Upload folder using huggingface_hub
Browse files- README.md +68 -0
- checkpoint-300/chat_template.jinja +45 -0
- checkpoint-300/config.json +61 -0
- checkpoint-300/generation_config.json +9 -0
- checkpoint-300/model.safetensors +3 -0
- checkpoint-300/optimizer.pt +3 -0
- checkpoint-300/rng_state.pth +3 -0
- checkpoint-300/scheduler.pt +3 -0
- checkpoint-300/tokenizer.json +0 -0
- checkpoint-300/tokenizer_config.json +20 -0
- checkpoint-300/trainer_state.json +532 -0
- checkpoint-300/training_args.bin +3 -0
- checkpoint-400/chat_template.jinja +45 -0
- checkpoint-400/config.json +61 -0
- checkpoint-400/generation_config.json +9 -0
- checkpoint-400/model.safetensors +3 -0
- checkpoint-400/optimizer.pt +3 -0
- checkpoint-400/rng_state.pth +3 -0
- checkpoint-400/scheduler.pt +3 -0
- checkpoint-400/tokenizer.json +0 -0
- checkpoint-400/tokenizer_config.json +20 -0
- checkpoint-400/trainer_state.json +698 -0
- checkpoint-400/training_args.bin +3 -0
- checkpoint-432/chat_template.jinja +45 -0
- checkpoint-432/config.json +61 -0
- checkpoint-432/generation_config.json +9 -0
- checkpoint-432/model.safetensors +3 -0
- checkpoint-432/optimizer.pt +3 -0
- checkpoint-432/rng_state.pth +3 -0
- checkpoint-432/scheduler.pt +3 -0
- checkpoint-432/tokenizer.json +0 -0
- checkpoint-432/tokenizer_config.json +20 -0
- checkpoint-432/trainer_state.json +743 -0
- checkpoint-432/training_args.bin +3 -0
- final_model/chat_template.jinja +45 -0
- final_model/config.json +61 -0
- final_model/generation_config.json +9 -0
- final_model/model.safetensors +3 -0
- final_model/tokenizer.json +0 -0
- final_model/tokenizer_config.json +20 -0
- final_model/training_args.bin +3 -0
- run_meta.json +39 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: LiquidAI/LFM2.5-1.2B-Instruct
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: dpo_fft_LFM2.5-1.2B-Instruct_argilla__distilabel-math-preference-dpo_20260222_195126
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- dpo
|
| 8 |
+
- trl
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for dpo_fft_LFM2.5-1.2B-Instruct_argilla__distilabel-math-preference-dpo_20260222_195126
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [LiquidAI/LFM2.5-1.2B-Instruct](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Instruct).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="None", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).
|
| 34 |
+
|
| 35 |
+
### Framework versions
|
| 36 |
+
|
| 37 |
+
- TRL: 0.28.0
|
| 38 |
+
- Transformers: 5.2.0
|
| 39 |
+
- Pytorch: 2.8.0
|
| 40 |
+
- Datasets: 4.5.0
|
| 41 |
+
- Tokenizers: 0.22.2
|
| 42 |
+
|
| 43 |
+
## Citations
|
| 44 |
+
|
| 45 |
+
Cite DPO as:
|
| 46 |
+
|
| 47 |
+
```bibtex
|
| 48 |
+
@inproceedings{rafailov2023direct,
|
| 49 |
+
title = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
|
| 50 |
+
author = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
|
| 51 |
+
year = 2023,
|
| 52 |
+
booktitle = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
|
| 53 |
+
url = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
|
| 54 |
+
editor = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
|
| 55 |
+
}
|
| 56 |
+
```
|
| 57 |
+
|
| 58 |
+
Cite TRL as:
|
| 59 |
+
|
| 60 |
+
```bibtex
|
| 61 |
+
@software{vonwerra2020trl,
|
| 62 |
+
title = {{TRL: Transformers Reinforcement Learning}},
|
| 63 |
+
author = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
|
| 64 |
+
license = {Apache-2.0},
|
| 65 |
+
url = {https://github.com/huggingface/trl},
|
| 66 |
+
year = {2020}
|
| 67 |
+
}
|
| 68 |
+
```
|
checkpoint-300/chat_template.jinja
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{- bos_token -}}
|
| 2 |
+
{%- set keep_past_thinking = keep_past_thinking | default(false) -%}
|
| 3 |
+
{%- set ns = namespace(system_prompt="") -%}
|
| 4 |
+
{%- if messages[0]["role"] == "system" -%}
|
| 5 |
+
{%- set ns.system_prompt = messages[0]["content"] -%}
|
| 6 |
+
{%- set messages = messages[1:] -%}
|
| 7 |
+
{%- endif -%}
|
| 8 |
+
{%- if tools -%}
|
| 9 |
+
{%- set ns.system_prompt = ns.system_prompt + ("\n" if ns.system_prompt else "") + "List of tools: [" -%}
|
| 10 |
+
{%- for tool in tools -%}
|
| 11 |
+
{%- if tool is not string -%}
|
| 12 |
+
{%- set tool = tool | tojson -%}
|
| 13 |
+
{%- endif -%}
|
| 14 |
+
{%- set ns.system_prompt = ns.system_prompt + tool -%}
|
| 15 |
+
{%- if not loop.last -%}
|
| 16 |
+
{%- set ns.system_prompt = ns.system_prompt + ", " -%}
|
| 17 |
+
{%- endif -%}
|
| 18 |
+
{%- endfor -%}
|
| 19 |
+
{%- set ns.system_prompt = ns.system_prompt + "]" -%}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if ns.system_prompt -%}
|
| 22 |
+
{{- "<|im_start|>system\n" + ns.system_prompt + "<|im_end|>\n" -}}
|
| 23 |
+
{%- endif -%}
|
| 24 |
+
{%- set ns.last_assistant_index = -1 -%}
|
| 25 |
+
{%- for message in messages -%}
|
| 26 |
+
{%- if message["role"] == "assistant" -%}
|
| 27 |
+
{%- set ns.last_assistant_index = loop.index0 -%}
|
| 28 |
+
{%- endif -%}
|
| 29 |
+
{%- endfor -%}
|
| 30 |
+
{%- for message in messages -%}
|
| 31 |
+
{{- "<|im_start|>" + message["role"] + "\n" -}}
|
| 32 |
+
{%- set content = message["content"] -%}
|
| 33 |
+
{%- if content is not string -%}
|
| 34 |
+
{%- set content = content | tojson -%}
|
| 35 |
+
{%- endif -%}
|
| 36 |
+
{%- if message["role"] == "assistant" and not keep_past_thinking and loop.index0 != ns.last_assistant_index -%}
|
| 37 |
+
{%- if "</think>" in content -%}
|
| 38 |
+
{%- set content = content.split("</think>")[-1] | trim -%}
|
| 39 |
+
{%- endif -%}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{- content + "<|im_end|>\n" -}}
|
| 42 |
+
{%- endfor -%}
|
| 43 |
+
{%- if add_generation_prompt -%}
|
| 44 |
+
{{- "<|im_start|>assistant\n" -}}
|
| 45 |
+
{%- endif -%}
|
checkpoint-300/config.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Lfm2ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"block_auto_adjust_ff_dim": true,
|
| 6 |
+
"block_dim": 2048,
|
| 7 |
+
"block_ff_dim": 12288,
|
| 8 |
+
"block_ffn_dim_multiplier": 1.0,
|
| 9 |
+
"block_mlp_init_scale": 1.0,
|
| 10 |
+
"block_multiple_of": 256,
|
| 11 |
+
"block_norm_eps": 1e-05,
|
| 12 |
+
"block_out_init_scale": 1.0,
|
| 13 |
+
"block_use_swiglu": true,
|
| 14 |
+
"block_use_xavier_init": true,
|
| 15 |
+
"bos_token_id": 1,
|
| 16 |
+
"conv_L_cache": 3,
|
| 17 |
+
"conv_bias": false,
|
| 18 |
+
"conv_dim": 2048,
|
| 19 |
+
"conv_use_xavier_init": true,
|
| 20 |
+
"dtype": "bfloat16",
|
| 21 |
+
"eos_token_id": 7,
|
| 22 |
+
"hidden_size": 2048,
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 12288,
|
| 25 |
+
"layer_types": [
|
| 26 |
+
"conv",
|
| 27 |
+
"conv",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"conv",
|
| 30 |
+
"conv",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"conv",
|
| 33 |
+
"conv",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"conv",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"conv",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"conv",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"conv"
|
| 42 |
+
],
|
| 43 |
+
"max_position_embeddings": 128000,
|
| 44 |
+
"model_type": "lfm2",
|
| 45 |
+
"norm_eps": 1e-05,
|
| 46 |
+
"num_attention_heads": 32,
|
| 47 |
+
"num_heads": 32,
|
| 48 |
+
"num_hidden_layers": 16,
|
| 49 |
+
"num_key_value_heads": 8,
|
| 50 |
+
"pad_token_id": 0,
|
| 51 |
+
"rope_parameters": {
|
| 52 |
+
"rope_theta": 1000000.0,
|
| 53 |
+
"rope_type": "default"
|
| 54 |
+
},
|
| 55 |
+
"tie_embedding": true,
|
| 56 |
+
"tie_word_embeddings": true,
|
| 57 |
+
"transformers_version": "5.2.0",
|
| 58 |
+
"use_cache": false,
|
| 59 |
+
"use_pos_enc": true,
|
| 60 |
+
"vocab_size": 65536
|
| 61 |
+
}
|
checkpoint-300/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
7
|
| 6 |
+
],
|
| 7 |
+
"pad_token_id": 0,
|
| 8 |
+
"transformers_version": "5.2.0"
|
| 9 |
+
}
|
checkpoint-300/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ea8a8b6c80f8066d94803546515de24808278c0cfb7efff8e029129e40dd94ff
|
| 3 |
+
size 2340697936
|
checkpoint-300/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0115530fa698b4ae187d35c21f6abaf53d807fda66ea7b7457a826086059b84
|
| 3 |
+
size 3178927435
|
checkpoint-300/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01f9a0f7843a37be87edd23f4e88aa93b38b95cc2c07503eeb1cf2e4632453a2
|
| 3 |
+
size 14645
|
checkpoint-300/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:426f979e5c76efd0e1e3fe301e36c368d85ae96e9cf26208baaf62ca2cac3280
|
| 3 |
+
size 1465
|
checkpoint-300/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-300/tokenizer_config.json
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"bos_token": "<|startoftext|>",
|
| 4 |
+
"clean_up_tokenization_spaces": false,
|
| 5 |
+
"eos_token": "<|im_end|>",
|
| 6 |
+
"is_local": false,
|
| 7 |
+
"legacy": false,
|
| 8 |
+
"model_input_names": [
|
| 9 |
+
"input_ids",
|
| 10 |
+
"attention_mask"
|
| 11 |
+
],
|
| 12 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 13 |
+
"pad_token": "<|pad|>",
|
| 14 |
+
"padding_side": "right",
|
| 15 |
+
"sp_model_kwargs": {},
|
| 16 |
+
"spaces_between_special_tokens": false,
|
| 17 |
+
"tokenizer_class": "TokenizersBackend",
|
| 18 |
+
"use_default_system_prompt": false,
|
| 19 |
+
"use_fast": true
|
| 20 |
+
}
|
checkpoint-300/trainer_state.json
ADDED
|
@@ -0,0 +1,532 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 300,
|
| 3 |
+
"best_metric": 0.6872237920761108,
|
| 4 |
+
"best_model_checkpoint": "models/dpo_fft_LFM2.5-1.2B-Instruct_argilla__distilabel-math-preference-dpo_20260222_195126/checkpoint-300",
|
| 5 |
+
"epoch": 2.0834782608695654,
|
| 6 |
+
"eval_steps": 100,
|
| 7 |
+
"global_step": 300,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.06956521739130435,
|
| 14 |
+
"grad_norm": 44.25,
|
| 15 |
+
"learning_rate": 1.0227272727272728e-07,
|
| 16 |
+
"logits/chosen": -1.0599091053009033,
|
| 17 |
+
"logits/rejected": -1.042307734489441,
|
| 18 |
+
"logps/chosen": -332.13507080078125,
|
| 19 |
+
"logps/rejected": -332.9361267089844,
|
| 20 |
+
"loss": 0.6927361965179444,
|
| 21 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 22 |
+
"rewards/chosen": 0.01432617288082838,
|
| 23 |
+
"rewards/margins": 0.0042648655362427235,
|
| 24 |
+
"rewards/rejected": 0.010061310604214668,
|
| 25 |
+
"step": 10
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.1391304347826087,
|
| 29 |
+
"grad_norm": 50.5,
|
| 30 |
+
"learning_rate": 2.159090909090909e-07,
|
| 31 |
+
"logits/chosen": -1.0978257656097412,
|
| 32 |
+
"logits/rejected": -1.09658682346344,
|
| 33 |
+
"logps/chosen": -329.22100830078125,
|
| 34 |
+
"logps/rejected": -312.2664794921875,
|
| 35 |
+
"loss": 0.7003747940063476,
|
| 36 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 37 |
+
"rewards/chosen": 0.004466252867132425,
|
| 38 |
+
"rewards/margins": -0.011238865554332733,
|
| 39 |
+
"rewards/rejected": 0.01570511981844902,
|
| 40 |
+
"step": 20
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"epoch": 0.20869565217391303,
|
| 44 |
+
"grad_norm": 54.75,
|
| 45 |
+
"learning_rate": 3.295454545454545e-07,
|
| 46 |
+
"logits/chosen": -1.1366004943847656,
|
| 47 |
+
"logits/rejected": -1.1707069873809814,
|
| 48 |
+
"logps/chosen": -328.1323547363281,
|
| 49 |
+
"logps/rejected": -305.90618896484375,
|
| 50 |
+
"loss": 0.7031519412994385,
|
| 51 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 52 |
+
"rewards/chosen": 0.03125340864062309,
|
| 53 |
+
"rewards/margins": -0.01610126718878746,
|
| 54 |
+
"rewards/rejected": 0.04735467582941055,
|
| 55 |
+
"step": 30
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 0.2782608695652174,
|
| 59 |
+
"grad_norm": 50.5,
|
| 60 |
+
"learning_rate": 4.4318181818181817e-07,
|
| 61 |
+
"logits/chosen": -1.0889461040496826,
|
| 62 |
+
"logits/rejected": -1.1017862558364868,
|
| 63 |
+
"logps/chosen": -338.80938720703125,
|
| 64 |
+
"logps/rejected": -328.07464599609375,
|
| 65 |
+
"loss": 0.6893259048461914,
|
| 66 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 67 |
+
"rewards/chosen": 0.09773217141628265,
|
| 68 |
+
"rewards/margins": 0.01154844556003809,
|
| 69 |
+
"rewards/rejected": 0.08618371188640594,
|
| 70 |
+
"step": 40
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"epoch": 0.34782608695652173,
|
| 74 |
+
"grad_norm": 53.5,
|
| 75 |
+
"learning_rate": 4.997951542310825e-07,
|
| 76 |
+
"logits/chosen": -1.1367381811141968,
|
| 77 |
+
"logits/rejected": -1.1468210220336914,
|
| 78 |
+
"logps/chosen": -329.12823486328125,
|
| 79 |
+
"logps/rejected": -323.8036804199219,
|
| 80 |
+
"loss": 0.6982084751129151,
|
| 81 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 82 |
+
"rewards/chosen": 0.1568203717470169,
|
| 83 |
+
"rewards/margins": -0.004218918737024069,
|
| 84 |
+
"rewards/rejected": 0.1610392928123474,
|
| 85 |
+
"step": 50
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"epoch": 0.41739130434782606,
|
| 89 |
+
"grad_norm": 64.0,
|
| 90 |
+
"learning_rate": 4.98158401695492e-07,
|
| 91 |
+
"logits/chosen": -1.0899083614349365,
|
| 92 |
+
"logits/rejected": -1.0698351860046387,
|
| 93 |
+
"logps/chosen": -335.6194152832031,
|
| 94 |
+
"logps/rejected": -329.00653076171875,
|
| 95 |
+
"loss": 0.6905360221862793,
|
| 96 |
+
"rewards/accuracies": 0.53125,
|
| 97 |
+
"rewards/chosen": 0.1543850153684616,
|
| 98 |
+
"rewards/margins": 0.010315253399312496,
|
| 99 |
+
"rewards/rejected": 0.14406974613666534,
|
| 100 |
+
"step": 60
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.48695652173913045,
|
| 104 |
+
"grad_norm": 72.5,
|
| 105 |
+
"learning_rate": 4.948956212473369e-07,
|
| 106 |
+
"logits/chosen": -1.1223671436309814,
|
| 107 |
+
"logits/rejected": -1.113906741142273,
|
| 108 |
+
"logps/chosen": -319.424560546875,
|
| 109 |
+
"logps/rejected": -317.76934814453125,
