Upload 74 files

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/.gitattributes

@@ -0,0 +1,37 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+onnx/model.onnx_data filter=lfs diff=lfs merge=lfs -text
+onnx/model_fp16.onnx_data filter=lfs diff=lfs merge=lfs -text

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/README.md

@@ -0,0 +1,305 @@
+---
+library_name: transformers
+license: apache-2.0
+language:
+- en
+pipeline_tag: text-generation
+tags:
+- safetensors
+- onnx
+- transformers.js
+base_model:
+- HuggingFaceTB/SmolLM2-1.7B
+---
+
+
+# SmolLM2
+
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/y45hIMNREW7w_XpHYB_0q.png)
+
+##  Table of Contents
+
+1. [Model Summary](#model-summary)
+2. [Evaluation](#evaluation)
+3. [Examples](#examples)
+4. [Limitations](#limitations)
+5. [Training](#training)
+6. [License](#license)
+7. [Citation](#citation)
+
+## Model Summary
+
+SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters. They are capable of solving a wide range of tasks while being lightweight enough to run on-device.
+
+The 1.7B variant demonstrates significant advances over its predecessor SmolLM1-1.7B, particularly in instruction following, knowledge, reasoning, and mathematics. It was trained on 11 trillion tokens using a diverse dataset combination: FineWeb-Edu, DCLM, The Stack, along with new mathematics and coding datasets that we curated and will release soon. We developed the instruct version through supervised fine-tuning (SFT) using a combination of public datasets and our own curated datasets. We then applied Direct Preference Optimization (DPO) using [UltraFeedback](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized).
+
+The instruct model additionally supports tasks such as text rewriting, summarization and function calling thanks to datasets developed by [Argilla](https://huggingface.co/argilla) such as [Synth-APIGen-v0.1](https://huggingface.co/datasets/argilla/Synth-APIGen-v0.1).
+You can find the SFT dataset here: https://huggingface.co/datasets/HuggingFaceTB/smoltalk.
+
+For more details refer to: https://github.com/huggingface/smollm. You will find pre-training, post-training, evaluation and local inference code.
+
+### How to use
+
+### Transformers
+```bash
+pip install transformers
+```
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+checkpoint = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
+
+device = "cuda" # for GPU usage or "cpu" for CPU usage
+tokenizer = AutoTokenizer.from_pretrained(checkpoint)
+# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
+model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
+
+messages = [{"role": "user", "content": "What is the capital of France."}]
+input_text=tokenizer.apply_chat_template(messages, tokenize=False)
+inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
+print(tokenizer.decode(outputs[0]))
+```
+
+
+### Chat in TRL
+You can also use the TRL CLI to chat with the model from the terminal:
+```bash
+pip install trl
+trl chat --model_name_or_path HuggingFaceTB/SmolLM2-1.7B-Instruct --device cpu
+```
+
+## Evaluation
+
+In this section, we report the evaluation results of SmolLM2. All evaluations are zero-shot unless stated otherwise, and we use [lighteval](https://github.com/huggingface/lighteval) to run them.
+
+## Base Pre-Trained Model
+
+| Metric           | SmolLM2-1.7B | Llama-1B    | Qwen2.5-1.5B | SmolLM1-1.7B |
+|------------------|--------------|-------------|---------------|--------------|
+| HellaSwag        | **68.7**     | 61.2        | 66.4          | 62.9         |
+| ARC (Average)    | **60.5**     | 49.2        | 58.5          | 59.9         |
+| PIQA             | **77.6**     | 74.8        | 76.1          | 76.0         |
+| MMLU-Pro (MCF)   | **19.4**     | 11.7        | 13.7          | 10.8         |
+| CommonsenseQA    | **43.6**     | 41.2        | 34.1          | 38.0         |
+| TriviaQA         | **36.7**     | 28.1        | 20.9          | 22.5         |
+| Winogrande       | **59.4**     | 57.8        | 59.3          | 54.7         |
+| OpenBookQA       | 42.2         | 38.4        | 40.0          | **42.4**     |
+| GSM8K (5-shot)   | 31.0         | 7.2         | **61.3**      | 5.5          |
+
+## Instruction Model
+
+| Metric                       | SmolLM2-1.7B-Instruct | Llama-1B-Instruct | Qwen2.5-1.5B-Instruct | SmolLM1-1.7B-Instruct |
+|:-----------------------------|:---------------------:|:-----------------:|:----------------------:|:----------------------:|
