Automated MNLP evaluation report (2026-05-23)

EVAL_REPORT.md

@@ -0,0 +1,87 @@
+# Automated MNLP evaluation report
+
+- **Model repo:** [`cs-552-2026-the-transformers/math_model`](https://huggingface.co/cs-552-2026-the-transformers/math_model)
+- **Owner(s):** group **the-transformers**
+- **Generated at:** 2026-05-23T14:37:35+00:00 (UTC)
+- **Pipeline:** [mnlp-project-ci](https://github.com/eric11eca/mnlp-project-ci)
+
+_This PR is opened automatically by the course CI. It is **non-blocking** — you do not need to merge it. The next nightly run will refresh this file._
+
+## Summary
+
+| Benchmark | Accuracy | Status |
+|---|---:|---|
+| Math | 0.3600 | ok |
+| Knowledge | — | not run |
+| Multilingual | — | not run |
+| Safety | — | not run |
+
+## Sample completions
+
+_Prompts are intentionally omitted to avoid revealing benchmark contents. For multi-completion problems, only one completion is shown per sample._
+
+### Math
+
+**Correct** (1 shown)
+
+- **reference**: `194`
+- **overall** (3/8 completions correct)
+- **extracted** (✓): `194`
+- **completion**:
+
+    ```text
+    <think>
+    To solve this problem, we need to understand the conditions under which a \(194 \times 194\) board can be tiled with \(1 \times 2\) and \(2 \times 1\) dominoes, and how marking cells affects this tiling.
+    
+    First, we note that a \(194 \times 194\) board has an area of \(194^2 = 37636\) cells. Each domino covers exactly 2 cells, so the number of dominoes required to tile the board is \(37636
+    
+    ... [truncated 147 chars] ...
+    
+    ontains at most one marked cell. This means that the marked cells can be placed in such a way that no two marked cells are in the same domino.
+    
+    To find the minimum number of marked cells, we need to consider the maximum number of dominoes that can be placed such that no two marked cells are in the same domino. This is equivalent to placing the marked cells in such a way that they are not in the same row or column, as dominoes span two rows and two columns.
+    
+    If we place a marked cell in each row and each column, we can ensure that no two marked cells are in the same row or column. This is similar to placing non-attacking rooks on a chessboard, where no two rooks attack each other.
+    
+    For a \(194 \times 194\) board, the maximum number of non-attacking rooks we can place is 194 (one per row and one per column). Therefore, the minimum number of marked cells required to ensure that no two marked cells are in the same row or column is 194.
+    
+    Thus, the least positive integer \(k\) for which one can choose \(k\) cells to mark so that there is exactly one tiling of the board by \(1 \times 2\) and \(2 \times 1\) dominoes in which every domino contains at most one marked cell is \(\boxed{194}\).
+    ```
+
+**Incorrect** (1 shown)
+
+- **reference**: `$g(x)=c, g(x)=\lceil x \rceil, g(x)=\lfloor x \rfloor$`
+- **overall** (0/8 completions correct)
+- **extracted** (✗): `h(x) = c \text{ for all } x \in \mathbb{Q} \text{ and some } c \in \mathbb{Z}`
+- **completion**:
+
+    ```text
+    <think>
+    Given the problem, we need to find all mappings $h:\mathbb{Q}\to\mathbb{Z}$ such that for every $x\in\mathbb{Q}$, every $a\in\mathbb{Z}$, and every positive integer $b$, the following equation holds:
+    \[
+    h(x)=h\!\left(\frac{h(bx-a)+a}{b}\right).
+    \]
+    
+    First, we can simplify the given equation by setting $x=0$. Substituting $x=0$ into the equation, we get:
+    \[
+    h(0) = h\!\left(\frac{h(0-a)+a}{b}
+    
+    ... [truncated 371 chars] ...
+    
+    r all $a$ and $b$. Therefore, $h$ must be constant on the set of all rational numbers. Let's denote this constant value by $c$. So, we have:
+    \[
+    h(x) = c \quad \text{for all} \quad x \in \mathbb{Q}.
+    \]
+    Now, we need to check if this constant function satisfies the given equation. Substituting $h(x) = c$ into the original equation, we get:
+    \[
+    c = c \quad \text{which is true for all} \quad c \in \mathbb{Z}.
+    \]
+    Thus, the constant function $h(x) = c$ for all $x \in \mathbb{Q}$ is a solution. To confirm that this is the only solution, we need to ensure that no other functions can satisfy the given equation. Suppose there exists a non-constant function $h$ that satisfies the equation. Then, there must exist some $x$ for which $h(x) \neq c$. However, the condition $h(x) = h\!\left(\frac{h(bx-a)+a}{b}\right)$ must hold for all $x$, $a$, and $b$, and since $h$ is constant on all rational numbers, this condition can only be satisfied if $h$ is constant. Therefore, the only possible solution is the constant function $h(x) = c$ for all $x \in \mathbb{Q}$.
+    
+    Hence, the solution to the problem is:
+    \[
+    </think>
+    
+    \boxed{h(x) = c \text{ for all } x \in \mathbb{Q} \text{ and some } c \in \mathbb{Z}}.
+    \]
+    ```

