Training in progress, step 267

README.md

@@ -1,18 +1,18 @@
 ---
-base_model: Ba2han/qwen-test-3-longer-2
+base_model: Ba2han/qwen3_from_scratch
 library_name: transformers
 model_name: model-sft-q2
 tags:
 - generated_from_trainer
-- trl
 - unsloth
 - sft
+- trl
 licence: license
 ---
 
 # Model Card for model-sft-q2
 
-This model is a fine-tuned version of [Ba2han/qwen-test-3-longer-2](https://huggingface.co/Ba2han/qwen-test-3-longer-2).
+This model is a fine-tuned version of [Ba2han/qwen3_from_scratch](https://huggingface.co/Ba2han/qwen3_from_scratch).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start
@@ -28,7 +28,7 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/batuhan409/huggingface/runs/j7uw0ag9) 
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/batuhan409/huggingface/runs/ye3goxhy) 
 
 
 This model was trained with SFT.

chat_template.jinja

@@ -0,0 +1,4 @@
+{% for message in messages %}{{'<|im_start|>' + message['role'] + '
+' + message['content'].strip() + '<|im_end|>' + '
+'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant
+' }}{% endif %}

config.json

@@ -1,17 +1,20 @@
 {
   "architectures": [
-    "Qwen3ForCausalLM"
+    "SquaredReLUQwen3ForCausalLM"
   ],
   "attention_bias": false,
   "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoModelForCausalLM": "patch.SquaredReLUQwen3ForCausalLM"
+  },
   "bos_token_id": 50030,
   "dtype": "bfloat16",
-  "eos_token_id": 50031,
-  "head_dim": 96,
-  "hidden_act": "silu",
-  "hidden_size": 1152,
+  "eos_token_id": 50049,
+  "head_dim": 128,
+  "hidden_act": "squared_relu",
+  "hidden_size": 1024,
   "initializer_range": 0.02,
-  "intermediate_size": 2880,
+  "intermediate_size": 2816,
   "layer_types": [
     "full_attention",
     "full_attention",
@@ -52,25 +55,27 @@
     "full_attention",
     "full_attention",
     "full_attention",
+    "full_attention",
+    "full_attention",
     "full_attention"
   ],
   "max_position_embeddings": 8192,
-  "max_window_layers": 40,
-  "model_name": "Ba2han/qwen-test-3-longer-2",
+  "max_window_layers": 28,
+  "model_name": "Ba2han/qwen3_from_scratch",
   "model_type": "qwen3",
   "num_attention_heads": 8,
-  "num_hidden_layers": 40,
-  "num_key_value_heads": 8,
+  "num_hidden_layers": 42,
+  "num_key_value_heads": 2,
   "pad_token_id": 50034,
   "rms_norm_eps": 1e-06,
   "rope_parameters": {
-    "rope_theta": 1000000,
+    "rope_theta": 10000.0,
     "rope_type": "default"
   },
   "sliding_window": null,
-  "tie_word_embeddings": false,
+  "tie_word_embeddings": true,
   "transformers_version": "5.5.0",
-  "unsloth_version": "2026.4.1",
+  "unsloth_version": "2026.5.7",
   "use_cache": false,
   "use_sliding_window": false,
   "vocab_size": 50050

generation_config.json

@@ -2,6 +2,7 @@
   "_from_model_config": true,
   "bos_token_id": 50030,
   "eos_token_id": [
+    50049,
     50031
   ],
   "max_length": 8192,
@@ -9,5 +10,5 @@
   "output_hidden_states": false,
   "pad_token_id": 50034,
   "transformers_version": "5.5.0",
-  "use_cache": true
+  "use_cache": false
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e96d39f6489966dc2411487d22641305de24d001d6d0f08631c48ce89fd00a7f
-size 1310260536
+oid sha256:c634698e8c168625d167a6eb13258edeefc99eca29d8affc045abb8526782e58
+size 1049614696

tokenizer_config.json

@@ -3,7 +3,7 @@
   "bos_token": "<|begin_of_text|>",
   "clean_up_tokenization_spaces": true,
   "eos_token": "<|im_end|>",
-  "additional_special_tokens": [
+  "extra_special_tokens": [
     "<|im_start|>",
     "<|im_end|>"
   ],
@@ -15,6 +15,168 @@
   "model_max_length": 8192,
   "pad_token": "<|finetune_right_pad_id|>",
   "padding_side": "right",
-  "tokenizer_class": "PreTrainedTokenizerFast",
-  "unk_token": null
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "added_tokens_decoder": {
+    "50030": {
+      "content": "<|begin_of_text|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50031": {
+      "content": "<|end_of_text|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50032": {
+      "content": "<|reserved_special_token_0|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50033": {
+      "content": "<|reserved_special_token_1|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50034": {
+      "content": "<|finetune_right_pad_id|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50035": {
+      "content": "<|reserved_special_token_2|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50036": {
+      "content": "<|start_header_id|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50037": {
+      "content": "<|end_header_id|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50038": {
+      "content": "<|eom_id|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50039": {
+      "content": "<|eot_id|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50040": {
+      "content": "<|python_tag|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50041": {
+      "content": "<|reserved_special_token_3|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50042": {
+      "content": "<|reserved_special_token_4|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50043": {
+      "content": "<|reserved_special_token_5|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50044": {
+      "content": "<|reserved_special_token_6|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50045": {
+      "content": "<|reserved_special_token_7|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50046": {
+      "content": "<|reserved_special_token_8|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50047": {
+      "content": "<|reserved_special_token_9|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50048": {
+      "content": "<|im_start|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    "50049": {
+      "content": "<|im_end|>",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    }
+  }
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6ce9f89c60b15bdcc2a04b969f0b2369be2edd0951e2475d4f933d4ddce91dd1
+oid sha256:80a6c8d882cb703ede24f73dc39c2a799e65102964d1fc9b53392cf548d27209
 size 5713