+| IFEval (Average prompt/inst) | **56.7**             | 53.5             | 47.4                  | 23.1                  |
+| MT-Bench                     | 6.13                | 5.48             | **6.52**              | 4.33                  |
+| OpenRewrite-Eval (micro_avg RougeL) | 44.9           | 39.2             | **46.9**              | NaN                   |
+| HellaSwag                    | **66.1**            | 56.1             | 60.9                  | 55.5                  |
+| ARC (Average)                | **51.7**            | 41.6             | 46.2                  | 43.7                  |
+| PIQA                         | **74.4**            | 72.3             | 73.2                  | 71.6                  |
+| MMLU-Pro (MCF)               | 19.3               | 12.7             | **24.2**              | 11.7                  |
+| BBH (3-shot)                 | 32.2               | 27.6             | **35.3**              | 25.7                  |
+| GSM8K (5-shot)               | **48.2**           | 26.8             | 42.8                  | 4.62                  |
+
+
+## Examples
+Below are some system and instruct prompts that work well for special tasks
+
+### Text rewriting
+
+```python
+system_prompt_rewrite = "You are an AI writing assistant. Your task is to rewrite the user's email to make it more professional and approachable while maintaining its main points and key message. Do not return any text other than the rewritten message."
+user_prompt_rewrite = "Rewrite the message below to make it more friendly and approachable while maintaining its main points and key message. Do not add any new information or return any text other than the rewritten message\nThe message:"
+messages = [{"role": "system", "content": system_prompt_rewrite}, {"role": "user", "content":f"{user_prompt_rewrite} The CI is failing after your last commit!"}]
+input_text=tokenizer.apply_chat_template(messages, tokenize=False)
+inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
+print(tokenizer.decode(outputs[0]))
+```
+```
+Hey there! I noticed that the CI isn't passing after your latest commit. Could you take a look and let me know what's going on? Thanks so much for your help!
+```
+
+### Summarization
+
+```python
+system_prompt_summarize = "Provide a concise, objective summary of the input text in up to three sentences, focusing on key actions and intentions without using second or third person pronouns."
+messages = [{"role": "system", "content": system_prompt_summarize}, {"role": "user", "content": INSERT_LONG_EMAIL}]
+input_text=tokenizer.apply_chat_template(messages, tokenize=False)
+inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
+print(tokenizer.decode(outputs[0]))
+```
+
+### Function calling
+
+SmolLM2-1.7B-Instruct can handle function calling, it scores 27% on the [BFCL Leaderboard](https://gorilla.cs.berkeley.edu/blogs/8_berkeley_function_calling_leaderboard.html). Here's how you can leverage it:
+
+```python
+import json
+import re
+from typing import Optional
+
+from jinja2 import Template
+import torch 
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from transformers.utils import get_json_schema
+
+
+system_prompt = Template("""You are an expert in composing functions. You are given a question and a set of possible functions. 
+Based on the question, you will need to make one or more function/tool calls to achieve the purpose. 
+If none of the functions can be used, point it out and refuse to answer. 
+If the given question lacks the parameters required by the function, also point it out.
+
+You have access to the following tools:
+<tools>{{ tools }}</tools>
+
+The output MUST strictly adhere to the following format, and NO other text MUST be included.
+The example format is as follows. Please make sure the parameter type is correct. If no function call is needed, please make the tool calls an empty list '[]'.
+<tool_call>[
+{"name": "func_name1", "arguments": {"argument1": "value1", "argument2": "value2"}},
+... (more tool calls as required)
+]</tool_call>""")
+
+
+def prepare_messages(
+    query: str,
+    tools: Optional[dict[str, any]] = None,
+    history: Optional[list[dict[str, str]]] = None
+) -> list[dict[str, str]]:
+    """Prepare the system and user messages for the given query and tools.
+    
+    Args:
+        query: The query to be answered.
+        tools: The tools available to the user. Defaults to None, in which case if a