|
| 110 |
+
"loss": 0.7009393215179444,
|
| 111 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 112 |
+
"rewards/chosen": 0.18402941524982452,
|
| 113 |
+
"rewards/margins": -0.010006649419665337,
|
| 114 |
+
"rewards/rejected": 0.19403605163097382,
|
| 115 |
+
"step": 70
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.5565217391304348,
|
| 119 |
+
"grad_norm": 52.5,
|
| 120 |
+
"learning_rate": 4.900281918608732e-07,
|
| 121 |
+
"logits/chosen": -1.1157792806625366,
|
| 122 |
+
"logits/rejected": -1.1078118085861206,
|
| 123 |
+
"logps/chosen": -314.5404052734375,
|
| 124 |
+
"logps/rejected": -311.91656494140625,
|
| 125 |
+
"loss": 0.6877760887145996,
|
| 126 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 127 |
+
"rewards/chosen": 0.19012701511383057,
|
| 128 |
+
"rewards/margins": 0.016789216548204422,
|
| 129 |
+
"rewards/rejected": 0.17333778738975525,
|
| 130 |
+
"step": 80
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.6260869565217392,
|
| 134 |
+
"grad_norm": 47.75,
|
| 135 |
+
"learning_rate": 4.835880067784441e-07,
|
| 136 |
+
"logits/chosen": -1.0738165378570557,
|
| 137 |
+
"logits/rejected": -1.0576341152191162,
|
| 138 |
+
"logps/chosen": -336.3821716308594,
|
| 139 |
+
"logps/rejected": -331.9998779296875,
|
| 140 |
+
"loss": 0.6956596374511719,
|
| 141 |
+
"rewards/accuracies": 0.4375,
|
| 142 |
+
"rewards/chosen": 0.21225734055042267,
|
| 143 |
+
"rewards/margins": 0.0009234324097633362,
|
| 144 |
+
"rewards/rejected": 0.21133390069007874,
|
| 145 |
+
"step": 90
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.6956521739130435,
|
| 149 |
+
"grad_norm": 49.0,
|
| 150 |
+
"learning_rate": 4.7561726453386744e-07,
|
| 151 |
+
"logits/chosen": -1.0457407236099243,
|
| 152 |
+
"logits/rejected": -1.038447618484497,
|
| 153 |
+
"logps/chosen": -330.8192138671875,
|
| 154 |
+
"logps/rejected": -323.5484924316406,
|
| 155 |
+
"loss": 0.6817105770111084,
|
| 156 |
+
"rewards/accuracies": 0.581250011920929,
|
| 157 |
+
"rewards/chosen": 0.26413798332214355,
|
| 158 |
+
"rewards/margins": 0.029450010508298874,
|
| 159 |
+
"rewards/rejected": 0.2346879541873932,
|
| 160 |
+
"step": 100
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.6956521739130435,
|
| 164 |
+
"eval_logits/chosen": -1.0708128213882446,
|
| 165 |
+
"eval_logits/rejected": -1.0917303562164307,
|
| 166 |
+
"eval_logps/chosen": -317.1337585449219,
|
| 167 |
+
"eval_logps/rejected": -318.0965270996094,
|
| 168 |
+
"eval_loss": 0.6924862265586853,
|
| 169 |
+
"eval_rewards/accuracies": 0.4838709533214569,
|
| 170 |
+
"eval_rewards/chosen": 0.2332489788532257,
|
| 171 |
+
"eval_rewards/margins": 0.0038641756400465965,
|
| 172 |
+
"eval_rewards/rejected": 0.22938479483127594,
|
| 173 |
+
"eval_runtime": 11.3087,
|
| 174 |
+
"eval_samples_per_second": 10.7,
|
| 175 |
+
"eval_steps_per_second": 2.741,
|
| 176 |
+
"step": 100
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 0.7652173913043478,
|
| 180 |
+
"grad_norm": 47.0,
|
| 181 |
+
"learning_rate": 4.6616819245164655e-07,
|
| 182 |
+
"logits/chosen": -1.0736799240112305,
|
| 183 |
+
"logits/rejected": -1.0783708095550537,
|
| 184 |
+
"logps/chosen": -314.97943115234375,
|
| 185 |
+
"logps/rejected": -308.35675048828125,
|
| 186 |
+
"loss": 0.6984636306762695,
|
| 187 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 188 |
+
"rewards/chosen": 0.22377920150756836,
|
| 189 |
+
"rewards/margins": -0.0050087496638298035,
|
| 190 |
+
"rewards/rejected": 0.22878792881965637,
|
| 191 |
+
"step": 110
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.8347826086956521,
|
| 195 |
+
"grad_norm": 51.25,
|
| 196 |
+
"learning_rate": 4.55302704433743e-07,
|
| 197 |
+
"logits/chosen": -1.048730731010437,
|
| 198 |
+
"logits/rejected": -1.0437910556793213,
|
| 199 |
+
"logps/chosen": -332.6585388183594,
|
| 200 |
+
"logps/rejected": -319.9508361816406,
|
| 201 |
+
"loss": 0.6809643268585205,
|
| 202 |
+
"rewards/accuracies": 0.59375,
|
| 203 |
+
"rewards/chosen": 0.3183898627758026,
|
| 204 |
+
"rewards/margins": 0.03039008378982544,
|
| 205 |
+
"rewards/rejected": 0.2879997789859772,
|
| 206 |
+
"step": 120
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.9043478260869565,
|
| 210 |
+
"grad_norm": 57.5,
|
| 211 |
+
"learning_rate": 4.4309199527622254e-07,
|
| 212 |
+
"logits/chosen": -1.0928480625152588,
|
| 213 |
+
"logits/rejected": -1.0909947156906128,
|
| 214 |
+
"logps/chosen": -333.54510498046875,
|
| 215 |
+
"logps/rejected": -321.5794372558594,
|
| 216 |
+
"loss": 0.6884570121765137,
|
| 217 |
+
"rewards/accuracies": 0.518750011920929,
|
| 218 |
+
"rewards/chosen": 0.3165169656276703,
|
| 219 |
+
"rewards/margins": 0.015730690211057663,
|
| 220 |
+
"rewards/rejected": 0.3007862865924835,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.9739130434782609,
|
| 225 |
+
"grad_norm": 50.75,
|
| 226 |
+
"learning_rate": 4.2961607417396517e-07,
|
| 227 |
+
"logits/chosen": -1.0702852010726929,
|
| 228 |
+
"logits/rejected": -1.0860363245010376,
|
| 229 |
+
"logps/chosen": -336.94122314453125,
|
| 230 |
+
"logps/rejected": -312.6028137207031,
|
| 231 |
+
"loss": 0.6929869651794434,
|
| 232 |
+
"rewards/accuracies": 0.5,
|
| 233 |
+
"rewards/chosen": 0.3366253077983856,
|
| 234 |
+
"rewards/margins": 0.008591671474277973,
|
| 235 |
+
"rewards/rejected": 0.3280336260795593,
|
| 236 |
+
"step": 140
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 1.0417391304347827,
|
| 240 |
+
"grad_norm": 58.25,
|
| 241 |
+
"learning_rate": 4.1496324047009244e-07,
|
| 242 |
+
"logits/chosen": -1.137653112411499,
|
| 243 |
+
"logits/rejected": -1.1385780572891235,
|
| 244 |
+
"logps/chosen": -297.7463073730469,
|
| 245 |
+
"logps/rejected": -304.8460388183594,
|
| 246 |
+
"loss": 0.6796726703643798,
|
| 247 |
+
"rewards/accuracies": 0.5320512652397156,
|
| 248 |
+
"rewards/chosen": 0.36476510763168335,
|
| 249 |
+
"rewards/margins": 0.029467487707734108,
|
| 250 |
+
"rewards/rejected": 0.3352976143360138,
|
| 251 |
+
"step": 150
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.111304347826087,
|
| 255 |
+
"grad_norm": 62.0,
|
| 256 |
+
"learning_rate": 3.9922950508520126e-07,
|
| 257 |
+
"logits/chosen": -1.0764613151550293,
|
| 258 |
+
"logits/rejected": -1.073861837387085,
|
| 259 |
+
"logps/chosen": -325.74151611328125,
|
| 260 |
+
"logps/rejected": -314.5641174316406,
|
| 261 |
+
"loss": 0.6664072513580322,
|
| 262 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 263 |
+
"rewards/chosen": 0.37418311834335327,
|
| 264 |
+
"rewards/margins": 0.06225377321243286,
|
| 265 |
+
"rewards/rejected": 0.3119293451309204,
|
| 266 |
+
"step": 160
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.1808695652173913,
|
| 270 |
+
"grad_norm": 48.25,
|
| 271 |
+
"learning_rate": 3.8251796141741945e-07,
|
| 272 |
+
"logits/chosen": -1.1032686233520508,
|
| 273 |
+
"logits/rejected": -1.1010105609893799,
|
| 274 |
+
"logps/chosen": -324.8043212890625,
|
| 275 |
+
"logps/rejected": -314.7898254394531,
|
| 276 |
+
"loss": 0.6656156063079834,
|
| 277 |
+
"rewards/accuracies": 0.643750011920929,
|
| 278 |
+
"rewards/chosen": 0.39548203349113464,
|
| 279 |
+
"rewards/margins": 0.06372012943029404,
|
| 280 |
+
"rewards/rejected": 0.33176189661026,
|
| 281 |
+
"step": 170
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 1.2504347826086957,
|
| 285 |
+
"grad_norm": 73.5,
|
| 286 |
+
"learning_rate": 3.649381098353834e-07,
|
| 287 |
+
"logits/chosen": -1.108781099319458,
|
| 288 |
+
"logits/rejected": -1.081012487411499,
|
| 289 |
+
"logps/chosen": -333.0024719238281,
|
| 290 |
+
"logps/rejected": -317.6542663574219,
|
| 291 |
+
"loss": 0.6682669162750244,
|
| 292 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 293 |
+
"rewards/chosen": 0.4114314019680023,
|
| 294 |
+
"rewards/margins": 0.06464624404907227,
|
| 295 |
+
"rewards/rejected": 0.34678512811660767,
|
| 296 |
+
"step": 180
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 1.32,
|
| 300 |
+
"grad_norm": 51.75,
|
| 301 |
+
"learning_rate": 3.466051401903162e-07,
|
| 302 |
+
"logits/chosen": -1.0279661417007446,
|
| 303 |
+
"logits/rejected": -1.028050422668457,
|
| 304 |
+
"logps/chosen": -342.05322265625,
|
| 305 |
+
"logps/rejected": -329.083984375,
|
| 306 |
+
"loss": 0.6815151691436767,
|
| 307 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 308 |
+
"rewards/chosen": 0.4475061893463135,
|
| 309 |
+
"rewards/margins": 0.03387105464935303,
|
| 310 |
+
"rewards/rejected": 0.41363516449928284,
|
| 311 |
+
"step": 190
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"epoch": 1.3895652173913042,
|
| 315 |
+
"grad_norm": 43.25,
|
| 316 |
+
"learning_rate": 3.276391770484606e-07,
|
| 317 |
+
"logits/chosen": -1.0554686784744263,
|
| 318 |
+
"logits/rejected": -1.0444362163543701,
|
| 319 |
+
"logps/chosen": -323.93359375,
|
| 320 |
+
"logps/rejected": -300.8354187011719,
|
| 321 |
+
"loss": 0.6729463577270508,
|
| 322 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 323 |
+
"rewards/chosen": 0.4051585793495178,
|
| 324 |
+
"rewards/margins": 0.04842006787657738,
|
| 325 |
+
"rewards/rejected": 0.35673850774765015,
|
| 326 |
+
"step": 200
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 1.3895652173913042,
|
| 330 |
+
"eval_logits/chosen": -1.059758186340332,
|
| 331 |
+
"eval_logits/rejected": -1.0798335075378418,
|
| 332 |
+
"eval_logps/chosen": -315.1194763183594,
|
| 333 |
+
"eval_logps/rejected": -316.16046142578125,
|
| 334 |
+
"eval_loss": 0.6918731331825256,
|
| 335 |
+
"eval_rewards/accuracies": 0.5483871102333069,
|
| 336 |
+
"eval_rewards/chosen": 0.4346778690814972,
|
| 337 |
+
"eval_rewards/margins": 0.011690356768667698,
|
| 338 |
+
"eval_rewards/rejected": 0.42298752069473267,
|
| 339 |
+
"eval_runtime": 11.2737,
|
| 340 |
+
"eval_samples_per_second": 10.733,
|
| 341 |
+
"eval_steps_per_second": 2.75,
|
| 342 |
+
"step": 200
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 1.4591304347826086,
|
| 346 |
+
"grad_norm": 44.5,
|
| 347 |
+
"learning_rate": 3.0816449258938656e-07,
|
| 348 |
+
"logits/chosen": -1.0483617782592773,
|
| 349 |
+
"logits/rejected": -1.0491857528686523,
|
| 350 |
+
"logps/chosen": -326.4630126953125,
|
| 351 |
+
"logps/rejected": -321.35467529296875,
|
| 352 |
+
"loss": 0.6760656356811523,
|
| 353 |
+
"rewards/accuracies": 0.59375,
|
| 354 |
+
"rewards/chosen": 0.4774439334869385,
|
| 355 |
+
"rewards/margins": 0.04994555562734604,
|
| 356 |
+
"rewards/rejected": 0.42749834060668945,
|
| 357 |
+
"step": 210
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 1.528695652173913,
|
| 361 |
+
"grad_norm": 69.0,
|
| 362 |
+
"learning_rate": 2.883086923275658e-07,
|
| 363 |
+
"logits/chosen": -1.0638173818588257,
|
| 364 |
+
"logits/rejected": -1.0861108303070068,
|
| 365 |
+
"logps/chosen": -323.24517822265625,
|
| 366 |
+
"logps/rejected": -316.22454833984375,
|
| 367 |
+
"loss": 0.6859281063079834,
|
| 368 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 369 |
+
"rewards/chosen": 0.464817613363266,
|
| 370 |
+
"rewards/margins": 0.025601129978895187,
|
| 371 |
+
"rewards/rejected": 0.4392164647579193,
|
| 372 |
+
"step": 220
|
| 373 |
+
},
|
| 374 |
+
{
|
| 375 |
+
"epoch": 1.5982608695652174,
|
| 376 |
+
"grad_norm": 51.25,
|
| 377 |
+
"learning_rate": 2.6820187899267197e-07,
|
| 378 |
+
"logits/chosen": -1.0582925081253052,
|
| 379 |
+
"logits/rejected": -1.0700470209121704,
|
| 380 |
+
"logps/chosen": -331.1497802734375,
|
| 381 |
+
"logps/rejected": -318.2004089355469,
|
| 382 |
+
"loss": 0.6922653675079345,
|
| 383 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 384 |
+
"rewards/chosen": 0.4878065586090088,
|
| 385 |
+
"rewards/margins": 0.014133572578430176,
|
| 386 |
+
"rewards/rejected": 0.4736729562282562,
|
| 387 |
+
"step": 230
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 1.6678260869565218,
|
| 391 |
+
"grad_norm": 44.0,
|
| 392 |
+
"learning_rate": 2.4797580004718036e-07,
|
| 393 |
+
"logits/chosen": -1.059127926826477,
|
| 394 |
+
"logits/rejected": -1.0511342287063599,
|
| 395 |
+
"logps/chosen": -329.36309814453125,
|
| 396 |
+
"logps/rejected": -324.93060302734375,
|
| 397 |
+
"loss": 0.6760355949401855,
|
| 398 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 399 |
+
"rewards/chosen": 0.4827202260494232,
|
| 400 |
+
"rewards/margins": 0.04902523010969162,
|
| 401 |
+
"rewards/rejected": 0.433694988489151,
|
| 402 |
+
"step": 240
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"epoch": 1.7373913043478262,
|
| 406 |
+
"grad_norm": 47.0,
|
| 407 |
+
"learning_rate": 2.277629844270543e-07,
|
| 408 |
+
"logits/chosen": -1.1127703189849854,
|
| 409 |
+
"logits/rejected": -1.119107961654663,
|
| 410 |
+
"logps/chosen": -313.205810546875,
|
| 411 |
+
"logps/rejected": -302.2017822265625,
|
| 412 |
+
"loss": 0.6791357040405274,
|
| 413 |
+
"rewards/accuracies": 0.59375,
|
| 414 |
+
"rewards/chosen": 0.46044379472732544,
|
| 415 |
+
"rewards/margins": 0.04207334667444229,
|
| 416 |
+
"rewards/rejected": 0.41837042570114136,
|
| 417 |
+
"step": 250
|
| 418 |
+
},
|
| 419 |
+
{
|
| 420 |
+
"epoch": 1.8069565217391306,
|
| 421 |
+
"grad_norm": 74.5,
|
| 422 |
+
"learning_rate": 2.0769587416192208e-07,
|
| 423 |
+
"logits/chosen": -1.0957348346710205,
|
| 424 |
+
"logits/rejected": -1.0845588445663452,
|
| 425 |
+
"logps/chosen": -326.6529235839844,
|
| 426 |
+
"logps/rejected": -328.163330078125,
|
| 427 |
+
"loss": 0.6897631645202636,
|
| 428 |
+
"rewards/accuracies": 0.550000011920929,
|
| 429 |
+
"rewards/chosen": 0.4732390344142914,
|
| 430 |
+
"rewards/margins": 0.017790159210562706,
|
| 431 |
+
"rewards/rejected": 0.4554489552974701,
|
| 432 |
+
"step": 260
|
| 433 |
+
},
|
| 434 |
+
{
|
| 435 |
+
"epoch": 1.8765217391304347,
|
| 436 |
+
"grad_norm": 54.25,
|
| 437 |
+