README.md

@@ -1,14 +1,15 @@
 ---
 base_model: Qwen/Qwen3-1.7B
 library_name: transformers
+model_name: math_sft_test
 tags:
 - generated_from_trainer
-- trl
 - sft
+- trl
 licence: license
 ---
 
-# math_model
+# Model Card for math_sft_test
 
 This model is a fine-tuned version of [Qwen/Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B).
 It has been trained using [TRL](https://github.com/huggingface/trl).
@@ -19,22 +20,23 @@ It has been trained using [TRL](https://github.com/huggingface/trl).
 from transformers import pipeline
 
 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?"
-generator = pipeline("text-generation", model="cs-552-2026-the-transformers/math_model", device="cuda")
+generator = pipeline("text-generation", model="None", device="cuda")
 output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
 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/jdecim/MNLP/runs/2btc84l3) 
+ 
+
 
 
 This model was trained with SFT.
 
 ### Framework versions
 
-- TRL: 0.27.2
-- Transformers: 5.8.0
+- TRL: 1.3.0
+- Transformers: 5.7.0
 - Pytorch: 2.10.0+cu128
 - Datasets: 4.8.5
 - Tokenizers: 0.22.2
@@ -46,12 +48,11 @@ This model was trained with SFT.
 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 Gallou{\'e}dec},
-	year         = 2020,
-	journal      = {GitHub repository},
-	publisher    = {GitHub},
-	howpublished = {\url{https://github.com/huggingface/trl}}
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  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},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
 }
 ```

chat_template.jinja

@@ -10,13 +10,9 @@
     {%- endfor %}
     {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
 {%- else %}
-    {#- BOXING INSTRUCTION injected so it is present even when the caller (e.g. CI)
-        passes no system message. Targets our no-box / multi-box failure mode. #}
-    {{- '<|im_start|>system\n' }}
     {%- if messages[0].role == 'system' %}
-        {{- messages[0].content + '\n\n' }}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
     {%- endif %}
-    {{- 'When solving a math problem, work through it step by step and write your final answer inside a \boxed{...}.<|im_end|>\n' }}
 {%- endif %}
 {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
 {%- for message in messages[::-1] %}

checkpoint-125/chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

checkpoint-125/config.json

@@ -0,0 +1,63 @@
+{
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.7.0",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

checkpoint-125/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "5.7.0"
+}

checkpoint-125/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7074b51b321535051b77436f30720111854403259c583a273488e061c0d23072
+size 3441185608

checkpoint-125/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b569f0710425a088cc5cd95c2def7bbb89539cdd1c3b41312c8d8fc136d2e186
+size 6882572207

checkpoint-125/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61c19bab1174704a4a4441475683bf1270277af15d2e2c95e964789128e482c4
+size 14645

checkpoint-125/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:770e85488b2d6f2b41a590e6c2bb599f3a174081e14b8c48d3638681b4a06597
+size 1465

checkpoint-125/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
+size 11422650

checkpoint-125/tokenizer_config.json

@@ -0,0 +1,30 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "local_files_only": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-125/trainer_state.json

@@ -0,0 +1,154 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 125,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 0.2860680317506194,
+      "epoch": 0.08,
+      "grad_norm": 7.46875,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.7724831104278564,
+      "mean_token_accuracy": 0.8434554561972618,
+      "num_tokens": 94518.0,
+      "step": 10
+    },
+    {
+      "entropy": 0.3743372498080134,
+      "epoch": 0.16,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.8761061946902657e-05,
+      "loss": 0.4372231483459473,
+      "mean_token_accuracy": 0.8732902690768242,
+      "num_tokens": 187445.0,
+      "step": 20
+    },
+    {
+      "entropy": 0.3575092004612088,
+      "epoch": 0.24,
+      "grad_norm": 2.625,
+      "learning_rate": 1.6991150442477876e-05,
+      "loss": 0.3412153959274292,
+      "mean_token_accuracy": 0.8936023108661175,
+      "num_tokens": 280610.0,
+      "step": 30
+    },
+    {
+      "entropy": 0.31294157058000566,
+      "epoch": 0.32,
+      "grad_norm": 2.421875,
+      "learning_rate": 1.5221238938053098e-05,
