21, 1.05s/it][A | |
| 37%|ββββ | 11/30 [00:10<00:20, 1.06s/it][A | |
| 40%|ββββ | 12/30 [00:11<00:19, 1.06s/it][A | |
| 43%|βββββ | 13/30 [00:12<00:18, 1.06s/it][A | |
| 47%|βββββ | 14/30 [00:13<00:17, 1.07s/it][A | |
| 50%|βββββ | 15/30 [00:14<00:16, 1.07s/it][A | |
| 53%|ββββββ | 16/30 [00:16<00:14, 1.07s/it][A | |
| 57%|ββββββ | 17/30 [00:17<00:15, 1.20s/it][A | |
| 60%|ββββββ | 18/30 [00:18<00:13, 1.14s/it][A | |
| 63%|βββββββ | 19/30 [00:19<00:12, 1.12s/it][A | |
| 67%|βββββββ | 20/30 [00:20<00:11, 1.11s/it][A | |
| 70%|βββββββ | 21/30 [00:21<00:09, 1.10s/it][A | |
| 73%|ββββββββ | 22/30 [00:22<00:08, 1.09s/it][A | |
| 77%|ββββββββ | 23/30 [00:23<00:07, 1.08s/it][A | |
| 80%|ββββββββ | 24/30 [00:24<00:06, 1.08s/it][A | |
| 83%|βββββββββ | 25/30 [00:26<00:05, 1.08s/it][A | |
| 87%|βββββββββ | 26/30 [00:27<00:04, 1.07s/it][A | |
| 90%|βββββββββ | 27/30 [00:28<00:03, 1.07s/it][A | |
| 93%|ββββββββββ| 28/30 [00:29<00:02, 1.07s/it][A | |
| 97%|ββββββββββ| 29/30 [00:30<00:01, 1.07s/it][A | |
| 100%|ββββββββββ| 30/30 [00:31<00:00, 1.10s/it][A | |
| [A{'eval_loss': '0.8184', 'eval_runtime': '32.84', 'eval_samples_per_second': '0.914', 'eval_steps_per_second': '0.914', 'eval_ppl': '2.267', 'memory/max_active (GiB)': '66.74', 'memory/max_allocated (GiB)': '66.74', 'memory/device_reserved (GiB)': '75.09', 'epoch': '1.981', 'tokens/train_per_sec_per_gpu': '0'} | |
| 100%|ββββββββββ| 45/45 [46:33<00:00, 59.77s/it] | |
| 100%|ββββββββββ| 30/30 [00:31<00:00, 1.10s/it][A | |
| [A[2026-06-24 13:39:15,755] [INFO] [axolotl.core.trainers.base._save:828] [PID:37810] Saving model checkpoint to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl/checkpoint-45 | |
| wandb: Adding directory to artifact (outputs/qwen36-27b-agent-memory-sft-lora-axolotl/checkpoint-45)... Done. 6.0s | |
| {'train_runtime': '2809', 'train_samples_per_second': '0.256', 'train_steps_per_second': '0.016', 'train_loss': '0.8578', 'memory/max_active (GiB)': '56.36', 'memory/max_allocated (GiB)': '56.36', 'memory/device_reserved (GiB)': '67.61', 'epoch': '1.981', 'tokens/train_per_sec_per_gpu': '0'} | |
| 100%|ββββββββββ| 45/45 [46:44<00:00, 59.77s/it] 100%|ββββββββββ| 45/45 [47:01<00:00, 62.69s/it] | |
| [2026-06-24 13:39:43,320] [INFO] [axolotl.train.save_trained_model:267] [PID:37810] Training completed! Saving trained model to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl. | |
| [2026-06-24 13:39:44,079] [INFO] [axolotl.train.save_trained_model:388] [PID:37810] Model successfully saved to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl | |
| [2026-06-24 13:39:44,270] [INFO] [axolotl.core.trainers.base._save:828] [PID:37810] Saving model checkpoint to ./outputs/qwen36-27b-agent-memory-sft-lora-axolotl | |
| [2026-06-24 13:39:45,109] [ERROR] [axolotl.telemetry.errors.wrapper:158] [PID:37810] Error captured in telemetry. Run ID: 37a2bad5-0b99-441f-a93c-4b23f0b23d19 | |
| Traceback (most recent call last): | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/utils/_http.py", line 761, in hf_raise_for_status | |
| response.raise_for_status() | |
| File "/root/miniconda3/lib/python3.10/site-packages/httpx/_models.py", line 829, in raise_for_status | |
| raise HTTPStatusError(message, request=request, response=self) | |
| httpx.HTTPStatusError: Client error '400 Bad Request' for url 'https://huggingface.co/api/validate-yaml' | |
| For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/400 | |
| The above exception was the direct cause of the following exception: | |
| Traceback (most recent call last): | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 11191, in _validate_yaml | |
| hf_raise_for_status(response) | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/utils/_http.py", line 841, in hf_raise_for_status | |
| raise _format(BadRequestError, message, response) from e | |
| huggingface_hub.errors.BadRequestError: (Request ID: Root=1-6a3bde21-0ec1b10c4b5c85cb56bb36a3;a4c1317d-0162-4f09-ae83-92201a21fe1f) | |
| Bad request: | |
| "datasets[0]" with value "./fine-tuning/data/seed/sft_train.jsonl" is not valid. If possible, use a dataset id from https://hf.co/datasets. | |
| The above exception was the direct cause of the following exception: | |
| Traceback (most recent call last): | |
| File "/root/miniconda3/lib/python3.10/runpy.py", line 196, in _run_module_as_main | |
| return _run_code(code, main_globals, None, | |
| File "/root/miniconda3/lib/python3.10/runpy.py", line 86, in _run_code | |
| exec(code, run_globals) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/cli/train.py", line 154, in <module> | |
| fire.Fire(do_cli) | |
| File "/root/miniconda3/lib/python3.10/site-packages/fire/core.py", line 135, in Fire | |
| component_trace = _Fire(component, args, parsed_flag_args, context, name) | |
| File "/root/miniconda3/lib/python3.10/site-packages/fire/core.py", line 468, in _Fire | |
| component, remaining_args = _CallAndUpdateTrace( | |
| File "/root/miniconda3/lib/python3.10/site-packages/fire/core.py", line 684, in _CallAndUpdateTrace | |
| component = fn(*varargs, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/cli/train.py", line 96, in do_cli | |
| do_train(parsed_cfg, parsed_cli_args) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/cli/train.py", line 50, in do_train | |
| model, tokenizer, trainer = train(cfg=cfg, dataset_meta=dataset_meta) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/telemetry/errors.py", line 127, in wrapper | |