"learning_rate": 1.8790595656469626e-07,
|
| 438 |
+
"logits/chosen": -1.0816611051559448,
|
| 439 |
+
"logits/rejected": -1.0837030410766602,
|
| 440 |
+
"logps/chosen": -331.3684387207031,
|
| 441 |
+
"logps/rejected": -313.75445556640625,
|
| 442 |
+
"loss": 0.6794800758361816,
|
| 443 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 444 |
+
"rewards/chosen": 0.4701307415962219,
|
| 445 |
+
"rewards/margins": 0.03732570633292198,
|
| 446 |
+
"rewards/rejected": 0.43280500173568726,
|
| 447 |
+
"step": 270
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 1.9460869565217391,
|
| 451 |
+
"grad_norm": 50.5,
|
| 452 |
+
"learning_rate": 1.6852290267685928e-07,
|
| 453 |
+
"logits/chosen": -1.0794267654418945,
|
| 454 |
+
"logits/rejected": -1.0937658548355103,
|
| 455 |
+
"logps/chosen": -331.467529296875,
|
| 456 |
+
"logps/rejected": -321.3943786621094,
|
| 457 |
+
"loss": 0.6855987071990967,
|
| 458 |
+
"rewards/accuracies": 0.543749988079071,
|
| 459 |
+
"rewards/chosen": 0.48015865683555603,
|
| 460 |
+
"rewards/margins": 0.03580061346292496,
|
| 461 |
+
"rewards/rejected": 0.44435805082321167,
|
| 462 |
+
"step": 280
|
| 463 |
+
},
|
| 464 |
+
{
|
| 465 |
+
"epoch": 2.013913043478261,
|
| 466 |
+
"grad_norm": 45.25,
|
| 467 |
+
"learning_rate": 1.4967371761464737e-07,
|
| 468 |
+
"logits/chosen": -1.0743939876556396,
|
| 469 |
+
"logits/rejected": -1.0771441459655762,
|
| 470 |
+
"logps/chosen": -322.74835205078125,
|
| 471 |
+
"logps/rejected": -319.2298278808594,
|
| 472 |
+
"loss": 0.690724229812622,
|
| 473 |
+
"rewards/accuracies": 0.4871794879436493,
|
| 474 |
+
"rewards/chosen": 0.4530966281890869,
|
| 475 |
+
"rewards/margins": 0.012794151902198792,
|
| 476 |
+
"rewards/rejected": 0.4403024911880493,
|
| 477 |
+
"step": 290
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 2.0834782608695654,
|
| 481 |
+
"grad_norm": 45.5,
|
| 482 |
+
"learning_rate": 1.3148190838338802e-07,
|
| 483 |
+
"logits/chosen": -1.0955299139022827,
|
| 484 |
+
"logits/rejected": -1.1200354099273682,
|
| 485 |
+
"logps/chosen": -331.6620178222656,
|
| 486 |
+
"logps/rejected": -313.5164489746094,
|
| 487 |
+
"loss": 0.6698547840118408,
|
| 488 |
+
"rewards/accuracies": 0.612500011920929,
|
| 489 |
+
"rewards/chosen": 0.44486236572265625,
|
| 490 |
+
"rewards/margins": 0.06079784035682678,
|
| 491 |
+
"rewards/rejected": 0.38406452536582947,
|
| 492 |
+
"step": 300
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"epoch": 2.0834782608695654,
|
| 496 |
+
"eval_logits/chosen": -1.0598626136779785,
|
| 497 |
+
"eval_logits/rejected": -1.080617070198059,
|
| 498 |
+
"eval_logps/chosen": -315.2421875,
|
| 499 |
+
"eval_logps/rejected": -316.3223876953125,
|
| 500 |
+
"eval_loss": 0.6872237920761108,
|
| 501 |
+
"eval_rewards/accuracies": 0.5403226017951965,
|
| 502 |
+
"eval_rewards/chosen": 0.42240414023399353,
|
| 503 |
+
"eval_rewards/margins": 0.015609556809067726,
|
| 504 |
+
"eval_rewards/rejected": 0.40679463744163513,
|
| 505 |
+
"eval_runtime": 11.2881,
|
| 506 |
+
"eval_samples_per_second": 10.719,
|
| 507 |
+
"eval_steps_per_second": 2.746,
|
| 508 |
+
"step": 300
|
| 509 |
+
}
|
| 510 |
+
],
|
| 511 |
+
"logging_steps": 10,
|
| 512 |
+
"max_steps": 432,
|
| 513 |
+
"num_input_tokens_seen": 0,
|
| 514 |
+
"num_train_epochs": 3,
|
| 515 |
+
"save_steps": 100,
|
| 516 |
+
"stateful_callbacks": {
|
| 517 |
+
"TrainerControl": {
|
| 518 |
+
"args": {
|
| 519 |
+
"should_epoch_stop": false,
|
| 520 |
+
"should_evaluate": false,
|
| 521 |
+
"should_log": false,
|
| 522 |
+
"should_save": true,
|
| 523 |
+
"should_training_stop": false
|
| 524 |
+
},
|
| 525 |
+
"attributes": {}
|
| 526 |
+
}
|
| 527 |
+
},
|
| 528 |
+
"total_flos": 0.0,
|
| 529 |
+
"train_batch_size": 4,
|
| 530 |
+
"trial_name": null,
|
| 531 |
+
"trial_params": null
|
| 532 |
+
}
|
checkpoint-300/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b05394a09844ad8fbc5bf2ac3f5b6be474605707ae7d97f68562a3f9312197c4
|
| 3 |
+
size 6225
|
checkpoint-400/chat_template.jinja
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{- bos_token -}}
|
| 2 |
+
{%- set keep_past_thinking = keep_past_thinking | default(false) -%}
|
| 3 |
+
{%- set ns = namespace(system_prompt="") -%}
|
| 4 |
+
{%- if messages[0]["role"] == "system" -%}
|
| 5 |
+
{%- set ns.system_prompt = messages[0]["content"] -%}
|
| 6 |
+
{%- set messages = messages[1:] -%}
|
| 7 |
+
{%- endif -%}
|
| 8 |
+
{%- if tools -%}
|
| 9 |
+
{%- set ns.system_prompt = ns.system_prompt + ("\n" if ns.system_prompt else "") + "List of tools: [" -%}
|
| 10 |
+
{%- for tool in tools -%}
|
| 11 |
+
{%- if tool is not string -%}
|
| 12 |
+
{%- set tool = tool | tojson -%}
|
| 13 |
+
{%- endif -%}
|
| 14 |
+
{%- set ns.system_prompt = ns.system_prompt + tool -%}
|
| 15 |
+
{%- if not loop.last -%}
|
| 16 |
+
{%- set ns.system_prompt = ns.system_prompt + ", " -%}
|
| 17 |
+
{%- endif -%}
|
| 18 |
+
{%- endfor -%}
|
| 19 |
+
{%- set ns.system_prompt = ns.system_prompt + "]" -%}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if ns.system_prompt -%}
|
| 22 |
+
{{- "<|im_start|>system\n" + ns.system_prompt + "<|im_end|>\n" -}}
|
| 23 |
+
{%- endif -%}
|
| 24 |
+
{%- set ns.last_assistant_index = -1 -%}
|
| 25 |
+
{%- for message in messages -%}
|
| 26 |
+
{%- if message["role"] == "assistant" -%}
|
| 27 |
+
{%- set ns.last_assistant_index = loop.index0 -%}
|
| 28 |
+
{%- endif -%}
|
| 29 |
+
{%- endfor -%}
|
| 30 |
+
{%- for message in messages -%}
|
| 31 |
+
{{- "<|im_start|>" + message["role"] + "\n" -}}
|
| 32 |
+
{%- set content = message["content"] -%}
|
| 33 |
+
{%- if content is not string -%}
|
| 34 |
+
{%- set content = content | tojson -%}
|
| 35 |
+
{%- endif -%}
|
| 36 |
+
{%- if message["role"] == "assistant" and not keep_past_thinking and loop.index0 != ns.last_assistant_index -%}
|
| 37 |
+
{%- if "</think>" in content -%}
|
| 38 |
+
{%- set content = content.split("</think>")[-1] | trim -%}
|
| 39 |
+
{%- endif -%}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{- content + "<|im_end|>\n" -}}
|
| 42 |
+
{%- endfor -%}
|
| 43 |
+
{%- if add_generation_prompt -%}
|
| 44 |
+
{{- "<|im_start|>assistant\n" -}}
|
| 45 |
+
{%- endif -%}
|
checkpoint-400/config.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Lfm2ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"block_auto_adjust_ff_dim": true,
|
| 6 |
+
"block_dim": 2048,
|
| 7 |
+
"block_ff_dim": 12288,
|
| 8 |
+
"block_ffn_dim_multiplier": 1.0,
|
| 9 |
+
"block_mlp_init_scale": 1.0,
|
| 10 |
+
"block_multiple_of": 256,
|
| 11 |
+
"block_norm_eps": 1e-05,
|
| 12 |
+
"block_out_init_scale": 1.0,
|
| 13 |
+
"block_use_swiglu": true,
|
| 14 |
+
"block_use_xavier_init": true,
|
| 15 |
+
"bos_token_id": 1,
|
| 16 |
+
"conv_L_cache": 3,
|
| 17 |
+
"conv_bias": false,
|
| 18 |
+
"conv_dim": 2048,
|
| 19 |
+
"conv_use_xavier_init": true,
|
| 20 |
+
"dtype": "bfloat16",
|
| 21 |
+
"eos_token_id": 7,
|
| 22 |
+
"hidden_size": 2048,
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 12288,
|
| 25 |
+
"layer_types": [
|
| 26 |
+
"conv",
|
| 27 |
+
"conv",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"conv",
|
| 30 |
+
"conv",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"conv",
|
| 33 |
+
"conv",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"conv",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"conv",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"conv",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"conv"
|
| 42 |
+
],
|
| 43 |
+
"max_position_embeddings": 128000,
|
| 44 |
+
"model_type": "lfm2",
|
| 45 |
+
"norm_eps": 1e-05,
|
| 46 |
+
"num_attention_heads": 32,
|
| 47 |
+
"num_heads": 32,
|
| 48 |
+
"num_hidden_layers": 16,
|
| 49 |
+
"num_key_value_heads": 8,
|
| 50 |
+
"pad_token_id": 0,
|
| 51 |
+
"rope_parameters": {
|
| 52 |
+
"rope_theta": 1000000.0,
|
| 53 |
+
"rope_type": "default"
|
| 54 |
+
},
|
| 55 |
+
"tie_embedding": true,
|
| 56 |
+
"tie_word_embeddings": true,
|
| 57 |
+
"transformers_version": "5.2.0",
|
| 58 |
+
"use_cache": false,
|
| 59 |
+
"use_pos_enc": true,
|
| 60 |
+
"vocab_size": 65536
|
| 61 |
+
}
|
checkpoint-400/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
7
|
| 6 |
+
],
|
| 7 |
+
"pad_token_id": 0,
|
| 8 |
+
"transformers_version": "5.2.0"
|
| 9 |
+
}
|
checkpoint-400/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a7c37aae343f39e5bff72a9e4c3f50518a498c1331c30c85404a17ed984d89c
|
| 3 |
+
size 2340697936
|
checkpoint-400/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5482a323203561585dedf00d2fbec8a8a40881ac4d483182c9dc68142423a846
|
| 3 |
+
size 3178927435
|
checkpoint-400/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f196323d7423b60f8e4ceb7dbf8715ee326c0d068e5ff164f13c63b279b9f1a0
|
| 3 |
+
size 14645
|
checkpoint-400/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:58d2a6b6ae30d2c6082649e4dba1804265132f00e29fa400bec0b33445a55f46
|
| 3 |
+
size 1465
|
checkpoint-400/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-400/tokenizer_config.json
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"bos_token": "<|startoftext|>",
|
| 4 |
+
"clean_up_tokenization_spaces": false,
|
| 5 |
+
"eos_token": "<|im_end|>",
|
| 6 |
+
"is_local": false,
|
| 7 |
+
"legacy": false,
|
| 8 |
+
"model_input_names": [
|
| 9 |
+
"input_ids",
|
| 10 |
+
"attention_mask"
|
| 11 |
+
],
|
| 12 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 13 |
+
"pad_token": "<|pad|>",
|
| 14 |
+
"padding_side": "right",
|
| 15 |
+
"sp_model_kwargs": {},
|
| 16 |
+
"spaces_between_special_tokens": false,
|
| 17 |
+
"tokenizer_class": "TokenizersBackend",
|
| 18 |
+
"use_default_system_prompt": false,
|
| 19 |
+
"use_fast": true
|
| 20 |
+
}
|
checkpoint-400/trainer_state.json
ADDED
|
@@ -0,0 +1,698 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 300,
|
| 3 |
+
"best_metric": 0.6872237920761108,
|
| 4 |
+
"best_model_checkpoint": "models/dpo_fft_LFM2.5-1.2B-Instruct_argilla__distilabel-math-preference-dpo_20260222_195126/checkpoint-300",
|
| 5 |
+
"epoch": 2.7791304347826085,
|
| 6 |
+
"eval_steps": 100,
|
| 7 |
+
"global_step": 400,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.06956521739130435,
|
| 14 |
+
"grad_norm": 44.25,
|
| 15 |
+
"learning_rate": 1.0227272727272728e-07,
|
| 16 |
+
"logits/chosen": -1.0599091053009033,
|
| 17 |
+
"logits/rejected": -1.042307734489441,
|
| 18 |
+
"logps/chosen": -332.13507080078125,
|
| 19 |
+
"logps/rejected": -332.9361267089844,
|
| 20 |
+
"loss": 0.6927361965179444,
|
| 21 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 22 |
+
"rewards/chosen": 0.01432617288082838,
|
| 23 |
+
"rewards/margins": 0.0042648655362427235,
|
| 24 |
+
"rewards/rejected": 0.010061310604214668,
|
| 25 |
+
"step": 10
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.1391304347826087,
|
| 29 |
+
"grad_norm": 50.5,
|
| 30 |
+
"learning_rate": 2.159090909090909e-07,
|
| 31 |
+
"logits/chosen": -1.0978257656097412,
|
| 32 |
+
"logits/rejected": -1.09658682346344,
|
| 33 |
+
"logps/chosen": -329.22100830078125,
|
| 34 |
+
"logps/rejected": -312.2664794921875,
|
| 35 |
+
"loss": 0.7003747940063476,
|
| 36 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 37 |
+
"rewards/chosen": 0.004466252867132425,
|
| 38 |
+
"rewards/margins": -0.011238865554332733,
|
| 39 |
+
"rewards/rejected": 0.01570511981844902,
|
| 40 |
+
"step": 20
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"epoch": 0.20869565217391303,
|
| 44 |
+
"grad_norm": 54.75,
|
| 45 |
+
"learning_rate": 3.295454545454545e-07,
|
| 46 |
+
"logits/chosen": -1.1366004943847656,
|
| 47 |
+
"logits/rejected": -1.1707069873809814,
|
| 48 |
+
"logps/chosen": -328.1323547363281,
|
| 49 |
+
"logps/rejected": -305.90618896484375,
|
| 50 |
+
"loss": 0.7031519412994385,
|
| 51 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 52 |
+
"rewards/chosen": 0.03125340864062309,
|
| 53 |
+
"rewards/margins": -0.01610126718878746,
|
| 54 |
+
"rewards/rejected": 0.04735467582941055,
|
| 55 |
+
"step": 30
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 0.2782608695652174,
|
| 59 |
+
"grad_norm": 50.5,
|
| 60 |
+
"learning_rate": 4.4318181818181817e-07,
|
| 61 |
+
"logits/chosen": -1.0889461040496826,
|
| 62 |
+
"logits/rejected": -1.1017862558364868,
|
| 63 |
+
"logps/chosen": -338.80938720703125,
|
| 64 |
+
"logps/rejected": -328.07464599609375,
|
| 65 |
+
"loss": 0.6893259048461914,
|
| 66 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 67 |
+
"rewards/chosen": 0.09773217141628265,
|
| 68 |
+
"rewards/margins": 0.01154844556003809,
|
| 69 |
+
"rewards/rejected": 0.08618371188640594,
|
| 70 |
+
"step": 40
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"epoch": 0.34782608695652173,
|
| 74 |
+
"grad_norm": 53.5,
|
| 75 |
+
"learning_rate": 4.997951542310825e-07,
|
| 76 |
+
"logits/chosen": -1.1367381811141968,
|
| 77 |
+
"logits/rejected": -1.1468210220336914,
|
| 78 |
+
"logps/chosen": -329.12823486328125,
|
| 79 |
+
"logps/rejected": -323.8036804199219,
|
| 80 |
+
"loss": 0.6982084751129151,
|
| 81 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 82 |
+
"rewards/chosen": 0.1568203717470169,
|
| 83 |
+
"rewards/margins": -0.004218918737024069,
|
| 84 |
+
"rewards/rejected": 0.1610392928123474,
|
| 85 |
+
"step": 50
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"epoch": 0.41739130434782606,
|
| 89 |
+
"grad_norm": 64.0,
|
| 90 |
+
"learning_rate": 4.98158401695492e-07,
|
| 91 |
+
"logits/chosen": -1.0899083614349365,
|
| 92 |
+
"logits/rejected": -1.0698351860046387,
|
| 93 |
+
"logps/chosen": -335.6194152832031,
|
| 94 |
+
"logps/rejected": -329.00653076171875,
|
| 95 |
+
"loss": 0.6905360221862793,
|
| 96 |
+
"rewards/accuracies": 0.53125,
|
| 97 |
+
"rewards/chosen": 0.1543850153684616,
|
| 98 |
+
"rewards/margins": 0.010315253399312496,
|
| 99 |
+