+      "loss": 0.3175334453582764,
+      "mean_token_accuracy": 0.8971926532685757,
+      "num_tokens": 379577.0,
+      "step": 40
+    },
+    {
+      "entropy": 0.30759106744080783,
+      "epoch": 0.4,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1.345132743362832e-05,
+      "loss": 0.32603118419647215,
+      "mean_token_accuracy": 0.8967837899923324,
+      "num_tokens": 477577.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.31574366632848977,
+      "epoch": 0.48,
+      "grad_norm": 1.9921875,
+      "learning_rate": 1.1681415929203541e-05,
+      "loss": 0.323792552947998,
+      "mean_token_accuracy": 0.8988608829677105,
+      "num_tokens": 573360.0,
+      "step": 60
+    },
+    {
+      "entropy": 0.3065364420413971,
+      "epoch": 0.56,
+      "grad_norm": 1.6171875,
+      "learning_rate": 9.911504424778762e-06,
+      "loss": 0.32266840934753416,
+      "mean_token_accuracy": 0.8985385209321975,
+      "num_tokens": 665761.0,
+      "step": 70
+    },
+    {
+      "entropy": 0.27206654362380506,
+      "epoch": 0.64,
+      "grad_norm": 2.1875,
+      "learning_rate": 8.141592920353984e-06,
+      "loss": 0.2854612350463867,
+      "mean_token_accuracy": 0.9085792370140553,
+      "num_tokens": 756921.0,
+      "step": 80
+    },
+    {
+      "entropy": 0.28962985165417193,
+      "epoch": 0.72,
+      "grad_norm": 1.8125,
+      "learning_rate": 6.371681415929204e-06,
+      "loss": 0.29041600227355957,
+      "mean_token_accuracy": 0.9059965394437313,
+      "num_tokens": 845171.0,
+      "step": 90
+    },
+    {
+      "entropy": 0.2753044681623578,
+      "epoch": 0.8,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.6017699115044254e-06,
+      "loss": 0.2797576427459717,
+      "mean_token_accuracy": 0.9101039692759514,
+      "num_tokens": 939550.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.2808386910706758,
+      "epoch": 0.88,
+      "grad_norm": 1.5859375,
+      "learning_rate": 2.831858407079646e-06,
+      "loss": 0.2893812894821167,
+      "mean_token_accuracy": 0.9063553236424923,
+      "num_tokens": 1041441.0,
+      "step": 110
+    },
+    {
+      "entropy": 0.2944059643894434,
+      "epoch": 0.96,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.0619469026548673e-06,
+      "loss": 0.2996867895126343,
+      "mean_token_accuracy": 0.9011975660920143,
+      "num_tokens": 1132794.0,
+      "step": 120
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 125,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2589656731283456e+16,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-125/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4e0921a533d54992c732b7cf0c8a29ef97e50bbd333d4e4c07ae7183a1feb0f
+size 5713

config.json

@@ -56,7 +56,7 @@
   },
   "sliding_window": null,
   "tie_word_embeddings": true,
-  "transformers_version": "5.8.0",
+  "transformers_version": "5.7.0",
   "use_cache": false,
   "use_sliding_window": false,
   "vocab_size": 151936

generation_config.json

@@ -6,9 +6,8 @@
     151643
   ],
   "pad_token_id": 151643,
-  "temperature": 0.7,
-  "top_p": 0.95,
+  "temperature": 0.6,
   "top_k": 20,
-  "repetition_penalty": 1.03,
+  "top_p": 0.95,
   "transformers_version": "4.51.0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c9179fe220727db304cb809795bd582e6dbc90a763ef3579708273b8b09858e5
+oid sha256:7074b51b321535051b77436f30720111854403259c583a273488e061c0d23072
 size 3441185608

tokenizer_config.json

@@ -27,5 +27,5 @@
   "split_special_tokens": false,
   "tokenizer_class": "Qwen2Tokenizer",
   "unk_token": null,
-  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {#- BOXING INSTRUCTION injected so it is present even when the caller (e.g. CI)\n        passes no system message. Targets our no-box / multi-box failure mode. #}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- 'When solving a math problem, work through it step by step and write your final answer inside a \\boxed{...}.<|im_end|>\\n' }}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if message.content is string %}\n        {%- set content = message.content %}\n    {%- else %}\n        {%- set content = '' %}\n    {%- endif %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}"
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if message.content is string %}\n        {%- set content = message.content %}\n    {%- else %}\n        {%- set content = '' %}\n    {%- endif %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}"
 }

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4e0921a533d54992c732b7cf0c8a29ef97e50bbd333d4e4c07ae7183a1feb0f
+size 5713