+            list without content will be passed to the model.
+        history: Exchange of messages, including the system_prompt from
+            the first query. Defaults to None, the first message in a conversation.
+    """
+    if tools is None:
+        tools = []
+    if history:
+        messages = history.copy()
+        messages.append({"role": "user", "content": query})
+    else:
+        messages = [
+            {"role": "system", "content": system_prompt.render(tools=json.dumps(tools))},
+            {"role": "user", "content": query}
+        ]
+    return messages
+
+
+def parse_response(text: str) -> str | dict[str, any]:
+    """Parses a response from the model, returning either the
+    parsed list with the tool calls parsed, or the
+    model thought or response if couldn't generate one.
+
+    Args:
+        text: Response from the model.
+    """
+    pattern = r"<tool_call>(.*?)</tool_call>"
+    matches = re.findall(pattern, text, re.DOTALL)
+    if matches:
+        return json.loads(matches[0])
+    return text
+
+
+model_name_smollm = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
+model = AutoModelForCausalLM.from_pretrained(model_name_smollm, device_map="auto", torch_dtype="auto", trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained(model_name_smollm)
+
+from datetime import datetime
+import random
+
+def get_current_time() -> str:
+    """Returns the current time in 24-hour format.
+
+    Returns:
+        str: Current time in HH:MM:SS format.
+    """
+    return datetime.now().strftime("%H:%M:%S")
+
+
+def get_random_number_between(min: int, max: int) -> int:
+    """
+    Gets a random number between min and max.
+
+    Args:
+        min: The minimum number.
+        max: The maximum number.
+
+    Returns:
+        A random number between min and max.
+    """
+    return random.randint(min, max)
+
+
+tools = [get_json_schema(get_random_number_between), get_json_schema(get_current_time)]
+
+toolbox = {"get_random_number_between": get_random_number_between, "get_current_time": get_current_time}
+
+query = "Give me a number between 1 and 300"
+
+messages = prepare_messages(query, tools=tools)
+
+inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
+result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+
+tool_calls = parse_response(result)
+# [{'name': 'get_random_number_between', 'arguments': {'min': 1, 'max': 300}}
+
+# Get tool responses
+tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
+# [63]
+
+# For the second turn, rebuild the history of messages:
+history = messages.copy()
+# Add the "parsed response"
+history.append({"role": "assistant", "content": result})
+query = "Can you give me the hour?"
+history.append({"role": "user", "content": query})
+
+inputs = tokenizer.apply_chat_template(history, add_generation_prompt=True, return_tensors="pt").to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
+result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+
+tool_calls = parse_response(result)
+tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
+# ['07:57:25']
+```
+More details such as parallel function calls and tools not available can be found [here](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct/blob/main/instructions_function_calling.md)
+
+## Limitations
+
+SmolLM2 models primarily understand and generate content in English. They can produce text on a variety of topics, but the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data. These models should be used as assistive tools rather than definitive sources of information. Users should always verify important information and critically evaluate any generated content.
+
+## Training
+
+### Model
+
+- **Architecture:** Transformer decoder
+- **Pretraining tokens:** 11T
+- **Precision:** bfloat16
+
+### Hardware
+
+- **GPUs:** 256 H100
+
+### Software
+
+- **Training Framework:** [nanotron](https://github.com/huggingface/nanotron/tree/main)
+- **Alignment Handbook** [alignment-handbook](https://github.com/huggingface/alignment-handbook/)
+
+## License
+
+[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
+
+## Citation
+```bash
+@misc{allal2024SmolLM2,
+      title={SmolLM2 - with great data, comes great performance}, 
+      author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Gabriel Martín Blázquez and Lewis Tunstall and Agustín Piqueres and Andres Marafioti and Cyril Zakka and Leandro von Werra and Thomas Wolf},
+      year={2024},
+}
+```

