base_model: Qwen/Qwen3-14B-Base

plugins:
  - axolotl.integrations.liger.LigerPlugin
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true

chat_template_jinja: "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set content = message.content %}\n        {%- set reasoning_content = '' %}\n        {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}"

datasets:
  - path: winglian/cuda-engineer-augment-v4-filtered
    type: chat_template
    split: train
    # split_thinking: true
    eot_tokens: ["<|im_end|>"]
  - path: axolotl-ai-internal/gpumode-py2triton-reasoning-v2-filtered
    type: chat_template
    split: train
    # split_thinking: true
    eot_tokens: ["<|im_end|>"]

dataset_prepared_path: last_run_prepared
val_set_size: 0.005
output_dir: ./outputs/out
save_only_model: true

sequence_len: 16384
sample_packing: true
pad_to_sequence_len: true

wandb_project: qwen3-14b-grpo-triton
wandb_entity: axolotl-ai
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 3
optimizer: adamw_torch_fused
max_grad_norm: 0.1
neftune_noise_alpha: 10
lr_scheduler: cosine
learning_rate: 1e-5

bf16: true
tf32: true

gradient_checkpointing: offload
gradient_checkpointing_kwargs:
  use_reentrant: false
logging_steps: 1
flash_attention: true

warmup_steps: 100
evals_per_epoch: 5
saves_per_epoch: 1
weight_decay: 0.01
deepspeed: deepspeed_configs/zero1.json

outputs/out

This model is a fine-tuned version of Qwen/Qwen3-14B-Base on the winglian/cuda-engineer-augment-v4-filtered and the axolotl-ai-internal/gpumode-py2triton-reasoning-v2-filtered datasets. It achieves the following results on the evaluation set:

Loss: 0.2262

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: 1e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 16
total_eval_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 100
num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.4626	0.0056	1	0.4989
0.3018	0.2	36	0.3577
0.2528	0.4	72	0.2954
0.2273	0.6	108	0.2686
0.2238	0.8	144	0.2540
0.2143	1.0	180	0.2458
0.1964	1.2	216	0.2387
0.1913	1.4	252	0.2357
0.1809	1.6	288	0.2327
0.1814	1.8	324	0.2296
0.1769	2.0	360	0.2271
0.1638	2.2	396	0.2253
0.1594	2.4	432	0.2257
0.154	2.6	468	0.2262
0.1578	2.8	504	0.2262
0.1571	3.0	540	0.2262

Framework versions

Transformers 4.51.3
Pytorch 2.6.0+cu124
Datasets 3.5.1
Tokenizers 0.21.1

winglian/qwen3-14b-triton-sft-v3

outputs/out

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

winglian/
qwen3-14b-triton-sft-v3