"rewards/rejected": 0.14406974613666534,
|
| 100 |
+
"step": 60
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.48695652173913045,
|
| 104 |
+
"grad_norm": 72.5,
|
| 105 |
+
"learning_rate": 4.948956212473369e-07,
|
| 106 |
+
"logits/chosen": -1.1223671436309814,
|
| 107 |
+
"logits/rejected": -1.113906741142273,
|
| 108 |
+
"logps/chosen": -319.424560546875,
|
| 109 |
+
"logps/rejected": -317.76934814453125,
|
| 110 |
+
"loss": 0.7009393215179444,
|
| 111 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 112 |
+
"rewards/chosen": 0.18402941524982452,
|
| 113 |
+
"rewards/margins": -0.010006649419665337,
|
| 114 |
+
"rewards/rejected": 0.19403605163097382,
|
| 115 |
+
"step": 70
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.5565217391304348,
|
| 119 |
+
"grad_norm": 52.5,
|
| 120 |
+
"learning_rate": 4.900281918608732e-07,
|
| 121 |
+
"logits/chosen": -1.1157792806625366,
|
| 122 |
+
"logits/rejected": -1.1078118085861206,
|
| 123 |
+
"logps/chosen": -314.5404052734375,
|
| 124 |
+
"logps/rejected": -311.91656494140625,
|
| 125 |
+
"loss": 0.6877760887145996,
|
| 126 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 127 |
+
"rewards/chosen": 0.19012701511383057,
|
| 128 |
+
"rewards/margins": 0.016789216548204422,
|
| 129 |
+
"rewards/rejected": 0.17333778738975525,
|
| 130 |
+
"step": 80
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.6260869565217392,
|
| 134 |
+
"grad_norm": 47.75,
|
| 135 |
+
"learning_rate": 4.835880067784441e-07,
|
| 136 |
+
"logits/chosen": -1.0738165378570557,
|
| 137 |
+
"logits/rejected": -1.0576341152191162,
|
| 138 |
+
"logps/chosen": -336.3821716308594,
|
| 139 |
+
"logps/rejected": -331.9998779296875,
|
| 140 |
+
"loss": 0.6956596374511719,
|
| 141 |
+
"rewards/accuracies": 0.4375,
|
| 142 |
+
"rewards/chosen": 0.21225734055042267,
|
| 143 |
+
"rewards/margins": 0.0009234324097633362,
|
| 144 |
+
"rewards/rejected": 0.21133390069007874,
|
| 145 |
+
"step": 90
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.6956521739130435,
|
| 149 |
+
"grad_norm": 49.0,
|
| 150 |
+
"learning_rate": 4.7561726453386744e-07,
|
| 151 |
+
"logits/chosen": -1.0457407236099243,
|
| 152 |
+
"logits/rejected": -1.038447618484497,
|
| 153 |
+
"logps/chosen": -330.8192138671875,
|
| 154 |
+
"logps/rejected": -323.5484924316406,
|
| 155 |
+
"loss": 0.6817105770111084,
|
| 156 |
+
"rewards/accuracies": 0.581250011920929,
|
| 157 |
+
"rewards/chosen": 0.26413798332214355,
|
| 158 |
+
"rewards/margins": 0.029450010508298874,
|
| 159 |
+
"rewards/rejected": 0.2346879541873932,
|
| 160 |
+
"step": 100
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.6956521739130435,
|
| 164 |
+
"eval_logits/chosen": -1.0708128213882446,
|
| 165 |
+
"eval_logits/rejected": -1.0917303562164307,
|
| 166 |
+
"eval_logps/chosen": -317.1337585449219,
|
| 167 |
+
"eval_logps/rejected": -318.0965270996094,
|
| 168 |
+
"eval_loss": 0.6924862265586853,
|
| 169 |
+
"eval_rewards/accuracies": 0.4838709533214569,
|
| 170 |
+
"eval_rewards/chosen": 0.2332489788532257,
|
| 171 |
+
"eval_rewards/margins": 0.0038641756400465965,
|
| 172 |
+
"eval_rewards/rejected": 0.22938479483127594,
|
| 173 |
+
"eval_runtime": 11.3087,
|
| 174 |
+
"eval_samples_per_second": 10.7,
|
| 175 |
+
"eval_steps_per_second": 2.741,
|
| 176 |
+
"step": 100
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 0.7652173913043478,
|
| 180 |
+
"grad_norm": 47.0,
|
| 181 |
+
"learning_rate": 4.6616819245164655e-07,
|
| 182 |
+
"logits/chosen": -1.0736799240112305,
|
| 183 |
+
"logits/rejected": -1.0783708095550537,
|
| 184 |
+
"logps/chosen": -314.97943115234375,
|
| 185 |
+
"logps/rejected": -308.35675048828125,
|
| 186 |
+
"loss": 0.6984636306762695,
|
| 187 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 188 |
+
"rewards/chosen": 0.22377920150756836,
|
| 189 |
+
"rewards/margins": -0.0050087496638298035,
|
| 190 |
+
"rewards/rejected": 0.22878792881965637,
|
| 191 |
+
"step": 110
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.8347826086956521,
|
| 195 |
+
"grad_norm": 51.25,
|
| 196 |
+
"learning_rate": 4.55302704433743e-07,
|
| 197 |
+
"logits/chosen": -1.048730731010437,
|
| 198 |
+
"logits/rejected": -1.0437910556793213,
|
| 199 |
+
"logps/chosen": -332.6585388183594,
|
| 200 |
+
"logps/rejected": -319.9508361816406,
|
| 201 |
+
"loss": 0.6809643268585205,
|
| 202 |
+
"rewards/accuracies": 0.59375,
|
| 203 |
+
"rewards/chosen": 0.3183898627758026,
|
| 204 |
+
"rewards/margins": 0.03039008378982544,
|
| 205 |
+
"rewards/rejected": 0.2879997789859772,
|
| 206 |
+
"step": 120
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.9043478260869565,
|
| 210 |
+
"grad_norm": 57.5,
|
| 211 |
+
"learning_rate": 4.4309199527622254e-07,
|
| 212 |
+
"logits/chosen": -1.0928480625152588,
|
| 213 |
+
"logits/rejected": -1.0909947156906128,
|
| 214 |
+
"logps/chosen": -333.54510498046875,
|
| 215 |
+
"logps/rejected": -321.5794372558594,
|
| 216 |
+
"loss": 0.6884570121765137,
|
| 217 |
+
"rewards/accuracies": 0.518750011920929,
|
| 218 |
+
"rewards/chosen": 0.3165169656276703,
|
| 219 |
+
"rewards/margins": 0.015730690211057663,
|
| 220 |
+
"rewards/rejected": 0.3007862865924835,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.9739130434782609,
|
| 225 |
+
"grad_norm": 50.75,
|
| 226 |
+
"learning_rate": 4.2961607417396517e-07,
|
| 227 |
+
"logits/chosen": -1.0702852010726929,
|
| 228 |
+
"logits/rejected": -1.0860363245010376,
|
| 229 |
+
"logps/chosen": -336.94122314453125,
|
| 230 |
+
"logps/rejected": -312.6028137207031,
|
| 231 |
+
"loss": 0.6929869651794434,
|
| 232 |
+
"rewards/accuracies": 0.5,
|
| 233 |
+
"rewards/chosen": 0.3366253077983856,
|
| 234 |
+
"rewards/margins": 0.008591671474277973,
|
| 235 |
+
"rewards/rejected": 0.3280336260795593,
|
| 236 |
+
"step": 140
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 1.0417391304347827,
|
| 240 |
+
"grad_norm": 58.25,
|
| 241 |
+
"learning_rate": 4.1496324047009244e-07,
|
| 242 |
+
"logits/chosen": -1.137653112411499,
|
| 243 |
+
"logits/rejected": -1.1385780572891235,
|
| 244 |
+
"logps/chosen": -297.7463073730469,
|
| 245 |
+
"logps/rejected": -304.8460388183594,
|
| 246 |
+
"loss": 0.6796726703643798,
|
| 247 |
+
"rewards/accuracies": 0.5320512652397156,
|
| 248 |
+
"rewards/chosen": 0.36476510763168335,
|
| 249 |
+
"rewards/margins": 0.029467487707734108,
|
| 250 |
+
"rewards/rejected": 0.3352976143360138,
|
| 251 |
+
"step": 150
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.111304347826087,
|
| 255 |
+
"grad_norm": 62.0,
|
| 256 |
+
"learning_rate": 3.9922950508520126e-07,
|
| 257 |
+
"logits/chosen": -1.0764613151550293,
|
| 258 |
+
"logits/rejected": -1.073861837387085,
|
| 259 |
+
"logps/chosen": -325.74151611328125,
|
| 260 |
+
"logps/rejected": -314.5641174316406,
|
| 261 |
+
"loss": 0.6664072513580322,
|
| 262 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 263 |
+
"rewards/chosen": 0.37418311834335327,
|
| 264 |
+
"rewards/margins": 0.06225377321243286,
|
| 265 |
+
"rewards/rejected": 0.3119293451309204,
|
| 266 |
+
"step": 160
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.1808695652173913,
|
| 270 |
+
"grad_norm": 48.25,
|
| 271 |
+
"learning_rate": 3.8251796141741945e-07,
|
| 272 |
+
"logits/chosen": -1.1032686233520508,
|
| 273 |
+
"logits/rejected": -1.1010105609893799,
|
| 274 |
+
"logps/chosen": -324.8043212890625,
|
| 275 |
+
"logps/rejected": -314.7898254394531,
|
| 276 |
+
"loss": 0.6656156063079834,
|
| 277 |
+
"rewards/accuracies": 0.643750011920929,
|
| 278 |
+
"rewards/chosen": 0.39548203349113464,
|
| 279 |
+
"rewards/margins": 0.06372012943029404,
|
| 280 |
+
"rewards/rejected": 0.33176189661026,
|
| 281 |
+
"step": 170
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 1.2504347826086957,
|
| 285 |
+
"grad_norm": 73.5,
|
| 286 |
+
"learning_rate": 3.649381098353834e-07,
|
| 287 |
+
"logits/chosen": -1.108781099319458,
|
| 288 |
+
"logits/rejected": -1.081012487411499,
|
| 289 |
+
"logps/chosen": -333.0024719238281,
|
| 290 |
+
"logps/rejected": -317.6542663574219,
|
| 291 |
+
"loss": 0.6682669162750244,
|
| 292 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 293 |
+
"rewards/chosen": 0.4114314019680023,
|
| 294 |
+
"rewards/margins": 0.06464624404907227,
|
| 295 |
+
"rewards/rejected": 0.34678512811660767,
|
| 296 |
+
"step": 180
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 1.32,
|
| 300 |
+
"grad_norm": 51.75,
|
| 301 |
+
"learning_rate": 3.466051401903162e-07,
|
| 302 |
+
"logits/chosen": -1.0279661417007446,
|
| 303 |
+
"logits/rejected": -1.028050422668457,
|
| 304 |
+
"logps/chosen": -342.05322265625,
|
| 305 |
+
"logps/rejected": -329.083984375,
|
| 306 |
+
"loss": 0.6815151691436767,
|
| 307 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 308 |
+
"rewards/chosen": 0.4475061893463135,
|
| 309 |
+
"rewards/margins": 0.03387105464935303,
|
| 310 |
+
"rewards/rejected": 0.41363516449928284,
|
| 311 |
+
"step": 190
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"epoch": 1.3895652173913042,
|
| 315 |
+
"grad_norm": 43.25,
|
| 316 |
+
"learning_rate": 3.276391770484606e-07,
|
| 317 |
+
"logits/chosen": -1.0554686784744263,
|
| 318 |
+
"logits/rejected": -1.0444362163543701,
|
| 319 |
+
"logps/chosen": -323.93359375,
|
| 320 |
+
"logps/rejected": -300.8354187011719,
|
| 321 |
+
"loss": 0.6729463577270508,
|
| 322 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 323 |
+
"rewards/chosen": 0.4051585793495178,
|
| 324 |
+
"rewards/margins": 0.04842006787657738,
|
| 325 |
+
"rewards/rejected": 0.35673850774765015,
|
| 326 |
+
"step": 200
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 1.3895652173913042,
|
| 330 |
+
"eval_logits/chosen": -1.059758186340332,
|
| 331 |
+
"eval_logits/rejected": -1.0798335075378418,
|
| 332 |
+
"eval_logps/chosen": -315.1194763183594,
|
| 333 |
+
"eval_logps/rejected": -316.16046142578125,
|
| 334 |
+
"eval_loss": 0.6918731331825256,
|
| 335 |
+
"eval_rewards/accuracies": 0.5483871102333069,
|
| 336 |
+
"eval_rewards/chosen": 0.4346778690814972,
|
| 337 |
+
"eval_rewards/margins": 0.011690356768667698,
|
| 338 |
+
"eval_rewards/rejected": 0.42298752069473267,
|
| 339 |
+
"eval_runtime": 11.2737,
|
| 340 |
+
"eval_samples_per_second": 10.733,
|
| 341 |
+
"eval_steps_per_second": 2.75,
|
| 342 |
+
"step": 200
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 1.4591304347826086,
|
| 346 |
+
"grad_norm": 44.5,
|
| 347 |
+
"learning_rate": 3.0816449258938656e-07,
|
| 348 |
+
"logits/chosen": -1.0483617782592773,
|
| 349 |
+
"logits/rejected": -1.0491857528686523,
|
| 350 |
+
"logps/chosen": -326.4630126953125,
|
| 351 |
+
"logps/rejected": -321.35467529296875,
|
| 352 |
+
"loss": 0.6760656356811523,
|
| 353 |
+
"rewards/accuracies": 0.59375,
|
| 354 |
+
"rewards/chosen": 0.4774439334869385,
|
| 355 |
+
"rewards/margins": 0.04994555562734604,
|
| 356 |
+
"rewards/rejected": 0.42749834060668945,
|
| 357 |
+
"step": 210
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 1.528695652173913,
|
| 361 |
+
"grad_norm": 69.0,
|
| 362 |
+
"learning_rate": 2.883086923275658e-07,
|
| 363 |
+
"logits/chosen": -1.0638173818588257,
|
| 364 |
+
"logits/rejected": -1.0861108303070068,
|
| 365 |
+
"logps/chosen": -323.24517822265625,
|
| 366 |
+
"logps/rejected": -316.22454833984375,
|
| 367 |
+
"loss": 0.6859281063079834,
|
| 368 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 369 |
+
"rewards/chosen": 0.464817613363266,
|
| 370 |
+
"rewards/margins": 0.025601129978895187,
|
| 371 |
+
"rewards/rejected": 0.4392164647579193,
|
| 372 |
+
"step": 220
|
| 373 |
+
},
|
| 374 |
+
{
|
| 375 |
+
"epoch": 1.5982608695652174,
|
| 376 |
+
"grad_norm": 51.25,
|
| 377 |
+
"learning_rate": 2.6820187899267197e-07,
|
| 378 |
+
"logits/chosen": -1.0582925081253052,
|
| 379 |
+
"logits/rejected": -1.0700470209121704,
|
| 380 |
+
"logps/chosen": -331.1497802734375,
|
| 381 |
+
"logps/rejected": -318.2004089355469,
|
| 382 |
+
"loss": 0.6922653675079345,
|
| 383 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 384 |
+
"rewards/chosen": 0.4878065586090088,
|
| 385 |
+
"rewards/margins": 0.014133572578430176,
|
| 386 |
+
"rewards/rejected": 0.4736729562282562,
|
| 387 |
+
"step": 230
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 1.6678260869565218,
|
| 391 |
+
"grad_norm": 44.0,
|
| 392 |
+
"learning_rate": 2.4797580004718036e-07,
|
| 393 |
+
"logits/chosen": -1.059127926826477,
|
| 394 |
+
"logits/rejected": -1.0511342287063599,
|
| 395 |
+
"logps/chosen": -329.36309814453125,
|
| 396 |
+
"logps/rejected": -324.93060302734375,
|
| 397 |
+
"loss": 0.6760355949401855,
|
| 398 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 399 |
+
"rewards/chosen": 0.4827202260494232,
|
| 400 |
+
"rewards/margins": 0.04902523010969162,
|
| 401 |
+
"rewards/rejected": 0.433694988489151,
|
| 402 |
+
"step": 240
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"epoch": 1.7373913043478262,
|
| 406 |
+
"grad_norm": 47.0,
|
| 407 |
+
"learning_rate": 2.277629844270543e-07,
|
| 408 |
+
"logits/chosen": -1.1127703189849854,
|
| 409 |
+
"logits/rejected": -1.119107961654663,
|
| 410 |
+
"logps/chosen": -313.205810546875,
|
| 411 |
+
"logps/rejected": -302.2017822265625,
|
| 412 |
+
"loss": 0.6791357040405274,
|
| 413 |
+
"rewards/accuracies": 0.59375,
|
| 414 |
+
"rewards/chosen": 0.46044379472732544,
|
| 415 |
+
"rewards/margins": 0.04207334667444229,
|
| 416 |
+
"rewards/rejected": 0.41837042570114136,
|
| 417 |
+
"step": 250
|
| 418 |
+
},
|
| 419 |
+
{
|
| 420 |
+
"epoch": 1.8069565217391306,
|
| 421 |
+
"grad_norm": 74.5,
|
| 422 |
+
"learning_rate": 2.0769587416192208e-07,
|
| 423 |
+
"logits/chosen": -1.0957348346710205,
|
| 424 |
+
"logits/rejected": -1.0845588445663452,
|
| 425 |
+
"logps/chosen": -326.6529235839844,
|
| 426 |
+
"logps/rejected": -328.163330078125,