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/all_results.json

@@ -0,0 +1,22 @@
+{
+    "epoch": 2.996074326092646,
+    "eval_logits/chosen": -0.34099623560905457,
+    "eval_logits/rejected": -0.3685227334499359,
+    "eval_logps/chosen": -310.2510070800781,
+    "eval_logps/rejected": -275.43145751953125,
+    "eval_loss": 0.587827205657959,
+    "eval_rewards/accuracies": 0.6746031641960144,
+    "eval_rewards/chosen": 0.01673175022006035,
+    "eval_rewards/margins": 0.5906793475151062,
+    "eval_rewards/rejected": -0.573947548866272,
+    "eval_runtime": 18.8462,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 106.122,
+    "eval_steps_per_second": 3.343,
+    "total_flos": 0.0,
+    "train_loss": 0.5334697115221363,
+    "train_runtime": 7355.3343,
+    "train_samples": 61134,
+    "train_samples_per_second": 24.935,
+    "train_steps_per_second": 0.195
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/config.json

@@ -0,0 +1,35 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 8192,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 24,
+  "num_key_value_heads": 32,
+  "pad_token_id": 2,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 130000,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.42.3",
+  "transformers.js_config": {
+    "kv_cache_dtype": {
+      "q4f16": "float16",
+      "fp16": "float16"
+    }
+  },
+  "use_cache": true,
+  "vocab_size": 49152
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 2.996074326092646,
+    "eval_logits/chosen": -0.34099623560905457,
+    "eval_logits/rejected": -0.3685227334499359,
+    "eval_logps/chosen": -310.2510070800781,
+    "eval_logps/rejected": -275.43145751953125,
+    "eval_loss": 0.587827205657959,
+    "eval_rewards/accuracies": 0.6746031641960144,
+    "eval_rewards/chosen": 0.01673175022006035,
+    "eval_rewards/margins": 0.5906793475151062,
+    "eval_rewards/rejected": -0.573947548866272,
+    "eval_runtime": 18.8462,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 106.122,
+    "eval_steps_per_second": 3.343
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 2,
+  "transformers_version": "4.42.3"
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/instructions_function_calling.md

