End of training

README.md

@@ -1,5 +1,5 @@
 ---
-base_model: Gensyn/Qwen2.5-0.5B-Instruct
+base_model: unsloth/Qwen2.5-0.5B-Instruct
 library_name: transformers
 model_name: Qwen2.5-0.5B-Instruct-Gensyn-Swarm-mangy_padded_cow
 tags:
@@ -14,7 +14,7 @@ licence: license
 
 # Model Card for Qwen2.5-0.5B-Instruct-Gensyn-Swarm-mangy_padded_cow
 
-This model is a fine-tuned version of [Gensyn/Qwen2.5-0.5B-Instruct](https://huggingface.co/Gensyn/Qwen2.5-0.5B-Instruct).
+This model is a fine-tuned version of [unsloth/Qwen2.5-0.5B-Instruct](https://huggingface.co/unsloth/Qwen2.5-0.5B-Instruct).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start
@@ -37,10 +37,10 @@ This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing
 
 ### Framework versions
 
-- TRL: 0.15.2
+- TRL: 0.17.0
 - Transformers: 4.51.3
-- Pytorch: 2.6.0
-- Datasets: 3.5.0
+- Pytorch: 2.7.0
+- Datasets: 3.5.1
 - Tokenizers: 0.21.1
 
 ## Citations
@@ -62,7 +62,7 @@ Cite TRL as:
 ```bibtex
 @misc{vonwerra2022trl,
 	title        = {{TRL: Transformer Reinforcement Learning}},
-	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
+	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
 	year         = 2020,
 	journal      = {GitHub repository},
 	publisher    = {GitHub},

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.0,
-    "train_runtime": 863.957,
-    "train_samples": 12,
-    "train_samples_per_second": 0.093,
-    "train_steps_per_second": 0.006
+    "train_loss": 0.27013154824574787,
+    "train_runtime": 1786.8594,
+    "train_samples": 23,
+    "train_samples_per_second": 0.013,
+    "train_steps_per_second": 0.002
 }

config.json

@@ -3,7 +3,6 @@
     "Qwen2ForCausalLM"
   ],
   "attention_dropout": 0.0,
-  "bos_token_id": 151643,
   "eos_token_id": 151645,
   "hidden_act": "silu",
   "hidden_size": 896,
@@ -15,13 +14,15 @@
   "num_attention_heads": 14,
   "num_hidden_layers": 24,
   "num_key_value_heads": 2,
+  "pad_token_id": 151654,
   "rms_norm_eps": 1e-06,
   "rope_scaling": null,
   "rope_theta": 1000000.0,
-  "sliding_window": 32768,
+  "sliding_window": null,
   "tie_word_embeddings": true,
   "torch_dtype": "float32",
   "transformers_version": "4.51.3",
+  "unsloth_fixed": true,
   "use_cache": true,
   "use_sliding_window": false,
   "vocab_size": 151936

generation_config.json

@@ -5,7 +5,8 @@
     151645,
     151643
   ],
-  "pad_token_id": 151643,
+  "max_length": 32768,
+  "pad_token_id": 151654,
   "repetition_penalty": 1.1,
   "temperature": 0.7,
   "top_k": 20,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cc6798eb79dc628e086c92abff71e217c8353df86dcc554b4d9ae22c5af6daa6
+oid sha256:042df9bab814993de332623fa5c40266568096ef8b903104ca3c676525f23d13
 size 1976163472

special_tokens_map.json

@@ -22,7 +22,7 @@
     "single_word": false
   },
   "pad_token": {
-    "content": "<|endoftext|>",
+    "content": "<|vision_pad|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5eee858c5123a4279c3e1f7b81247343f356ac767940b2692a928ad929543214
-size 11422063
+oid sha256:64e71213db910f5cafa86d35091f37393dcc344b1bbc34091d1b3eed4cca01d5
+size 11422064

tokenizer_config.json

@@ -200,8 +200,9 @@
   "eos_token": "<|im_end|>",
   "errors": "replace",
   "extra_special_tokens": {},
-  "model_max_length": 131072,
-  "pad_token": "<|endoftext|>",
+  "model_max_length": 32768,
+  "pad_token": "<|vision_pad|>",
+  "padding_side": "left",
   "split_special_tokens": false,
   "tokenizer_class": "Qwen2Tokenizer",
   "unk_token": null

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.0,
-    "train_runtime": 863.957,
-    "train_samples": 12,
-    "train_samples_per_second": 0.093,
-    "train_steps_per_second": 0.006
+    "train_loss": 0.27013154824574787,
+    "train_runtime": 1786.8594,
+    "train_samples": 23,
+    "train_samples_per_second": 0.013,
+    "train_steps_per_second": 0.002
 }

trainer_state.json

@@ -2,77 +2,98 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 2.6666666666666665,
+  "epoch": 0.5217391304347826,
   "eval_steps": 500,
-  "global_step": 5,
+  "global_step": 3,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 163.66666666666666,
-      "epoch": 1.0,
-      "grad_norm": 3.171876650220895e-10,
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completions/clipped_ratio": 0.125,
+      "completions/max_length": 759.0,
+      "completions/max_terminated_length": 544.0,
+      "completions/mean_length": 344.3125,
+      "completions/mean_terminated_length": 246.29166412353516,
+      "completions/min_length": 64.5,
+      "completions/min_terminated_length": 64.5,
+      "epoch": 0.34782608695652173,
+      "grad_norm": 6.1356120109558105,
       "kl": 0.0,
       "learning_rate": 5e-07,
-      "loss": 0.0,