|
| 427 |
+
"loss": 0.6897631645202636,
|
| 428 |
+
"rewards/accuracies": 0.550000011920929,
|
| 429 |
+
"rewards/chosen": 0.4732390344142914,
|
| 430 |
+
"rewards/margins": 0.017790159210562706,
|
| 431 |
+
"rewards/rejected": 0.4554489552974701,
|
| 432 |
+
"step": 260
|
| 433 |
+
},
|
| 434 |
+
{
|
| 435 |
+
"epoch": 1.8765217391304347,
|
| 436 |
+
"grad_norm": 54.25,
|
| 437 |
+
"learning_rate": 1.8790595656469626e-07,
|
| 438 |
+
"logits/chosen": -1.0816611051559448,
|
| 439 |
+
"logits/rejected": -1.0837030410766602,
|
| 440 |
+
"logps/chosen": -331.3684387207031,
|
| 441 |
+
"logps/rejected": -313.75445556640625,
|
| 442 |
+
"loss": 0.6794800758361816,
|
| 443 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 444 |
+
"rewards/chosen": 0.4701307415962219,
|
| 445 |
+
"rewards/margins": 0.03732570633292198,
|
| 446 |
+
"rewards/rejected": 0.43280500173568726,
|
| 447 |
+
"step": 270
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 1.9460869565217391,
|
| 451 |
+
"grad_norm": 50.5,
|
| 452 |
+
"learning_rate": 1.6852290267685928e-07,
|
| 453 |
+
"logits/chosen": -1.0794267654418945,
|
| 454 |
+
"logits/rejected": -1.0937658548355103,
|
| 455 |
+
"logps/chosen": -331.467529296875,
|
| 456 |
+
"logps/rejected": -321.3943786621094,
|
| 457 |
+
"loss": 0.6855987071990967,
|
| 458 |
+
"rewards/accuracies": 0.543749988079071,
|
| 459 |
+
"rewards/chosen": 0.48015865683555603,
|
| 460 |
+
"rewards/margins": 0.03580061346292496,
|
| 461 |
+
"rewards/rejected": 0.44435805082321167,
|
| 462 |
+
"step": 280
|
| 463 |
+
},
|
| 464 |
+
{
|
| 465 |
+
"epoch": 2.013913043478261,
|
| 466 |
+
"grad_norm": 45.25,
|
| 467 |
+
"learning_rate": 1.4967371761464737e-07,
|
| 468 |
+
"logits/chosen": -1.0743939876556396,
|
| 469 |
+
"logits/rejected": -1.0771441459655762,
|
| 470 |
+
"logps/chosen": -322.74835205078125,
|
| 471 |
+
"logps/rejected": -319.2298278808594,
|
| 472 |
+
"loss": 0.690724229812622,
|
| 473 |
+
"rewards/accuracies": 0.4871794879436493,
|
| 474 |
+
"rewards/chosen": 0.4530966281890869,
|
| 475 |
+
"rewards/margins": 0.012794151902198792,
|
| 476 |
+
"rewards/rejected": 0.4403024911880493,
|
| 477 |
+
"step": 290
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 2.0834782608695654,
|
| 481 |
+
"grad_norm": 45.5,
|
| 482 |
+
"learning_rate": 1.3148190838338802e-07,
|
| 483 |
+
"logits/chosen": -1.0955299139022827,
|
| 484 |
+
"logits/rejected": -1.1200354099273682,
|
| 485 |
+
"logps/chosen": -331.6620178222656,
|
| 486 |
+
"logps/rejected": -313.5164489746094,
|
| 487 |
+
"loss": 0.6698547840118408,
|
| 488 |
+
"rewards/accuracies": 0.612500011920929,
|
| 489 |
+
"rewards/chosen": 0.44486236572265625,
|
| 490 |
+
"rewards/margins": 0.06079784035682678,
|
| 491 |
+
"rewards/rejected": 0.38406452536582947,
|
| 492 |
+
"step": 300
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"epoch": 2.0834782608695654,
|
| 496 |
+
"eval_logits/chosen": -1.0598626136779785,
|
| 497 |
+
"eval_logits/rejected": -1.080617070198059,
|
| 498 |
+
"eval_logps/chosen": -315.2421875,
|
| 499 |
+
"eval_logps/rejected": -316.3223876953125,
|
| 500 |
+
"eval_loss": 0.6872237920761108,
|
| 501 |
+
"eval_rewards/accuracies": 0.5403226017951965,
|
| 502 |
+
"eval_rewards/chosen": 0.42240414023399353,
|
| 503 |
+
"eval_rewards/margins": 0.015609556809067726,
|
| 504 |
+
"eval_rewards/rejected": 0.40679463744163513,
|
| 505 |
+
"eval_runtime": 11.2881,
|
| 506 |
+
"eval_samples_per_second": 10.719,
|
| 507 |
+
"eval_steps_per_second": 2.746,
|
| 508 |
+
"step": 300
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"epoch": 2.1530434782608694,
|
| 512 |
+
"grad_norm": 47.5,
|
| 513 |
+
"learning_rate": 1.1406667461278538e-07,
|
| 514 |
+
"logits/chosen": -1.0953623056411743,
|
| 515 |
+
"logits/rejected": -1.0638518333435059,
|
| 516 |
+
"logps/chosen": -317.5690002441406,
|
| 517 |
+
"logps/rejected": -313.86895751953125,
|
| 518 |
+
"loss": 0.6786671161651612,
|
| 519 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 520 |
+
"rewards/chosen": 0.43794316053390503,
|
| 521 |
+
"rewards/margins": 0.040073662996292114,
|
| 522 |
+
"rewards/rejected": 0.39786943793296814,
|
| 523 |
+
"step": 310
|
| 524 |
+
},
|
| 525 |
+
{
|
| 526 |
+
"epoch": 2.222608695652174,
|
| 527 |
+
"grad_norm": 75.5,
|
| 528 |
+
"learning_rate": 9.754212751576385e-08,
|
| 529 |
+
"logits/chosen": -1.036792516708374,
|
| 530 |
+
"logits/rejected": -1.0298279523849487,
|
| 531 |
+
"logps/chosen": -322.68511962890625,
|
| 532 |
+
"logps/rejected": -324.95562744140625,
|
| 533 |
+
"loss": 0.6964206695556641,
|
| 534 |
+
"rewards/accuracies": 0.5,
|
| 535 |
+
"rewards/chosen": 0.4462190568447113,
|
| 536 |
+
"rewards/margins": 0.005142434034496546,
|
| 537 |
+
"rewards/rejected": 0.4410766065120697,
|
| 538 |
+
"step": 320
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"epoch": 2.292173913043478,
|
| 542 |
+
"grad_norm": 50.0,
|
| 543 |
+
"learning_rate": 8.201654218854689e-08,
|
| 544 |
+
"logits/chosen": -1.076971173286438,
|
| 545 |
+
"logits/rejected": -1.0646440982818604,
|
| 546 |
+
"logps/chosen": -314.86492919921875,
|
| 547 |
+
"logps/rejected": -309.55389404296875,
|
| 548 |
+
"loss": 0.6790688037872314,
|
| 549 |
+
"rewards/accuracies": 0.581250011920929,
|
| 550 |
+
"rewards/chosen": 0.4310247302055359,
|
| 551 |
+
"rewards/margins": 0.03967974707484245,
|
| 552 |
+
"rewards/rejected": 0.39134499430656433,
|
| 553 |
+
"step": 330
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 2.3617391304347826,
|
| 557 |
+
"grad_norm": 48.0,
|
| 558 |
+
"learning_rate": 6.759164815118493e-08,
|
| 559 |
+
"logits/chosen": -1.0706030130386353,
|
| 560 |
+
"logits/rejected": -1.0911849737167358,
|
| 561 |
+
"logps/chosen": -337.18853759765625,
|
| 562 |
+
"logps/rejected": -322.5784912109375,
|
| 563 |
+
"loss": 0.6748550415039063,
|
| 564 |
+
"rewards/accuracies": 0.625,
|
| 565 |
+
"rewards/chosen": 0.5078705549240112,
|
| 566 |
+
"rewards/margins": 0.05118296295404434,
|
| 567 |
+
"rewards/rejected": 0.4566876292228699,
|
| 568 |
+
"step": 340
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"epoch": 2.431304347826087,
|
| 572 |
+
"grad_norm": 60.5,
|
| 573 |
+
"learning_rate": 5.4361962777179273e-08,
|
| 574 |
+
"logits/chosen": -1.0652002096176147,
|
| 575 |
+
"logits/rejected": -1.0517784357070923,
|
| 576 |
+
"logps/chosen": -330.02325439453125,
|
| 577 |
+
"logps/rejected": -326.01873779296875,
|
| 578 |
+
"loss": 0.691571044921875,
|
| 579 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 580 |
+
"rewards/chosen": 0.46054545044898987,
|
| 581 |
+
"rewards/margins": 0.012338491156697273,
|
| 582 |
+
"rewards/rejected": 0.44820699095726013,
|
| 583 |
+
"step": 350
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"epoch": 2.5008695652173913,
|
| 587 |
+
"grad_norm": 51.75,
|
| 588 |
+
"learning_rate": 4.2414171979824e-08,
|
| 589 |
+
"logits/chosen": -1.0523455142974854,
|
| 590 |
+
"logits/rejected": -1.0446560382843018,
|
| 591 |
+
"logps/chosen": -336.7765197753906,
|
| 592 |
+
"logps/rejected": -316.00067138671875,
|
| 593 |
+
"loss": 0.6638113975524902,
|
| 594 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 595 |
+
"rewards/chosen": 0.46467241644859314,
|
| 596 |
+
"rewards/margins": 0.06902624666690826,
|
| 597 |
+
"rewards/rejected": 0.39564621448516846,
|
| 598 |
+
"step": 360
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"epoch": 2.5704347826086957,
|
| 602 |
+
"grad_norm": 49.0,
|
| 603 |
+
"learning_rate": 3.1826562213243844e-08,
|
| 604 |
+
"logits/chosen": -1.0767319202423096,
|
| 605 |
+
"logits/rejected": -1.0776704549789429,
|
| 606 |
+
"logps/chosen": -329.4648742675781,
|
| 607 |
+
"logps/rejected": -319.5013732910156,
|
| 608 |
+
"loss": 0.6690452575683594,
|
| 609 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 610 |
+
"rewards/chosen": 0.46328702569007874,
|
| 611 |
+
"rewards/margins": 0.061441998928785324,
|
| 612 |
+
"rewards/rejected": 0.40184497833251953,
|
| 613 |
+
"step": 370
|
| 614 |
+
},
|
| 615 |
+
{
|
| 616 |
+
"epoch": 2.64,
|
| 617 |
+
"grad_norm": 56.5,
|
| 618 |
+
"learning_rate": 2.2668507509871953e-08,
|
| 619 |
+
"logits/chosen": -1.1089640855789185,
|
| 620 |
+
"logits/rejected": -1.0823930501937866,
|
| 621 |
+
"logps/chosen": -328.79351806640625,
|
| 622 |
+
"logps/rejected": -311.1128845214844,
|
| 623 |
+
"loss": 0.645816707611084,
|
| 624 |
+
"rewards/accuracies": 0.675000011920929,
|
| 625 |
+
"rewards/chosen": 0.4771113991737366,
|
| 626 |
+
"rewards/margins": 0.11158710718154907,
|
| 627 |
+
"rewards/rejected": 0.3655242323875427,
|
| 628 |
+
"step": 380
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"epoch": 2.7095652173913045,
|
| 632 |
+
"grad_norm": 51.0,
|
| 633 |
+
"learning_rate": 1.5000014915493465e-08,
|
| 634 |
+
"logits/chosen": -1.1323436498641968,
|
| 635 |
+
"logits/rejected": -1.1117795705795288,
|
| 636 |
+
"logps/chosen": -316.7195739746094,
|
| 637 |
+
"logps/rejected": -307.0235900878906,
|
| 638 |
+
"loss": 0.6755961894989013,
|
| 639 |
+
"rewards/accuracies": 0.581250011920929,
|
| 640 |
+
"rewards/chosen": 0.45080599188804626,
|
| 641 |
+
"rewards/margins": 0.048326827585697174,
|
| 642 |
+
"rewards/rejected": 0.4024791717529297,
|
| 643 |
+
"step": 390
|
| 644 |
+
},
|
| 645 |
+
{
|
| 646 |
+
"epoch": 2.7791304347826085,
|
| 647 |
+
"grad_norm": 77.0,
|
| 648 |
+
"learning_rate": 8.871331300335321e-09,
|
| 649 |
+
"logits/chosen": -1.0839048624038696,
|
| 650 |
+
"logits/rejected": -1.0926908254623413,
|
| 651 |
+
"logps/chosen": -324.00946044921875,
|
| 652 |
+
"logps/rejected": -318.5250549316406,
|
| 653 |
+
"loss": 0.6874616622924805,
|
| 654 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 655 |
+
"rewards/chosen": 0.45788002014160156,
|
| 656 |
+
"rewards/margins": 0.025683864951133728,
|
| 657 |
+
"rewards/rejected": 0.4321961998939514,
|
| 658 |
+
"step": 400
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"epoch": 2.7791304347826085,
|
| 662 |
+
"eval_logits/chosen": -1.059512972831726,
|
| 663 |
+
"eval_logits/rejected": -1.0796120166778564,
|
| 664 |
+
"eval_logps/chosen": -315.1502990722656,
|
| 665 |
+
"eval_logps/rejected": -316.1904296875,
|
| 666 |
+
"eval_loss": 0.6914414763450623,
|
| 667 |
+
"eval_rewards/accuracies": 0.5322580933570862,
|
| 668 |
+
"eval_rewards/chosen": 0.4315947890281677,
|
| 669 |
+
"eval_rewards/margins": 0.01160100195556879,
|
| 670 |
+
"eval_rewards/rejected": 0.4199938178062439,
|
| 671 |
+
"eval_runtime": 11.25,
|
| 672 |
+
"eval_samples_per_second": 10.756,
|
| 673 |
+
"eval_steps_per_second": 2.756,
|
| 674 |
+
"step": 400
|
| 675 |
+
}
|
| 676 |
+
],
|
| 677 |
+
"logging_steps": 10,
|
| 678 |
+
"max_steps": 432,
|
| 679 |
+
"num_input_tokens_seen": 0,
|
| 680 |
+
"num_train_epochs": 3,
|
| 681 |
+
"save_steps": 100,
|
| 682 |
+
"stateful_callbacks": {
|
| 683 |
+
"TrainerControl": {
|
| 684 |
+
"args": {
|
| 685 |
+
"should_epoch_stop": false,
|
| 686 |
+
"should_evaluate": false,
|
| 687 |
+
"should_log": false,
|
| 688 |
+
"should_save": true,
|
| 689 |
+
"should_training_stop": false
|
| 690 |
+
},
|
| 691 |
+
"attributes": {}
|
| 692 |
+
}
|
| 693 |
+
},
|
| 694 |
+
"total_flos": 0.0,
|
| 695 |
+
"train_batch_size": 4,
|
| 696 |
+
"trial_name": null,
|
| 697 |
+
"trial_params": null
|
| 698 |
+
}
|
checkpoint-400/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b05394a09844ad8fbc5bf2ac3f5b6be474605707ae7d97f68562a3f9312197c4
|
| 3 |
+
size 6225
|
checkpoint-432/chat_template.jinja
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{- bos_token -}}
|
| 2 |
+
{%- set keep_past_thinking = keep_past_thinking | default(false) -%}
|
| 3 |
+
{%- set ns = namespace(system_prompt="") -%}
|
| 4 |
+
{%- if messages[0]["role"] == "system" -%}
|
| 5 |
+
{%- set ns.system_prompt = messages[0]["content"] -%}
|
| 6 |
+
{%- set messages = messages[1:] -%}
|
| 7 |
+
{%- endif -%}
|
| 8 |
+
{%- if tools -%}
|
| 9 |
+
{%- set ns.system_prompt = ns.system_prompt + ("\n" if ns.system_prompt else "") + "List of tools: [" -%}
|
| 10 |
+
{%- for tool in tools -%}
|
| 11 |
+
{%- if tool is not string -%}
|
| 12 |
+
{%- set tool = tool | tojson -%}
|
| 13 |
+
{%- endif -%}
|
| 14 |
+
{%- set ns.system_prompt = ns.system_prompt + tool -%}
|
| 15 |
+
{%- if not loop.last -%}
|
| 16 |
+
{%- set ns.system_prompt = ns.system_prompt + ", " -%}
|
| 17 |
+
{%- endif -%}
|
| 18 |
+
{%- endfor -%}
|
| 19 |
+
{%- set ns.system_prompt = ns.system_prompt + "]" -%}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if ns.system_prompt -%}
|
| 22 |
+
{{- "<|im_start|>system\n" + ns.system_prompt + "<|im_end|>\n" -}}
|
| 23 |
+
{%- endif -%}
|
| 24 |
+
{%- set ns.last_assistant_index = -1 -%}
|
| 25 |
+
{%- for message in messages -%}
|
| 26 |
+
{%- if message["role"] == "assistant" -%}
|
| 27 |
+
{%- set ns.last_assistant_index = loop.index0 -%}
|
| 28 |
+
{%- endif -%}
|
| 29 |
+
{%- endfor -%}
|
| 30 |
+
{%- for message in messages -%}
|
| 31 |
+
{{- "<|im_start|>" + message["role"] + "\n" -}}
|
| 32 |
+
{%- set content = message["content"] -%}
|
| 33 |
+
{%- if content is not string -%}
|
| 34 |
+
{%- set content = content | tojson -%}
|
| 35 |
+
{%- endif -%}
|
| 36 |
+
{%- if message["role"] == "assistant" and not keep_past_thinking and loop.index0 != ns.last_assistant_index -%}
|
| 37 |
+
{%- if "</think>" in content -%}
|
| 38 |
+
{%- set content = content.split("</think>")[-1] | trim -%}
|
| 39 |
+
{%- endif -%}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{- content + "<|im_end|>\n" -}}
|
| 42 |
+
{%- endfor -%}
|
| 43 |
+
{%- if add_generation_prompt -%}
|
| 44 |
+
{{- "<|im_start|>assistant\n" -}}
|
| 45 |
+
{%- endif -%}
|
checkpoint-432/config.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Lfm2ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"block_auto_adjust_ff_dim": true,