@@ -0,0 +1,183 @@
+## Quick start
+Instructions for funtion calling: 
+
+```python
+import json
+import re
+from typing import Optional
+
+from jinja2 import Template
+import torch 
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from transformers.utils import get_json_schema
+
+
+system_prompt = Template("""You are an expert in composing functions. You are given a question and a set of possible functions. 
+Based on the question, you will need to make one or more function/tool calls to achieve the purpose. 
+If none of the functions can be used, point it out and refuse to answer. 
+If the given question lacks the parameters required by the function, also point it out.
+
+You have access to the following tools:
+<tools>{{ tools }}</tools>
+
+The output MUST strictly adhere to the following format, and NO other text MUST be included.
+The example format is as follows. Please make sure the parameter type is correct. If no function call is needed, please make the tool calls an empty list '[]'.
+<tool_call>[
+{"name": "func_name1", "arguments": {"argument1": "value1", "argument2": "value2"}},
+... (more tool calls as required)
+]</tool_call>""")
+
+
+def prepare_messages(
+    query: str,
+    tools: Optional[dict[str, any]] = None,
+    history: Optional[list[dict[str, str]]] = None
+) -> list[dict[str, str]]:
+    """Prepare the system and user messages for the given query and tools.
+    
+    Args:
+        query: The query to be answered.
+        tools: The tools available to the user. Defaults to None, in which case if a
+            list without content will be passed to the model.
+        history: Exchange of messages, including the system_prompt from
+            the first query. Defaults to None, the first message in a conversation.
+    """
+    if tools is None:
+        tools = []
+    if history:
+        messages = history.copy()
+        messages.append({"role": "user", "content": query})
+    else:
+        messages = [
+            {"role": "system", "content": system_prompt.render(tools=json.dumps(tools))},
+            {"role": "user", "content": query}
+        ]
+    return messages
+
+
+def parse_response(text: str) -> str | dict[str, any]:
+    """Parses a response from the model, returning either the
+    parsed list with the tool calls parsed, or the
+    model thought or response if couldn't generate one.
+
+    Args:
+        text: Response from the model.
+    """
+    pattern = r"<tool_call>(.*?)</tool_call>"
+    matches = re.findall(pattern, text, re.DOTALL)
+    if matches:
+        return json.loads(matches[0])
+    return text
+
+model_name_smollm = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
+model = AutoModelForCausalLM.from_pretrained(model_name_smollm, device_map="auto", torch_dtype="auto", trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained(model_name_smollm)
+
+from datetime import datetime
+import random
+
+def get_current_time() -> str:
+    """Returns the current time in 24-hour format.
+
+    Returns:
+        str: Current time in HH:MM:SS format.
+    """
+    return datetime.now().strftime("%H:%M:%S")
+
+
+def get_random_number_between(min: int, max: int) -> int:
+    """
+    Gets a random number between min and max.
+
+    Args:
+        min: The minimum number.
+        max: The maximum number.
+
+    Returns:
+        A random number between min and max.
+    """
+    return random.randint(min, max)
+
+
+tools = [get_json_schema(get_random_number_between), get_json_schema(get_current_time)]
+
+toolbox = {"get_random_number_between": get_random_number_between, "get_current_time": get_current_time}
+
+query = "Give me a number between 1 and 300"
+
+messages = prepare_messages(query, tools=tools)
+
+inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
+result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+
+tool_calls = parse_response(result)
+# [{'name': 'get_random_number_between', 'arguments': {'min': 1, 'max': 300}}
+
+# Get tool responses
+tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
+# [63]
+
+# For the second turn, rebuild the history of messages:
+history = messages.copy()
+# Add the "parsed response"
+history.append({"role": "assistant", "content": result})
+query = "Can you give me the hour?"
+history.append({"role": "user", "content": query})
+
+inputs = tokenizer.apply_chat_template(history, add_generation_prompt=True, return_tensors="pt").to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
+result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+
+tool_calls = parse_response(result)
+tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
+# ['07:57:25']
+```
+
+#### Parallel function calls
+
+Multiple calls required by the same query.
+
+```python
+query = "Can you give me the hour and a random number between 1 and 50?"
+
+messages = prepare_messages(query, tools=tools)
+
+inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
+result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+
+tool_calls = parse_response(result)
+tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
+# ['09:24:52', 50]
+
+query = "Can you give me a random number between 1 and 10, other between 200 and 210 and another one between 55 and 60?"
+
+messages = prepare_messages(query, tools=tools)
+
+inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
+result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+
+tool_calls = parse_response(result)
+tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
+# [7, 202, 60]
+```
+
+#### Tools not available
+
+```python
+query = "Can you open a new page with youtube?"
+
+messages = prepare_messages(query, tools=tools)
+
+inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
+result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+
+tool_calls = parse_response(result)
+# []
+
+# The response will be something similar to the following:
+# "The query cannot be answered with the provided tools. Please make sure the tools are correctly installed and imported. If the tools are not installed, install them using pip: 'pip install -r tools.txt'. If the tools are already installed, ensure they are correctly configured. If the tools are not correctly configured, please contact the support team. The output MUST strictly adhere to the following format, and NO other text MUST be included.\n\n<tool_call>[]</tool_call>"
+```

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f55217be716b6a997b97b9d8d7eb6fad02e00858f5010ec24f64603c3a98a0e8
+size 3422777952

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/runs/Oct31_06-24-59_ip-26-0-174-36/events.out.tfevents.1730356365.ip-26-0-174-36.3169719.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6bfce1916438dd2e6553aa0a62d418087b3ae04f8af75e714ad1f01b7663db6
+size 114828

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/runs/Oct31_06-24-59_ip-26-0-174-36/events.out.tfevents.1730363825.ip-26-0-174-36.3169719.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3d7723fd0715ce6dcbccf7bb2097f59490b0ac670f798f5378ef5abb7d1301d
+size 828

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/special_tokens_map.json

@@ -0,0 +1,34 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": {
+    "content": "<|im_start|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/tokenizer_config.json

@@ -0,0 +1,154 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<repo_name>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<file_sep>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<jupyter_script>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": "<|im_start|>",
+  "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful AI assistant named SmolLM, trained by Hugging Face<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "model_max_length": 2048,
+  "pad_token": "<|im_end|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 2.996074326092646,
+    "total_flos": 0.0,
+    "train_loss": 0.5334697115221363,
+    "train_runtime": 7355.3343,
+    "train_samples": 61134,
+    "train_samples_per_second": 24.935,
+    "train_steps_per_second": 0.195
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/trainer_state.json