-      "reward": 7.75,
-      "reward_std": 0.0,
-      "rewards/concensus_correctness_reward_func": 2.5,
-      "rewards/consensus_reward_func": 2.0,
-      "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.5,
-      "rewards/question_recreation_reward_func": 1.0,
-      "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.5,
-      "rewards/xmlcount_reward_func": 1.25,
+      "loss": 0.1231,
+      "num_tokens": 9605.0,
+      "reward": 0.0872948095202446,
+      "reward_std": 0.036231483216397464,
+      "rewards/concensus_correctness_reward_func/mean": 0.0,
+      "rewards/concensus_correctness_reward_func/std": 0.0,
+      "rewards/consensus_reward_func/mean": 0.0,
+      "rewards/consensus_reward_func/std": 0.0,
+      "rewards/cumulative_reward_2/mean": 0.0,
+      "rewards/cumulative_reward_2/std": 0.0,
+      "rewards/final_correctness_reward_func/mean": 0.0,
+      "rewards/final_correctness_reward_func/std": 0.0,
+      "rewards/question_recreation_reward_func/mean": 0.0872948169708252,
+      "rewards/question_recreation_reward_func/std": 0.042022028006613255,
+      "rewards/soft_format_reward_func/mean": 0.0,
+      "rewards/soft_format_reward_func/std": 0.0,
+      "rewards/strict_format_reward_func/mean": 0.0,
+      "rewards/strict_format_reward_func/std": 0.0,
+      "rewards/xmlcount_reward_func/mean": 0.0,
+      "rewards/xmlcount_reward_func/std": 0.0,
       "step": 2
     },
     {
-      "completion_length": 161.16666666666666,
-      "epoch": 2.0,
-      "grad_norm": 5.25152259456263e-08,
-      "kl": 0.0,
-      "learning_rate": 2.5e-07,
-      "loss": 0.0,
-      "reward": 7.75,
-      "reward_std": 0.0,
-      "rewards/concensus_correctness_reward_func": 2.5,
-      "rewards/consensus_reward_func": 2.0,
-      "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.5,
-      "rewards/question_recreation_reward_func": 1.0,
-      "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.5,
-      "rewards/xmlcount_reward_func": 1.25,
-      "step": 4
-    },
-    {
-      "completion_length": 160.0,
-      "epoch": 2.6666666666666665,
-      "kl": 0.0,
-      "reward": 8.25,
-      "reward_std": 0.0,
-      "rewards/concensus_correctness_reward_func": 2.75,
-      "rewards/consensus_reward_func": 2.0,
-      "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.75,
-      "rewards/question_recreation_reward_func": 1.0,
-      "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.5,
-      "rewards/xmlcount_reward_func": 1.25,
-      "step": 5,
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 369.0,
+      "completions/max_terminated_length": 369.0,
+      "completions/mean_length": 126.25,
+      "completions/mean_terminated_length": 126.25,
+      "completions/min_length": 2.0,
+      "completions/min_terminated_length": 2.0,
+      "epoch": 0.5217391304347826,
+      "kl": 0.0009940905874827877,
+      "num_tokens": 12663.0,
+      "reward": 2.516864776611328,
+      "reward_std": 3.5382814407348633,
+      "rewards/concensus_correctness_reward_func/mean": 2.5,
+      "rewards/concensus_correctness_reward_func/std": 7.071068286895752,
+      "rewards/consensus_reward_func/mean": 0.0,
+      "rewards/consensus_reward_func/std": 0.0,
+      "rewards/cumulative_reward_2/mean": 0.0,
+      "rewards/cumulative_reward_2/std": 0.0,
+      "rewards/final_correctness_reward_func/mean": 0.0,
+      "rewards/final_correctness_reward_func/std": 0.0,
+      "rewards/question_recreation_reward_func/mean": 0.01686476171016693,
+      "rewards/question_recreation_reward_func/std": 0.01196382101625204,
+      "rewards/soft_format_reward_func/mean": 0.0,
+      "rewards/soft_format_reward_func/std": 0.0,
+      "rewards/strict_format_reward_func/mean": 0.0,
+      "rewards/strict_format_reward_func/std": 0.0,
+      "rewards/xmlcount_reward_func/mean": 0.0,
+      "rewards/xmlcount_reward_func/std": 0.0,
+      "step": 3,
       "total_flos": 0.0,
-      "train_loss": 0.0,
-      "train_runtime": 863.957,
-      "train_samples_per_second": 0.093,
-      "train_steps_per_second": 0.006
+      "train_loss": 0.27013154824574787,
+      "train_runtime": 1786.8594,
+      "train_samples_per_second": 0.013,
+      "train_steps_per_second": 0.002
     }
   ],
   "logging_steps": 2,
-  "max_steps": 5,
-  "num_input_tokens_seen": 0,
-  "num_train_epochs": 5,
+  "max_steps": 3,
+  "num_input_tokens_seen": 12663,
+  "num_train_epochs": 1,
   "save_steps": 25,
   "stateful_callbacks": {
     "TrainerControl": {

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:85368bab1dd51be2cbdf5e0109659802ffed3c70a6bb03cb0591e9dcae7e152d
-size 5944
+oid sha256:6a71230f32e732eea5fcb1a0593b7bc74ff3864c9b7b760cda833ba20848a0ed
+size 6929