|
| 6 |
+
"block_dim": 2048,
|
| 7 |
+
"block_ff_dim": 12288,
|
| 8 |
+
"block_ffn_dim_multiplier": 1.0,
|
| 9 |
+
"block_mlp_init_scale": 1.0,
|
| 10 |
+
"block_multiple_of": 256,
|
| 11 |
+
"block_norm_eps": 1e-05,
|
| 12 |
+
"block_out_init_scale": 1.0,
|
| 13 |
+
"block_use_swiglu": true,
|
| 14 |
+
"block_use_xavier_init": true,
|
| 15 |
+
"bos_token_id": 1,
|
| 16 |
+
"conv_L_cache": 3,
|
| 17 |
+
"conv_bias": false,
|
| 18 |
+
"conv_dim": 2048,
|
| 19 |
+
"conv_use_xavier_init": true,
|
| 20 |
+
"dtype": "bfloat16",
|
| 21 |
+
"eos_token_id": 7,
|
| 22 |
+
"hidden_size": 2048,
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 12288,
|
| 25 |
+
"layer_types": [
|
| 26 |
+
"conv",
|
| 27 |
+
"conv",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"conv",
|
| 30 |
+
"conv",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"conv",
|
| 33 |
+
"conv",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"conv",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"conv",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"conv",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"conv"
|
| 42 |
+
],
|
| 43 |
+
"max_position_embeddings": 128000,
|
| 44 |
+
"model_type": "lfm2",
|
| 45 |
+
"norm_eps": 1e-05,
|
| 46 |
+
"num_attention_heads": 32,
|
| 47 |
+
"num_heads": 32,
|
| 48 |
+
"num_hidden_layers": 16,
|
| 49 |
+
"num_key_value_heads": 8,
|
| 50 |
+
"pad_token_id": 0,
|
| 51 |
+
"rope_parameters": {
|
| 52 |
+
"rope_theta": 1000000.0,
|
| 53 |
+
"rope_type": "default"
|
| 54 |
+
},
|
| 55 |
+
"tie_embedding": true,
|
| 56 |
+
"tie_word_embeddings": true,
|
| 57 |
+
"transformers_version": "5.2.0",
|
| 58 |
+
"use_cache": false,
|
| 59 |
+
"use_pos_enc": true,
|
| 60 |
+
"vocab_size": 65536
|
| 61 |
+
}
|
checkpoint-432/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
7
|
| 6 |
+
],
|
| 7 |
+
"pad_token_id": 0,
|
| 8 |
+
"transformers_version": "5.2.0"
|
| 9 |
+
}
|
checkpoint-432/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3d360e6c4593561d9ab3c7c38f9c56f3b0e9007f2104a3d408741552c249a101
|
| 3 |
+
size 2340697936
|
checkpoint-432/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:293e1dd43e9549c2668c55fc867d914e17d6d112d4671b18f856bb9bbd8177d7
|
| 3 |
+
size 3178927435
|
checkpoint-432/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f196323d7423b60f8e4ceb7dbf8715ee326c0d068e5ff164f13c63b279b9f1a0
|
| 3 |
+
size 14645
|
checkpoint-432/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:500bc375cdd19b81329befa9b7152f30c7d8b2e4876303d1a5d1646dc97c1f9f
|
| 3 |
+
size 1465
|
checkpoint-432/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-432/tokenizer_config.json
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"bos_token": "<|startoftext|>",
|
| 4 |
+
"clean_up_tokenization_spaces": false,
|
| 5 |
+
"eos_token": "<|im_end|>",
|
| 6 |
+
"is_local": false,
|
| 7 |
+
"legacy": false,
|
| 8 |
+
"model_input_names": [
|
| 9 |
+
"input_ids",
|
| 10 |
+
"attention_mask"
|
| 11 |
+
],
|
| 12 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 13 |
+
"pad_token": "<|pad|>",
|
| 14 |
+
"padding_side": "right",
|
| 15 |
+
"sp_model_kwargs": {},
|
| 16 |
+
"spaces_between_special_tokens": false,
|
| 17 |
+
"tokenizer_class": "TokenizersBackend",
|
| 18 |
+
"use_default_system_prompt": false,
|
| 19 |
+
"use_fast": true
|
| 20 |
+
}
|
checkpoint-432/trainer_state.json
ADDED
|
@@ -0,0 +1,743 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 300,
|
| 3 |
+
"best_metric": 0.6872237920761108,
|
| 4 |
+
"best_model_checkpoint": "models/dpo_fft_LFM2.5-1.2B-Instruct_argilla__distilabel-math-preference-dpo_20260222_195126/checkpoint-300",
|
| 5 |
+
"epoch": 3.0,
|
| 6 |
+
"eval_steps": 100,
|
| 7 |
+
"global_step": 432,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.06956521739130435,
|
| 14 |
+
"grad_norm": 44.25,
|
| 15 |
+
"learning_rate": 1.0227272727272728e-07,
|
| 16 |
+
"logits/chosen": -1.0599091053009033,
|
| 17 |
+
"logits/rejected": -1.042307734489441,
|
| 18 |
+
"logps/chosen": -332.13507080078125,
|
| 19 |
+
"logps/rejected": -332.9361267089844,
|
| 20 |
+
"loss": 0.6927361965179444,
|
| 21 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 22 |
+
"rewards/chosen": 0.01432617288082838,
|
| 23 |
+
"rewards/margins": 0.0042648655362427235,
|
| 24 |
+
"rewards/rejected": 0.010061310604214668,
|
| 25 |
+
"step": 10
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.1391304347826087,
|
| 29 |
+
"grad_norm": 50.5,
|
| 30 |
+
"learning_rate": 2.159090909090909e-07,
|
| 31 |
+
"logits/chosen": -1.0978257656097412,
|
| 32 |
+
"logits/rejected": -1.09658682346344,
|
| 33 |
+
"logps/chosen": -329.22100830078125,
|
| 34 |
+
"logps/rejected": -312.2664794921875,
|
| 35 |
+
"loss": 0.7003747940063476,
|
| 36 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 37 |
+
"rewards/chosen": 0.004466252867132425,
|
| 38 |
+
"rewards/margins": -0.011238865554332733,
|
| 39 |
+
"rewards/rejected": 0.01570511981844902,
|
| 40 |
+
"step": 20
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"epoch": 0.20869565217391303,
|
| 44 |
+
"grad_norm": 54.75,
|
| 45 |
+
"learning_rate": 3.295454545454545e-07,
|
| 46 |
+
"logits/chosen": -1.1366004943847656,
|
| 47 |
+
"logits/rejected": -1.1707069873809814,
|
| 48 |
+
"logps/chosen": -328.1323547363281,
|
| 49 |
+
"logps/rejected": -305.90618896484375,
|
| 50 |
+
"loss": 0.7031519412994385,
|
| 51 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 52 |
+
"rewards/chosen": 0.03125340864062309,
|
| 53 |
+
"rewards/margins": -0.01610126718878746,
|
| 54 |
+
"rewards/rejected": 0.04735467582941055,
|
| 55 |
+
"step": 30
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 0.2782608695652174,
|
| 59 |
+
"grad_norm": 50.5,
|
| 60 |
+
"learning_rate": 4.4318181818181817e-07,
|
| 61 |
+
"logits/chosen": -1.0889461040496826,
|
| 62 |
+
"logits/rejected": -1.1017862558364868,
|
| 63 |
+
"logps/chosen": -338.80938720703125,
|
| 64 |
+
"logps/rejected": -328.07464599609375,
|
| 65 |
+
"loss": 0.6893259048461914,
|
| 66 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 67 |
+
"rewards/chosen": 0.09773217141628265,
|
| 68 |
+
"rewards/margins": 0.01154844556003809,
|
| 69 |
+
"rewards/rejected": 0.08618371188640594,
|
| 70 |
+
"step": 40
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"epoch": 0.34782608695652173,
|
| 74 |
+
"grad_norm": 53.5,
|
| 75 |
+
"learning_rate": 4.997951542310825e-07,
|
| 76 |
+
"logits/chosen": -1.1367381811141968,
|
| 77 |
+
"logits/rejected": -1.1468210220336914,
|
| 78 |
+
"logps/chosen": -329.12823486328125,
|
| 79 |
+
"logps/rejected": -323.8036804199219,
|
| 80 |
+
"loss": 0.6982084751129151,
|
| 81 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 82 |
+
"rewards/chosen": 0.1568203717470169,
|
| 83 |
+
"rewards/margins": -0.004218918737024069,
|
| 84 |
+
"rewards/rejected": 0.1610392928123474,
|
| 85 |
+
"step": 50
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"epoch": 0.41739130434782606,
|
| 89 |
+
"grad_norm": 64.0,
|
| 90 |
+
"learning_rate": 4.98158401695492e-07,
|
| 91 |
+
"logits/chosen": -1.0899083614349365,
|
| 92 |
+
"logits/rejected": -1.0698351860046387,
|
| 93 |
+
"logps/chosen": -335.6194152832031,
|
| 94 |
+
"logps/rejected": -329.00653076171875,
|
| 95 |
+
"loss": 0.6905360221862793,
|
| 96 |
+
"rewards/accuracies": 0.53125,
|
| 97 |
+
"rewards/chosen": 0.1543850153684616,
|
| 98 |
+
"rewards/margins": 0.010315253399312496,
|
| 99 |
+
"rewards/rejected": 0.14406974613666534,
|
| 100 |
+
"step": 60
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.48695652173913045,
|
| 104 |
+
"grad_norm": 72.5,
|
| 105 |
+
"learning_rate": 4.948956212473369e-07,
|
| 106 |
+
"logits/chosen": -1.1223671436309814,
|
| 107 |
+
"logits/rejected": -1.113906741142273,
|
| 108 |
+
"logps/chosen": -319.424560546875,
|
| 109 |
+
"logps/rejected": -317.76934814453125,
|
| 110 |
+
"loss": 0.7009393215179444,
|
| 111 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 112 |
+
"rewards/chosen": 0.18402941524982452,
|
| 113 |
+
"rewards/margins": -0.010006649419665337,
|
| 114 |
+
"rewards/rejected": 0.19403605163097382,
|
| 115 |
+
"step": 70
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.5565217391304348,
|
| 119 |
+
"grad_norm": 52.5,
|
| 120 |
+
"learning_rate": 4.900281918608732e-07,
|
| 121 |
+
"logits/chosen": -1.1157792806625366,
|
| 122 |
+
"logits/rejected": -1.1078118085861206,
|
| 123 |
+
"logps/chosen": -314.5404052734375,
|
| 124 |
+
"logps/rejected": -311.91656494140625,
|
| 125 |
+
"loss": 0.6877760887145996,
|
| 126 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 127 |
+
"rewards/chosen": 0.19012701511383057,
|
| 128 |
+
"rewards/margins": 0.016789216548204422,
|
| 129 |
+
"rewards/rejected": 0.17333778738975525,
|
| 130 |
+
"step": 80
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.6260869565217392,
|
| 134 |
+
"grad_norm": 47.75,
|
| 135 |
+
"learning_rate": 4.835880067784441e-07,
|
| 136 |
+
"logits/chosen": -1.0738165378570557,
|
| 137 |
+
"logits/rejected": -1.0576341152191162,
|
| 138 |
+
"logps/chosen": -336.3821716308594,
|
| 139 |
+
"logps/rejected": -331.9998779296875,
|
| 140 |
+
"loss": 0.6956596374511719,
|
| 141 |
+
"rewards/accuracies": 0.4375,
|
| 142 |
+
"rewards/chosen": 0.21225734055042267,
|
| 143 |
+
"rewards/margins": 0.0009234324097633362,
|
| 144 |
+
"rewards/rejected": 0.21133390069007874,
|
| 145 |
+
"step": 90
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.6956521739130435,
|
| 149 |
+
"grad_norm": 49.0,
|
| 150 |
+
"learning_rate": 4.7561726453386744e-07,
|
| 151 |
+
"logits/chosen": -1.0457407236099243,
|
| 152 |
+
"logits/rejected": -1.038447618484497,
|
| 153 |
+
"logps/chosen": -330.8192138671875,
|
| 154 |
+
"logps/rejected": -323.5484924316406,
|
| 155 |
+
"loss": 0.6817105770111084,
|
| 156 |
+
"rewards/accuracies": 0.581250011920929,
|
| 157 |
+
"rewards/chosen": 0.26413798332214355,
|
| 158 |
+
"rewards/margins": 0.029450010508298874,
|
| 159 |
+
"rewards/rejected": 0.2346879541873932,
|
| 160 |
+
"step": 100
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.6956521739130435,
|
| 164 |
+
"eval_logits/chosen": -1.0708128213882446,
|
| 165 |
+
"eval_logits/rejected": -1.0917303562164307,
|
| 166 |
+
"eval_logps/chosen": -317.1337585449219,
|
| 167 |
+
"eval_logps/rejected": -318.0965270996094,
|
| 168 |
+
"eval_loss": 0.6924862265586853,
|
| 169 |
+
"eval_rewards/accuracies": 0.4838709533214569,
|
| 170 |
+
"eval_rewards/chosen": 0.2332489788532257,
|
| 171 |
+
"eval_rewards/margins": 0.0038641756400465965,
|
| 172 |
+
"eval_rewards/rejected": 0.22938479483127594,
|
| 173 |
+
"eval_runtime": 11.3087,
|
| 174 |
+
"eval_samples_per_second": 10.7,
|
| 175 |
+
"eval_steps_per_second": 2.741,
|
| 176 |
+
"step": 100
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 0.7652173913043478,
|
| 180 |
+
"grad_norm": 47.0,
|
| 181 |
+
"learning_rate": 4.6616819245164655e-07,
|
| 182 |
+
"logits/chosen": -1.0736799240112305,
|
| 183 |
+
"logits/rejected": -1.0783708095550537,
|
| 184 |
+
"logps/chosen": -314.97943115234375,
|
| 185 |
+
"logps/rejected": -308.35675048828125,
|
| 186 |
+
"loss": 0.6984636306762695,
|
| 187 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 188 |
+
"rewards/chosen": 0.22377920150756836,
|
| 189 |
+
"rewards/margins": -0.0050087496638298035,
|
| 190 |
+
"rewards/rejected": 0.22878792881965637,
|
| 191 |
+
"step": 110
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.8347826086956521,
|
| 195 |
+
"grad_norm": 51.25,
|
| 196 |
+
"learning_rate": 4.55302704433743e-07,
|
| 197 |
+
"logits/chosen": -1.048730731010437,
|
| 198 |
+
"logits/rejected": -1.0437910556793213,
|
| 199 |
+
"logps/chosen": -332.6585388183594,
|
| 200 |
+
"logps/rejected": -319.9508361816406,
|
| 201 |
+
"loss": 0.6809643268585205,
|
| 202 |
+
"rewards/accuracies": 0.59375,
|
| 203 |
+
"rewards/chosen": 0.3183898627758026,
|
| 204 |
+
"rewards/margins": 0.03039008378982544,
|
| 205 |
+
"rewards/rejected": 0.2879997789859772,
|
| 206 |
+
"step": 120
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.9043478260869565,
|
| 210 |
+
"grad_norm": 57.5,
|
| 211 |
+
"learning_rate": 4.4309199527622254e-07,
|
| 212 |
+
"logits/chosen": -1.0928480625152588,
|
| 213 |
+
"logits/rejected": -1.0909947156906128,
|
| 214 |
+
"logps/chosen": -333.54510498046875,
|
| 215 |
+
"logps/rejected": -321.5794372558594,
|
| 216 |
+
"loss": 0.6884570121765137,
|
| 217 |
+
"rewards/accuracies": 0.518750011920929,
|
| 218 |
+
"rewards/chosen": 0.3165169656276703,
|
| 219 |
+
"rewards/margins": 0.015730690211057663,
|
| 220 |
+
"rewards/rejected": 0.3007862865924835,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.9739130434782609,
|
| 225 |
+
"grad_norm": 50.75,
|
| 226 |
+
"learning_rate": 4.2961607417396517e-07,
|
| 227 |
+
"logits/chosen": -1.0702852010726929,
|
| 228 |
+
"logits/rejected": -1.0860363245010376,
|
| 229 |
+
"logps/chosen": -336.94122314453125,
|
| 230 |
+
"logps/rejected": -312.6028137207031,
|
| 231 |
+
"loss": 0.6929869651794434,
|
| 232 |
+
"rewards/accuracies": 0.5,
|
| 233 |
+
"rewards/chosen": 0.3366253077983856,
|
| 234 |
+
"rewards/margins": 0.008591671474277973,
|
| 235 |
+
"rewards/rejected": 0.3280336260795593,
|
| 236 |
+
"step": 140
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 1.0417391304347827,
|
| 240 |
+
"grad_norm": 58.25,
|
| 241 |
+
"learning_rate": 4.1496324047009244e-07,
|
| 242 |
+
"logits/chosen": -1.137653112411499,
|
| 243 |
+
"logits/rejected": -1.1385780572891235,
|
| 244 |
+
"logps/chosen": -297.7463073730469,
|
| 245 |
+
"logps/rejected": -304.8460388183594,
|
| 246 |
+
"loss": 0.6796726703643798,
|
| 247 |
+
"rewards/accuracies": 0.5320512652397156,
|
| 248 |
+
"rewards/chosen": 0.36476510763168335,
|
| 249 |
+