@@ -0,0 +1,2426 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.996074326092646,
+  "eval_steps": 100,
+  "global_step": 1431,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.002093692750588851,
+      "grad_norm": 53.581159797745435,
+      "learning_rate": 6.9444444444444435e-09,
+      "logits/chosen": -0.48425233364105225,
+      "logits/rejected": -0.32109448313713074,
+      "logps/chosen": -276.5158996582031,
+      "logps/rejected": -302.22406005859375,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.02093692750588851,
+      "grad_norm": 47.92244929429545,
+      "learning_rate": 6.944444444444444e-08,
+      "logits/chosen": -0.20008452236652374,
+      "logits/rejected": -0.1900922954082489,
+      "logps/chosen": -337.452392578125,
+      "logps/rejected": -293.0785217285156,
+      "loss": 0.7231,
+      "rewards/accuracies": 0.3541666567325592,
+      "rewards/chosen": -0.022724969312548637,
+      "rewards/margins": -0.036751341074705124,
+      "rewards/rejected": 0.014026367105543613,
+      "step": 10
+    },
+    {
+      "epoch": 0.04187385501177702,
+      "grad_norm": 41.37562973132045,
+      "learning_rate": 1.3888888888888888e-07,
+      "logits/chosen": -0.042457859963178635,
+      "logits/rejected": -0.12332990020513535,
+      "logps/chosen": -298.8910217285156,
+      "logps/rejected": -271.74114990234375,
+      "loss": 0.7114,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": 0.016111990436911583,
+      "rewards/margins": -0.0034494102001190186,
+      "rewards/rejected": 0.0195614043623209,
+      "step": 20
+    },
+    {
+      "epoch": 0.06281078251766553,
+      "grad_norm": 50.166233912169,
+      "learning_rate": 2.0833333333333333e-07,
+      "logits/chosen": -0.0645759329199791,
+      "logits/rejected": -0.09565907716751099,
+      "logps/chosen": -323.93743896484375,
+      "logps/rejected": -261.7337341308594,
+      "loss": 0.7165,
+      "rewards/accuracies": 0.4625000059604645,
+      "rewards/chosen": -0.056218355894088745,
+      "rewards/margins": 0.005095779895782471,
+      "rewards/rejected": -0.061314135789871216,
+      "step": 30
+    },
+    {
+      "epoch": 0.08374771002355404,
+      "grad_norm": 45.45611748658909,
+      "learning_rate": 2.7777777777777776e-07,
+      "logits/chosen": -0.13370418548583984,
+      "logits/rejected": -0.16684015095233917,
+      "logps/chosen": -272.07928466796875,
+      "logps/rejected": -251.01589965820312,
+      "loss": 0.7166,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.006498994771391153,
+      "rewards/margins": 0.06555742770433426,
+      "rewards/rejected": -0.05905843526124954,
+      "step": 40
+    },
+    {
+      "epoch": 0.10468463752944256,
+      "grad_norm": 55.30410487113166,
+      "learning_rate": 3.472222222222222e-07,
+      "logits/chosen": -0.0792478546500206,
+      "logits/rejected": -0.0551227442920208,
+      "logps/chosen": -292.2696838378906,
+      "logps/rejected": -266.82965087890625,
+      "loss": 0.7243,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.08804600685834885,
+      "rewards/margins": -0.04349173232913017,
+      "rewards/rejected": -0.044554274529218674,
+      "step": 50
+    },
+    {
+      "epoch": 0.12562156503533106,
+      "grad_norm": 56.63523816899612,
+      "learning_rate": 4.1666666666666667e-07,
+      "logits/chosen": -0.1644289791584015,
+      "logits/rejected": -0.19546563923358917,
+      "logps/chosen": -269.5990295410156,
+      "logps/rejected": -258.08721923828125,
+      "loss": 0.7419,
+      "rewards/accuracies": 0.518750011920929,
+      "rewards/chosen": 0.04003779590129852,
+      "rewards/margins": 0.05106702446937561,
+      "rewards/rejected": -0.011029230430722237,
+      "step": 60
+    },
+    {
+      "epoch": 0.14655849254121958,
+      "grad_norm": 50.48428456914726,
+      "learning_rate": 4.861111111111111e-07,