| return func(*args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/train.py", line 639, in train | |
| create_model_card(cfg, trainer) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/train.py", line 424, in create_model_card | |
| trainer.push_to_hub() | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/core/trainers/base.py", line 675, in push_to_hub | |
| return super().push_to_hub(*args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/transformers/trainer.py", line 4073, in push_to_hub | |
| return upload_folder( | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 88, in _inner_fn | |
| return fn(*args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 2113, in _inner | |
| return fn(self, *args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 5803, in upload_folder | |
| add_operations = self._prepare_upload_folder_additions( | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 11130, in _prepare_upload_folder_additions | |
| self._validate_yaml( | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 11195, in _validate_yaml | |
| raise ValueError(f"Invalid metadata in README.md.\n{message}") from e | |
| ValueError: Invalid metadata in README.md. | |
| - "datasets[0]" with value "./fine-tuning/data/seed/sft_train.jsonl" is not valid. If possible, use a dataset id from https://hf.co/datasets. | |
| Traceback (most recent call last): | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/utils/_http.py", line 761, in hf_raise_for_status | |
| response.raise_for_status() | |
| File "/root/miniconda3/lib/python3.10/site-packages/httpx/_models.py", line 829, in raise_for_status | |
| raise HTTPStatusError(message, request=request, response=self) | |
| httpx.HTTPStatusError: Client error '400 Bad Request' for url 'https://huggingface.co/api/validate-yaml' | |
| For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/400 | |
| The above exception was the direct cause of the following exception: | |
| Traceback (most recent call last): | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 11191, in _validate_yaml | |
| hf_raise_for_status(response) | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/utils/_http.py", line 841, in hf_raise_for_status | |
| raise _format(BadRequestError, message, response) from e | |
| huggingface_hub.errors.BadRequestError: (Request ID: Root=1-6a3bde21-0ec1b10c4b5c85cb56bb36a3;a4c1317d-0162-4f09-ae83-92201a21fe1f) | |
| Bad request: | |
| "datasets[0]" with value "./fine-tuning/data/seed/sft_train.jsonl" is not valid. If possible, use a dataset id from https://hf.co/datasets. | |
| The above exception was the direct cause of the following exception: | |
| Traceback (most recent call last): | |
| File "/root/miniconda3/lib/python3.10/runpy.py", line 196, in _run_module_as_main | |
| return _run_code(code, main_globals, None, | |
| File "/root/miniconda3/lib/python3.10/runpy.py", line 86, in _run_code | |
| exec(code, run_globals) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/cli/train.py", line 154, in <module> | |
| fire.Fire(do_cli) | |
| File "/root/miniconda3/lib/python3.10/site-packages/fire/core.py", line 135, in Fire | |
| component_trace = _Fire(component, args, parsed_flag_args, context, name) | |
| File "/root/miniconda3/lib/python3.10/site-packages/fire/core.py", line 468, in _Fire | |
| component, remaining_args = _CallAndUpdateTrace( | |
| File "/root/miniconda3/lib/python3.10/site-packages/fire/core.py", line 684, in _CallAndUpdateTrace | |
| component = fn(*varargs, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/cli/train.py", line 96, in do_cli | |
| do_train(parsed_cfg, parsed_cli_args) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/cli/train.py", line 50, in do_train | |
| model, tokenizer, trainer = train(cfg=cfg, dataset_meta=dataset_meta) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/telemetry/errors.py", line 127, in wrapper | |
| return func(*args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/train.py", line 639, in train | |
| create_model_card(cfg, trainer) | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/train.py", line 424, in create_model_card | |
| trainer.push_to_hub() | |
| File "/root/miniconda3/lib/python3.10/site-packages/axolotl/core/trainers/base.py", line 675, in push_to_hub | |
| return super().push_to_hub(*args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/transformers/trainer.py", line 4073, in push_to_hub | |
| return upload_folder( | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 88, in _inner_fn | |
| return fn(*args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 2113, in _inner | |
| return fn(self, *args, **kwargs) | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 5803, in upload_folder | |
| add_operations = self._prepare_upload_folder_additions( | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 11130, in _prepare_upload_folder_additions | |
| self._validate_yaml( | |
| File "/root/miniconda3/lib/python3.10/site-packages/huggingface_hub/hf_api.py", line 11195, in _validate_yaml | |
| raise ValueError(f"Invalid metadata in README.md.\n{message}") from e | |
| ValueError: Invalid metadata in README.md. | |
| - "datasets[0]" with value "./fine-tuning/data/seed/sft_train.jsonl" is not valid. If possible, use a dataset id from https://hf.co/datasets. | |