"rewards/margins": 0.029467487707734108,
|
| 250 |
+
"rewards/rejected": 0.3352976143360138,
|
| 251 |
+
"step": 150
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.111304347826087,
|
| 255 |
+
"grad_norm": 62.0,
|
| 256 |
+
"learning_rate": 3.9922950508520126e-07,
|
| 257 |
+
"logits/chosen": -1.0764613151550293,
|
| 258 |
+
"logits/rejected": -1.073861837387085,
|
| 259 |
+
"logps/chosen": -325.74151611328125,
|
| 260 |
+
"logps/rejected": -314.5641174316406,
|
| 261 |
+
"loss": 0.6664072513580322,
|
| 262 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 263 |
+
"rewards/chosen": 0.37418311834335327,
|
| 264 |
+
"rewards/margins": 0.06225377321243286,
|
| 265 |
+
"rewards/rejected": 0.3119293451309204,
|
| 266 |
+
"step": 160
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.1808695652173913,
|
| 270 |
+
"grad_norm": 48.25,
|
| 271 |
+
"learning_rate": 3.8251796141741945e-07,
|
| 272 |
+
"logits/chosen": -1.1032686233520508,
|
| 273 |
+
"logits/rejected": -1.1010105609893799,
|
| 274 |
+
"logps/chosen": -324.8043212890625,
|
| 275 |
+
"logps/rejected": -314.7898254394531,
|
| 276 |
+
"loss": 0.6656156063079834,
|
| 277 |
+
"rewards/accuracies": 0.643750011920929,
|
| 278 |
+
"rewards/chosen": 0.39548203349113464,
|
| 279 |
+
"rewards/margins": 0.06372012943029404,
|
| 280 |
+
"rewards/rejected": 0.33176189661026,
|
| 281 |
+
"step": 170
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 1.2504347826086957,
|
| 285 |
+
"grad_norm": 73.5,
|
| 286 |
+
"learning_rate": 3.649381098353834e-07,
|
| 287 |
+
"logits/chosen": -1.108781099319458,
|
| 288 |
+
"logits/rejected": -1.081012487411499,
|
| 289 |
+
"logps/chosen": -333.0024719238281,
|
| 290 |
+
"logps/rejected": -317.6542663574219,
|
| 291 |
+
"loss": 0.6682669162750244,
|
| 292 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 293 |
+
"rewards/chosen": 0.4114314019680023,
|
| 294 |
+
"rewards/margins": 0.06464624404907227,
|
| 295 |
+
"rewards/rejected": 0.34678512811660767,
|
| 296 |
+
"step": 180
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 1.32,
|
| 300 |
+
"grad_norm": 51.75,
|
| 301 |
+
"learning_rate": 3.466051401903162e-07,
|
| 302 |
+
"logits/chosen": -1.0279661417007446,
|
| 303 |
+
"logits/rejected": -1.028050422668457,
|
| 304 |
+
"logps/chosen": -342.05322265625,
|
| 305 |
+
"logps/rejected": -329.083984375,
|
| 306 |
+
"loss": 0.6815151691436767,
|
| 307 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 308 |
+
"rewards/chosen": 0.4475061893463135,
|
| 309 |
+
"rewards/margins": 0.03387105464935303,
|
| 310 |
+
"rewards/rejected": 0.41363516449928284,
|
| 311 |
+
"step": 190
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"epoch": 1.3895652173913042,
|
| 315 |
+
"grad_norm": 43.25,
|
| 316 |
+
"learning_rate": 3.276391770484606e-07,
|
| 317 |
+
"logits/chosen": -1.0554686784744263,
|
| 318 |
+
"logits/rejected": -1.0444362163543701,
|
| 319 |
+
"logps/chosen": -323.93359375,
|
| 320 |
+
"logps/rejected": -300.8354187011719,
|
| 321 |
+
"loss": 0.6729463577270508,
|
| 322 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 323 |
+
"rewards/chosen": 0.4051585793495178,
|
| 324 |
+
"rewards/margins": 0.04842006787657738,
|
| 325 |
+
"rewards/rejected": 0.35673850774765015,
|
| 326 |
+
"step": 200
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 1.3895652173913042,
|
| 330 |
+
"eval_logits/chosen": -1.059758186340332,
|
| 331 |
+
"eval_logits/rejected": -1.0798335075378418,
|
| 332 |
+
"eval_logps/chosen": -315.1194763183594,
|
| 333 |
+
"eval_logps/rejected": -316.16046142578125,
|
| 334 |
+
"eval_loss": 0.6918731331825256,
|
| 335 |
+
"eval_rewards/accuracies": 0.5483871102333069,
|
| 336 |
+
"eval_rewards/chosen": 0.4346778690814972,
|
| 337 |
+
"eval_rewards/margins": 0.011690356768667698,
|
| 338 |
+
"eval_rewards/rejected": 0.42298752069473267,
|
| 339 |
+
"eval_runtime": 11.2737,
|
| 340 |
+
"eval_samples_per_second": 10.733,
|
| 341 |
+
"eval_steps_per_second": 2.75,
|
| 342 |
+
"step": 200
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 1.4591304347826086,
|
| 346 |
+
"grad_norm": 44.5,
|
| 347 |
+
"learning_rate": 3.0816449258938656e-07,
|
| 348 |
+
"logits/chosen": -1.0483617782592773,
|
| 349 |
+
"logits/rejected": -1.0491857528686523,
|
| 350 |
+
"logps/chosen": -326.4630126953125,
|
| 351 |
+
"logps/rejected": -321.35467529296875,
|
| 352 |
+
"loss": 0.6760656356811523,
|
| 353 |
+
"rewards/accuracies": 0.59375,
|
| 354 |
+
"rewards/chosen": 0.4774439334869385,
|
| 355 |
+
"rewards/margins": 0.04994555562734604,
|
| 356 |
+
"rewards/rejected": 0.42749834060668945,
|
| 357 |
+
"step": 210
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 1.528695652173913,
|
| 361 |
+
"grad_norm": 69.0,
|
| 362 |
+
"learning_rate": 2.883086923275658e-07,
|
| 363 |
+
"logits/chosen": -1.0638173818588257,
|
| 364 |
+
"logits/rejected": -1.0861108303070068,
|
| 365 |
+
"logps/chosen": -323.24517822265625,
|
| 366 |
+
"logps/rejected": -316.22454833984375,
|
| 367 |
+
"loss": 0.6859281063079834,
|
| 368 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 369 |
+
"rewards/chosen": 0.464817613363266,
|
| 370 |
+
"rewards/margins": 0.025601129978895187,
|
| 371 |
+
"rewards/rejected": 0.4392164647579193,
|
| 372 |
+
"step": 220
|
| 373 |
+
},
|
| 374 |
+
{
|
| 375 |
+
"epoch": 1.5982608695652174,
|
| 376 |
+
"grad_norm": 51.25,
|
| 377 |
+
"learning_rate": 2.6820187899267197e-07,
|
| 378 |
+
"logits/chosen": -1.0582925081253052,
|
| 379 |
+
"logits/rejected": -1.0700470209121704,
|
| 380 |
+
"logps/chosen": -331.1497802734375,
|
| 381 |
+
"logps/rejected": -318.2004089355469,
|
| 382 |
+
"loss": 0.6922653675079345,
|
| 383 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 384 |
+
"rewards/chosen": 0.4878065586090088,
|
| 385 |
+
"rewards/margins": 0.014133572578430176,
|
| 386 |
+
"rewards/rejected": 0.4736729562282562,
|
| 387 |
+
"step": 230
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 1.6678260869565218,
|
| 391 |
+
"grad_norm": 44.0,
|
| 392 |
+
"learning_rate": 2.4797580004718036e-07,
|
| 393 |
+
"logits/chosen": -1.059127926826477,
|
| 394 |
+
"logits/rejected": -1.0511342287063599,
|
| 395 |
+
"logps/chosen": -329.36309814453125,
|
| 396 |
+
"logps/rejected": -324.93060302734375,
|
| 397 |
+
"loss": 0.6760355949401855,
|
| 398 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 399 |
+
"rewards/chosen": 0.4827202260494232,
|
| 400 |
+
"rewards/margins": 0.04902523010969162,
|
| 401 |
+
"rewards/rejected": 0.433694988489151,
|
| 402 |
+
"step": 240
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"epoch": 1.7373913043478262,
|
| 406 |
+
"grad_norm": 47.0,
|
| 407 |
+
"learning_rate": 2.277629844270543e-07,
|
| 408 |
+
"logits/chosen": -1.1127703189849854,
|
| 409 |
+
"logits/rejected": -1.119107961654663,
|
| 410 |
+
"logps/chosen": -313.205810546875,
|
| 411 |
+
"logps/rejected": -302.2017822265625,
|
| 412 |
+
"loss": 0.6791357040405274,
|
| 413 |
+
"rewards/accuracies": 0.59375,
|
| 414 |
+
"rewards/chosen": 0.46044379472732544,
|
| 415 |
+
"rewards/margins": 0.04207334667444229,
|
| 416 |
+
"rewards/rejected": 0.41837042570114136,
|
| 417 |
+
"step": 250
|
| 418 |
+
},
|
| 419 |
+
{
|
| 420 |
+
"epoch": 1.8069565217391306,
|
| 421 |
+
"grad_norm": 74.5,
|
| 422 |
+
"learning_rate": 2.0769587416192208e-07,
|
| 423 |
+
"logits/chosen": -1.0957348346710205,
|
| 424 |
+
"logits/rejected": -1.0845588445663452,
|
| 425 |
+
"logps/chosen": -326.6529235839844,
|
| 426 |
+
"logps/rejected": -328.163330078125,
|
| 427 |
+
"loss": 0.6897631645202636,
|
| 428 |
+
"rewards/accuracies": 0.550000011920929,
|
| 429 |
+
"rewards/chosen": 0.4732390344142914,
|
| 430 |
+
"rewards/margins": 0.017790159210562706,
|
| 431 |
+
"rewards/rejected": 0.4554489552974701,
|
| 432 |
+
"step": 260
|
| 433 |
+
},
|
| 434 |
+
{
|
| 435 |
+
"epoch": 1.8765217391304347,
|
| 436 |
+
"grad_norm": 54.25,
|
| 437 |
+
"learning_rate": 1.8790595656469626e-07,
|
| 438 |
+
"logits/chosen": -1.0816611051559448,
|
| 439 |
+
"logits/rejected": -1.0837030410766602,
|
| 440 |
+
"logps/chosen": -331.3684387207031,
|
| 441 |
+
"logps/rejected": -313.75445556640625,
|
| 442 |
+
"loss": 0.6794800758361816,
|
| 443 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 444 |
+
"rewards/chosen": 0.4701307415962219,
|
| 445 |
+
"rewards/margins": 0.03732570633292198,
|
| 446 |
+
"rewards/rejected": 0.43280500173568726,
|
| 447 |
+
"step": 270
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 1.9460869565217391,
|
| 451 |
+
"grad_norm": 50.5,
|
| 452 |
+
"learning_rate": 1.6852290267685928e-07,
|
| 453 |
+
"logits/chosen": -1.0794267654418945,
|
| 454 |
+
"logits/rejected": -1.0937658548355103,
|
| 455 |
+
"logps/chosen": -331.467529296875,
|
| 456 |
+
"logps/rejected": -321.3943786621094,
|
| 457 |
+
"loss": 0.6855987071990967,
|
| 458 |
+
"rewards/accuracies": 0.543749988079071,
|
| 459 |
+
"rewards/chosen": 0.48015865683555603,
|
| 460 |
+
"rewards/margins": 0.03580061346292496,
|
| 461 |
+
"rewards/rejected": 0.44435805082321167,
|
| 462 |
+
"step": 280
|
| 463 |
+
},
|
| 464 |
+
{
|
| 465 |
+
"epoch": 2.013913043478261,
|
| 466 |
+
"grad_norm": 45.25,
|
| 467 |
+
"learning_rate": 1.4967371761464737e-07,
|
| 468 |
+
"logits/chosen": -1.0743939876556396,
|
| 469 |
+
"logits/rejected": -1.0771441459655762,
|
| 470 |
+
"logps/chosen": -322.74835205078125,
|
| 471 |
+
"logps/rejected": -319.2298278808594,
|
| 472 |
+
"loss": 0.690724229812622,
|
| 473 |
+
"rewards/accuracies": 0.4871794879436493,
|
| 474 |
+
"rewards/chosen": 0.4530966281890869,
|
| 475 |
+
"rewards/margins": 0.012794151902198792,
|
| 476 |
+
"rewards/rejected": 0.4403024911880493,
|
| 477 |
+
"step": 290
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 2.0834782608695654,
|
| 481 |
+
"grad_norm": 45.5,
|
| 482 |
+
"learning_rate": 1.3148190838338802e-07,
|
| 483 |
+
"logits/chosen": -1.0955299139022827,
|
| 484 |
+
"logits/rejected": -1.1200354099273682,
|
| 485 |
+
"logps/chosen": -331.6620178222656,
|
| 486 |
+
"logps/rejected": -313.5164489746094,
|
| 487 |
+
"loss": 0.6698547840118408,
|
| 488 |
+
"rewards/accuracies": 0.612500011920929,
|
| 489 |
+
"rewards/chosen": 0.44486236572265625,
|
| 490 |
+
"rewards/margins": 0.06079784035682678,
|
| 491 |
+
"rewards/rejected": 0.38406452536582947,
|
| 492 |
+
"step": 300
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"epoch": 2.0834782608695654,
|
| 496 |
+
"eval_logits/chosen": -1.0598626136779785,
|
| 497 |
+
"eval_logits/rejected": -1.080617070198059,
|
| 498 |
+
"eval_logps/chosen": -315.2421875,
|
| 499 |
+
"eval_logps/rejected": -316.3223876953125,
|
| 500 |
+
"eval_loss": 0.6872237920761108,
|
| 501 |
+
"eval_rewards/accuracies": 0.5403226017951965,
|
| 502 |
+
"eval_rewards/chosen": 0.42240414023399353,
|
| 503 |
+
"eval_rewards/margins": 0.015609556809067726,
|
| 504 |
+
"eval_rewards/rejected": 0.40679463744163513,
|
| 505 |
+
"eval_runtime": 11.2881,
|
| 506 |
+
"eval_samples_per_second": 10.719,
|
| 507 |
+
"eval_steps_per_second": 2.746,
|
| 508 |
+
"step": 300
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"epoch": 2.1530434782608694,
|
| 512 |
+
"grad_norm": 47.5,
|
| 513 |
+
"learning_rate": 1.1406667461278538e-07,
|
| 514 |
+
"logits/chosen": -1.0953623056411743,
|
| 515 |
+
"logits/rejected": -1.0638518333435059,
|
| 516 |
+
"logps/chosen": -317.5690002441406,
|
| 517 |
+
"logps/rejected": -313.86895751953125,
|
| 518 |
+
"loss": 0.6786671161651612,
|
| 519 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 520 |
+
"rewards/chosen": 0.43794316053390503,
|
| 521 |
+
"rewards/margins": 0.040073662996292114,
|
| 522 |
+
"rewards/rejected": 0.39786943793296814,
|
| 523 |
+
"step": 310
|
| 524 |
+
},
|
| 525 |
+
{
|
| 526 |
+
"epoch": 2.222608695652174,
|
| 527 |
+
"grad_norm": 75.5,
|
| 528 |
+
"learning_rate": 9.754212751576385e-08,
|
| 529 |
+
"logits/chosen": -1.036792516708374,
|
| 530 |
+
"logits/rejected": -1.0298279523849487,
|
| 531 |
+
"logps/chosen": -322.68511962890625,
|
| 532 |
+
"logps/rejected": -324.95562744140625,
|
| 533 |
+
"loss": 0.6964206695556641,
|
| 534 |
+
"rewards/accuracies": 0.5,
|
| 535 |
+
"rewards/chosen": 0.4462190568447113,
|
| 536 |
+
"rewards/margins": 0.005142434034496546,
|
| 537 |
+
"rewards/rejected": 0.4410766065120697,
|
| 538 |
+
"step": 320
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"epoch": 2.292173913043478,
|
| 542 |
+
"grad_norm": 50.0,
|
| 543 |
+
"learning_rate": 8.201654218854689e-08,
|
| 544 |
+
"logits/chosen": -1.076971173286438,
|
| 545 |
+
"logits/rejected": -1.0646440982818604,
|
| 546 |
+
"logps/chosen": -314.86492919921875,
|
| 547 |
+
"logps/rejected": -309.55389404296875,
|
| 548 |
+
"loss": 0.6790688037872314,
|
| 549 |
+
"rewards/accuracies": 0.581250011920929,
|
| 550 |
+
"rewards/chosen": 0.4310247302055359,
|
| 551 |
+
"rewards/margins": 0.03967974707484245,
|
| 552 |
+
"rewards/rejected": 0.39134499430656433,
|
| 553 |
+
"step": 330
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 2.3617391304347826,
|
| 557 |
+
"grad_norm": 48.0,
|
| 558 |
+
"learning_rate": 6.759164815118493e-08,
|
| 559 |
+
"logits/chosen": -1.0706030130386353,
|
| 560 |
+
"logits/rejected": -1.0911849737167358,
|
| 561 |
+
"logps/chosen": -337.18853759765625,
|
| 562 |
+
"logps/rejected": -322.5784912109375,
|
| 563 |
+
"loss": 0.6748550415039063,
|
| 564 |
+
"rewards/accuracies": 0.625,
|
| 565 |
+
"rewards/chosen": 0.5078705549240112,
|
| 566 |
+
"rewards/margins": 0.05118296295404434,
|
| 567 |
+
"rewards/rejected": 0.4566876292228699,
|
| 568 |
+
"step": 340
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"epoch": 2.431304347826087,
|
| 572 |
+
"grad_norm": 60.5,
|
| 573 |
+
"learning_rate": 5.4361962777179273e-08,
|
| 574 |
+
"logits/chosen": -1.0652002096176147,
|
| 575 |
+
"logits/rejected": -1.0517784357070923,