+      "logits/chosen": -0.1528700441122055,
+      "logits/rejected": -0.12956030666828156,
+      "logps/chosen": -334.1960144042969,
+      "logps/rejected": -295.05865478515625,
+      "loss": 0.7096,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -0.012429716996848583,
+      "rewards/margins": 0.03767075017094612,
+      "rewards/rejected": -0.05010046809911728,
+      "step": 70
+    },
+    {
+      "epoch": 0.16749542004710807,
+      "grad_norm": 49.10207941889299,
+      "learning_rate": 5.555555555555555e-07,
+      "logits/chosen": 0.019698064774274826,
+      "logits/rejected": -0.05375131219625473,
+      "logps/chosen": -324.42889404296875,
+      "logps/rejected": -256.3900451660156,
+      "loss": 0.7125,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.0010342865716665983,
+      "rewards/margins": 0.04288307949900627,
+      "rewards/rejected": -0.0418488010764122,
+      "step": 80
+    },
+    {
+      "epoch": 0.1884323475529966,
+      "grad_norm": 48.35701594054669,
+      "learning_rate": 6.249999999999999e-07,
+      "logits/chosen": -0.18192127346992493,
+      "logits/rejected": -0.15880808234214783,
+      "logps/chosen": -282.22161865234375,
+      "logps/rejected": -251.54296875,
+      "loss": 0.6976,
+      "rewards/accuracies": 0.4937500059604645,
+      "rewards/chosen": -0.012317812070250511,
+      "rewards/margins": -0.011760599911212921,
+      "rewards/rejected": -0.0005572110530920327,
+      "step": 90
+    },
+    {
+      "epoch": 0.2093692750588851,
+      "grad_norm": 44.92592146146067,
+      "learning_rate": 6.944444444444444e-07,
+      "logits/chosen": -0.11696960031986237,
+      "logits/rejected": -0.13915769755840302,
+      "logps/chosen": -316.96527099609375,
+      "logps/rejected": -277.3512268066406,
+      "loss": 0.6787,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": 0.03938128799200058,
+      "rewards/margins": 0.08465500175952911,
+      "rewards/rejected": -0.045273713767528534,
+      "step": 100
+    },
+    {
+      "epoch": 0.2093692750588851,
+      "eval_logits/chosen": -0.3141140341758728,
+      "eval_logits/rejected": -0.3377441465854645,
+      "eval_logps/chosen": -310.252685546875,
+      "eval_logps/rejected": -274.4240417480469,
+      "eval_loss": 0.6966869235038757,
+      "eval_rewards/accuracies": 0.5515872836112976,
+      "eval_rewards/chosen": 0.01587979681789875,
+      "eval_rewards/margins": 0.08612197637557983,
+      "eval_rewards/rejected": -0.07024218887090683,
+      "eval_runtime": 19.212,
+      "eval_samples_per_second": 104.102,
+      "eval_steps_per_second": 3.279,
+      "step": 100
+    },
+    {
+      "epoch": 0.23030620256477363,
+      "grad_norm": 42.34774239140926,
+      "learning_rate": 7.638888888888888e-07,
+      "logits/chosen": -0.15527260303497314,
+      "logits/rejected": -0.1998877376317978,
+      "logps/chosen": -317.2166748046875,
+      "logps/rejected": -267.8033447265625,
+      "loss": 0.6869,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": -0.0031774670351296663,
+      "rewards/margins": 0.06427686661481857,
+      "rewards/rejected": -0.06745433807373047,
+      "step": 110
+    },
+    {
+      "epoch": 0.2512431300706621,
+      "grad_norm": 42.69543501672853,
+      "learning_rate": 8.333333333333333e-07,
+      "logits/chosen": -0.08257915079593658,
+      "logits/rejected": -0.12443940341472626,
+      "logps/chosen": -275.12579345703125,
+      "logps/rejected": -245.96347045898438,
+      "loss": 0.6706,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": 0.050213612616062164,
+      "rewards/margins": 0.10280221700668335,
+      "rewards/rejected": -0.052588604390621185,
+      "step": 120
+    },
+    {
+      "epoch": 0.2721800575765506,
+      "grad_norm": 50.775507526530916,
+      "learning_rate": 9.027777777777778e-07,