|
| 576 |
+
"logps/chosen": -330.02325439453125,
|
| 577 |
+
"logps/rejected": -326.01873779296875,
|
| 578 |
+
"loss": 0.691571044921875,
|
| 579 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 580 |
+
"rewards/chosen": 0.46054545044898987,
|
| 581 |
+
"rewards/margins": 0.012338491156697273,
|
| 582 |
+
"rewards/rejected": 0.44820699095726013,
|
| 583 |
+
"step": 350
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"epoch": 2.5008695652173913,
|
| 587 |
+
"grad_norm": 51.75,
|
| 588 |
+
"learning_rate": 4.2414171979824e-08,
|
| 589 |
+
"logits/chosen": -1.0523455142974854,
|
| 590 |
+
"logits/rejected": -1.0446560382843018,
|
| 591 |
+
"logps/chosen": -336.7765197753906,
|
| 592 |
+
"logps/rejected": -316.00067138671875,
|
| 593 |
+
"loss": 0.6638113975524902,
|
| 594 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 595 |
+
"rewards/chosen": 0.46467241644859314,
|
| 596 |
+
"rewards/margins": 0.06902624666690826,
|
| 597 |
+
"rewards/rejected": 0.39564621448516846,
|
| 598 |
+
"step": 360
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"epoch": 2.5704347826086957,
|
| 602 |
+
"grad_norm": 49.0,
|
| 603 |
+
"learning_rate": 3.1826562213243844e-08,
|
| 604 |
+
"logits/chosen": -1.0767319202423096,
|
| 605 |
+
"logits/rejected": -1.0776704549789429,
|
| 606 |
+
"logps/chosen": -329.4648742675781,
|
| 607 |
+
"logps/rejected": -319.5013732910156,
|
| 608 |
+
"loss": 0.6690452575683594,
|
| 609 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 610 |
+
"rewards/chosen": 0.46328702569007874,
|
| 611 |
+
"rewards/margins": 0.061441998928785324,
|
| 612 |
+
"rewards/rejected": 0.40184497833251953,
|
| 613 |
+
"step": 370
|
| 614 |
+
},
|
| 615 |
+
{
|
| 616 |
+
"epoch": 2.64,
|
| 617 |
+
"grad_norm": 56.5,
|
| 618 |
+
"learning_rate": 2.2668507509871953e-08,
|
| 619 |
+
"logits/chosen": -1.1089640855789185,
|
| 620 |
+
"logits/rejected": -1.0823930501937866,
|
| 621 |
+
"logps/chosen": -328.79351806640625,
|
| 622 |
+
"logps/rejected": -311.1128845214844,
|
| 623 |
+
"loss": 0.645816707611084,
|
| 624 |
+
"rewards/accuracies": 0.675000011920929,
|
| 625 |
+
"rewards/chosen": 0.4771113991737366,
|
| 626 |
+
"rewards/margins": 0.11158710718154907,
|
| 627 |
+
"rewards/rejected": 0.3655242323875427,
|
| 628 |
+
"step": 380
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"epoch": 2.7095652173913045,
|
| 632 |
+
"grad_norm": 51.0,
|
| 633 |
+
"learning_rate": 1.5000014915493465e-08,
|
| 634 |
+
"logits/chosen": -1.1323436498641968,
|
| 635 |
+
"logits/rejected": -1.1117795705795288,
|
| 636 |
+
"logps/chosen": -316.7195739746094,
|
| 637 |
+
"logps/rejected": -307.0235900878906,
|
| 638 |
+
"loss": 0.6755961894989013,
|
| 639 |
+
"rewards/accuracies": 0.581250011920929,
|
| 640 |
+
"rewards/chosen": 0.45080599188804626,
|
| 641 |
+
"rewards/margins": 0.048326827585697174,
|
| 642 |
+
"rewards/rejected": 0.4024791717529297,
|
| 643 |
+
"step": 390
|
| 644 |
+
},
|
| 645 |
+
{
|
| 646 |
+
"epoch": 2.7791304347826085,
|
| 647 |
+
"grad_norm": 77.0,
|
| 648 |
+
"learning_rate": 8.871331300335321e-09,
|
| 649 |
+
"logits/chosen": -1.0839048624038696,
|
| 650 |
+
"logits/rejected": -1.0926908254623413,
|
| 651 |
+
"logps/chosen": -324.00946044921875,
|
| 652 |
+
"logps/rejected": -318.5250549316406,
|
| 653 |
+
"loss": 0.6874616622924805,
|
| 654 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 655 |
+
"rewards/chosen": 0.45788002014160156,
|
| 656 |
+
"rewards/margins": 0.025683864951133728,
|
| 657 |
+
"rewards/rejected": 0.4321961998939514,
|
| 658 |
+
"step": 400
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"epoch": 2.7791304347826085,
|
| 662 |
+
"eval_logits/chosen": -1.059512972831726,
|
| 663 |
+
"eval_logits/rejected": -1.0796120166778564,
|
| 664 |
+
"eval_logps/chosen": -315.1502990722656,
|
| 665 |
+
"eval_logps/rejected": -316.1904296875,
|
| 666 |
+
"eval_loss": 0.6914414763450623,
|
| 667 |
+
"eval_rewards/accuracies": 0.5322580933570862,
|
| 668 |
+
"eval_rewards/chosen": 0.4315947890281677,
|
| 669 |
+
"eval_rewards/margins": 0.01160100195556879,
|
| 670 |
+
"eval_rewards/rejected": 0.4199938178062439,
|
| 671 |
+
"eval_runtime": 11.25,
|
| 672 |
+
"eval_samples_per_second": 10.756,
|
| 673 |
+
"eval_steps_per_second": 2.756,
|
| 674 |
+
"step": 400
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 2.8486956521739133,
|
| 678 |
+
"grad_norm": 41.75,
|
| 679 |
+
"learning_rate": 4.3226141225268796e-09,
|
| 680 |
+
"logits/chosen": -1.0775182247161865,
|
| 681 |
+
"logits/rejected": -1.0758936405181885,
|
| 682 |
+
"logps/chosen": -314.0296325683594,
|
| 683 |
+
"logps/rejected": -312.0280456542969,
|
| 684 |
+
"loss": 0.6733474254608154,
|
| 685 |
+
"rewards/accuracies": 0.574999988079071,
|
| 686 |
+
"rewards/chosen": 0.450728178024292,
|
| 687 |
+
"rewards/margins": 0.04884380102157593,
|
| 688 |
+
"rewards/rejected": 0.40188440680503845,
|
| 689 |
+
"step": 410
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 2.9182608695652172,
|
| 693 |
+
"grad_norm": 47.25,
|
| 694 |
+
"learning_rate": 1.383668301212393e-09,
|
| 695 |
+
"logits/chosen": -1.0620819330215454,
|
| 696 |
+
"logits/rejected": -1.0514990091323853,
|
| 697 |
+
"logps/chosen": -332.62652587890625,
|
| 698 |
+
"logps/rejected": -328.03424072265625,
|
| 699 |
+
"loss": 0.6754622459411621,
|
| 700 |
+
"rewards/accuracies": 0.574999988079071,
|
| 701 |
+
"rewards/chosen": 0.459503710269928,
|
| 702 |
+
"rewards/margins": 0.04998493567109108,
|
| 703 |
+
"rewards/rejected": 0.4095187783241272,
|
| 704 |
+
"step": 420
|
| 705 |
+
},
|
| 706 |
+
{
|
| 707 |
+
"epoch": 2.9878260869565216,
|
| 708 |
+
"grad_norm": 52.25,
|
| 709 |
+
"learning_rate": 7.375092342298828e-11,
|
| 710 |
+
"logits/chosen": -1.0443447828292847,
|
| 711 |
+
"logits/rejected": -1.0698919296264648,
|
| 712 |
+
"logps/chosen": -325.37957763671875,
|
| 713 |
+
"logps/rejected": -314.8664245605469,
|
| 714 |
+
"loss": 0.6853389263153076,
|
| 715 |
+
"rewards/accuracies": 0.5625,
|
| 716 |
+
"rewards/chosen": 0.4191187918186188,
|
| 717 |
+
"rewards/margins": 0.025067320093512535,
|
| 718 |
+
"rewards/rejected": 0.3940514922142029,
|
| 719 |
+
"step": 430
|
| 720 |
+
}
|
| 721 |
+
],
|
| 722 |
+
"logging_steps": 10,
|
| 723 |
+
"max_steps": 432,
|
| 724 |
+
"num_input_tokens_seen": 0,
|
| 725 |
+
"num_train_epochs": 3,
|
| 726 |
+
"save_steps": 100,
|
| 727 |
+
"stateful_callbacks": {
|
| 728 |
+
"TrainerControl": {
|
| 729 |
+
"args": {
|
| 730 |
+
"should_epoch_stop": false,
|
| 731 |
+
"should_evaluate": false,
|
| 732 |
+
"should_log": false,
|
| 733 |
+
"should_save": true,
|
| 734 |
+
"should_training_stop": true
|
| 735 |
+
},
|
| 736 |
+
"attributes": {}
|
| 737 |
+
}
|
| 738 |
+
},
|
| 739 |
+
"total_flos": 0.0,
|
| 740 |
+
"train_batch_size": 4,
|
| 741 |
+
"trial_name": null,
|
| 742 |
+
"trial_params": null
|
| 743 |
+
}
|
checkpoint-432/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b05394a09844ad8fbc5bf2ac3f5b6be474605707ae7d97f68562a3f9312197c4
|
| 3 |
+
size 6225
|
final_model/chat_template.jinja
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{- bos_token -}}
|
| 2 |
+
{%- set keep_past_thinking = keep_past_thinking | default(false) -%}
|
| 3 |
+
{%- set ns = namespace(system_prompt="") -%}
|
| 4 |
+
{%- if messages[0]["role"] == "system" -%}
|
| 5 |
+
{%- set ns.system_prompt = messages[0]["content"] -%}
|
| 6 |
+
{%- set messages = messages[1:] -%}
|
| 7 |
+
{%- endif -%}
|
| 8 |
+
{%- if tools -%}
|
| 9 |
+
{%- set ns.system_prompt = ns.system_prompt + ("\n" if ns.system_prompt else "") + "List of tools: [" -%}
|
| 10 |
+
{%- for tool in tools -%}
|
| 11 |
+
{%- if tool is not string -%}
|
| 12 |
+
{%- set tool = tool | tojson -%}
|
| 13 |
+
{%- endif -%}
|
| 14 |
+
{%- set ns.system_prompt = ns.system_prompt + tool -%}
|
| 15 |
+
{%- if not loop.last -%}
|
| 16 |
+
{%- set ns.system_prompt = ns.system_prompt + ", " -%}
|
| 17 |
+
{%- endif -%}
|
| 18 |
+
{%- endfor -%}
|
| 19 |
+
{%- set ns.system_prompt = ns.system_prompt + "]" -%}
|
| 20 |
+
{%- endif -%}
|
| 21 |
+
{%- if ns.system_prompt -%}
|
| 22 |
+
{{- "<|im_start|>system\n" + ns.system_prompt + "<|im_end|>\n" -}}
|
| 23 |
+
{%- endif -%}
|
| 24 |
+
{%- set ns.last_assistant_index = -1 -%}
|
| 25 |
+
{%- for message in messages -%}
|
| 26 |
+
{%- if message["role"] == "assistant" -%}
|
| 27 |
+
{%- set ns.last_assistant_index = loop.index0 -%}
|
| 28 |
+
{%- endif -%}
|
| 29 |
+
{%- endfor -%}
|
| 30 |
+
{%- for message in messages -%}
|
| 31 |
+
{{- "<|im_start|>" + message["role"] + "\n" -}}
|
| 32 |
+
{%- set content = message["content"] -%}
|
| 33 |
+
{%- if content is not string -%}
|
| 34 |
+
{%- set content = content | tojson -%}
|
| 35 |
+
{%- endif -%}
|
| 36 |
+
{%- if message["role"] == "assistant" and not keep_past_thinking and loop.index0 != ns.last_assistant_index -%}
|
| 37 |
+
{%- if "</think>" in content -%}
|
| 38 |
+
{%- set content = content.split("</think>")[-1] | trim -%}
|
| 39 |
+
{%- endif -%}
|
| 40 |
+
{%- endif -%}
|
| 41 |
+
{{- content + "<|im_end|>\n" -}}
|
| 42 |
+
{%- endfor -%}
|
| 43 |
+
{%- if add_generation_prompt -%}
|
| 44 |
+
{{- "<|im_start|>assistant\n" -}}
|
| 45 |
+
{%- endif -%}
|
final_model/config.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Lfm2ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"block_auto_adjust_ff_dim": true,
|
| 6 |
+
"block_dim": 2048,
|
| 7 |
+
"block_ff_dim": 12288,
|
| 8 |
+
"block_ffn_dim_multiplier": 1.0,
|
| 9 |
+
"block_mlp_init_scale": 1.0,
|
| 10 |
+
"block_multiple_of": 256,
|
| 11 |
+
"block_norm_eps": 1e-05,
|
| 12 |
+
"block_out_init_scale": 1.0,
|
| 13 |
+
"block_use_swiglu": true,
|
| 14 |
+
"block_use_xavier_init": true,
|
| 15 |
+
"bos_token_id": 1,
|
| 16 |
+
"conv_L_cache": 3,
|
| 17 |
+
"conv_bias": false,
|
| 18 |
+
"conv_dim": 2048,
|
| 19 |
+
"conv_use_xavier_init": true,
|
| 20 |
+
"dtype": "bfloat16",
|
| 21 |
+
"eos_token_id": 7,
|
| 22 |
+
"hidden_size": 2048,
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 12288,
|
| 25 |
+
"layer_types": [
|
| 26 |
+
"conv",
|
| 27 |
+
"conv",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"conv",
|
| 30 |
+
"conv",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"conv",
|
| 33 |
+
"conv",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"conv",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"conv",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"conv",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"conv"
|
| 42 |
+
],
|
| 43 |
+
"max_position_embeddings": 128000,
|
| 44 |
+
"model_type": "lfm2",
|
| 45 |
+
"norm_eps": 1e-05,
|
| 46 |
+
"num_attention_heads": 32,
|
| 47 |
+
"num_heads": 32,
|
| 48 |
+
"num_hidden_layers": 16,
|
| 49 |
+
"num_key_value_heads": 8,
|
| 50 |
+
"pad_token_id": 0,
|
| 51 |
+
"rope_parameters": {
|
| 52 |
+
"rope_theta": 1000000.0,
|
| 53 |
+
"rope_type": "default"
|
| 54 |
+
},
|
| 55 |
+
"tie_embedding": true,
|
| 56 |
+
"tie_word_embeddings": true,
|
| 57 |
+
"transformers_version": "5.2.0",
|
| 58 |
+
"use_cache": false,
|
| 59 |
+
"use_pos_enc": true,
|
| 60 |
+
"vocab_size": 65536
|
| 61 |
+
}
|
final_model/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
7
|
| 6 |
+
],
|
| 7 |
+
"pad_token_id": 0,
|
| 8 |
+
"transformers_version": "5.2.0"
|
| 9 |
+
}
|
final_model/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ea8a8b6c80f8066d94803546515de24808278c0cfb7efff8e029129e40dd94ff
|
| 3 |
+
size 2340697936
|
final_model/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
final_model/tokenizer_config.json
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"bos_token": "<|startoftext|>",
|
| 4 |
+
"clean_up_tokenization_spaces": false,
|
| 5 |
+
"eos_token": "<|im_end|>",
|
| 6 |
+
"is_local": false,
|
| 7 |
+
"legacy": false,
|
| 8 |
+
"model_input_names": [
|
| 9 |
+
"input_ids",
|
| 10 |
+
"attention_mask"
|
| 11 |
+
],
|
| 12 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 13 |
+
"pad_token": "<|pad|>",
|
| 14 |
+
"padding_side": "right",
|
| 15 |
+
"sp_model_kwargs": {},
|
| 16 |
+
"spaces_between_special_tokens": false,
|
| 17 |
+
"tokenizer_class": "TokenizersBackend",
|
| 18 |
+
"use_default_system_prompt": false,
|
| 19 |
+
"use_fast": true
|
| 20 |
+
}
|
final_model/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b05394a09844ad8fbc5bf2ac3f5b6be474605707ae7d97f68562a3f9312197c4
|
| 3 |
+
size 6225
|
run_meta.json
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"run_name": "dpo_fft_LFM2.5-1.2B-Instruct_argilla__distilabel-math-preference-dpo_20260222_195126",
|
| 3 |
+
"model": "LiquidAI/LFM2.5-1.2B-Instruct",
|
| 4 |
+
"dataset": "argilla/distilabel-math-preference-dpo",
|
| 5 |
+
"timestamp": "20260222_195126",
|
| 6 |
+
"args": {
|
| 7 |
+
"dataset": "argilla/distilabel-math-preference-dpo",
|
| 8 |
+
"dataset_split": "train",
|
| 9 |
+
"instruction_col": null,
|
| 10 |
+
"chosen_col": null,
|
| 11 |
+
"rejected_col": null,
|
| 12 |
+
"max_samples": null,
|
| 13 |
+
"seed": 42,
|
| 14 |
+
"model_name": "LiquidAI/LFM2.5-1.2B-Instruct",
|
| 15 |
+
"ref_4bit": false,
|
| 16 |
+
"num_epochs": 3,
|
| 17 |
+
"batch_size": 4,
|
| 18 |
+
"grad_accum": 4,
|
| 19 |
+
"learning_rate": 5e-07,
|
| 20 |
+
"beta": 0.1,
|
| 21 |
+
"max_length": 1024,
|
| 22 |
+
"max_prompt_length": 512,
|
| 23 |
+
"warmup_ratio": 0.1,
|
| 24 |
+
"optim": "paged_adamw_8bit",
|
| 25 |
+
"logging_steps": 10,
|
| 26 |
+
"save_steps": 100,
|
| 27 |
+
"eval_ratio": 0.05,
|
| 28 |
+
"output_dir": "models",
|
| 29 |
+
"run_name": null
|
| 30 |
+
},
|
| 31 |
+
"train_metrics": {
|
| 32 |
+
"train_runtime": 1608.535,
|
| 33 |
+
"train_samples_per_second": 4.284,
|
| 34 |
+
"train_steps_per_second": 0.269,
|
| 35 |
+
"total_flos": 0.0,
|
| 36 |
+
"train_loss": 0.6827363446354866,
|
| 37 |
+
"epoch": 3.0
|
| 38 |
+
}
|
| 39 |
+
}
|