+      "logits/chosen": -0.14645054936408997,
+      "logits/rejected": -0.2005746066570282,
+      "logps/chosen": -313.00238037109375,
+      "logps/rejected": -257.37945556640625,
+      "loss": 0.6662,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": 0.12051638215780258,
+      "rewards/margins": 0.18545812368392944,
+      "rewards/rejected": -0.06494174152612686,
+      "step": 130
+    },
+    {
+      "epoch": 0.29311698508243916,
+      "grad_norm": 45.04510687958917,
+      "learning_rate": 9.722222222222222e-07,
+      "logits/chosen": -0.18603107333183289,
+      "logits/rejected": -0.14804694056510925,
+      "logps/chosen": -336.45330810546875,
+      "logps/rejected": -270.71844482421875,
+      "loss": 0.6837,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.1061311587691307,
+      "rewards/margins": 0.20069436728954315,
+      "rewards/rejected": -0.09456320852041245,
+      "step": 140
+    },
+    {
+      "epoch": 0.31405391258832765,
+      "grad_norm": 47.07473020466515,
+      "learning_rate": 9.999463737538052e-07,
+      "logits/chosen": -0.09479381889104843,
+      "logits/rejected": -0.18823939561843872,
+      "logps/chosen": -300.7477111816406,
+      "logps/rejected": -261.5872497558594,
+      "loss": 0.6701,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": 0.014584893360733986,
+      "rewards/margins": 0.1319422572851181,
+      "rewards/rejected": -0.11735737323760986,
+      "step": 150
+    },
+    {
+      "epoch": 0.33499084009421615,
+      "grad_norm": 40.24946321953983,
+      "learning_rate": 9.996186994612174e-07,
+      "logits/chosen": -0.07166972011327744,
+      "logits/rejected": -0.05619664117693901,
+      "logps/chosen": -277.3063049316406,
+      "logps/rejected": -264.21929931640625,
+      "loss": 0.6571,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.002465812023729086,
+      "rewards/margins": 0.21548476815223694,
+      "rewards/rejected": -0.21795055270195007,
+      "step": 160
+    },
+    {
+      "epoch": 0.3559277676001047,
+      "grad_norm": 43.25422484420196,
+      "learning_rate": 9.989933382359422e-07,
+      "logits/chosen": -0.22567839920520782,
+      "logits/rejected": -0.2581842541694641,
+      "logps/chosen": -270.6056213378906,
+      "logps/rejected": -234.0198211669922,
+      "loss": 0.6614,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.0034525603987276554,
+      "rewards/margins": 0.24006681144237518,
+      "rewards/rejected": -0.2435193508863449,
+      "step": 170
+    },
+    {
+      "epoch": 0.3768646951059932,
+      "grad_norm": 39.47362541006732,
+      "learning_rate": 9.980706626858607e-07,
+      "logits/chosen": -0.15520626306533813,
+      "logits/rejected": -0.13547591865062714,
+      "logps/chosen": -275.5711364746094,
+      "logps/rejected": -280.95904541015625,
+      "loss": 0.6323,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.01858050376176834,
+      "rewards/margins": 0.3474423885345459,
+      "rewards/rejected": -0.32886189222335815,
+      "step": 180
+    },
+    {
+      "epoch": 0.39780162261188173,
+      "grad_norm": 45.3164057858113,
+      "learning_rate": 9.968512225671258e-07,
+      "logits/chosen": -0.07492430508136749,
+      "logits/rejected": -0.07894851267337799,
+      "logps/chosen": -287.3063659667969,
+      "logps/rejected": -286.6043395996094,
+      "loss": 0.649,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.07111965119838715,
+      "rewards/margins": 0.22652164101600647,
+      "rewards/rejected": -0.2976413071155548,
+      "step": 190
+    },
+    {
+      "epoch": 0.4187385501177702,
+      "grad_norm": 51.9773301288437,
+      "learning_rate": 9.953357444566038e-07,
+      "logits/chosen": -0.09818016737699509,
+      "logits/rejected": -0.11751595884561539,
+      "logps/chosen": -284.3682861328125,
+      "logps/rejected": -264.2423095703125,
+      "loss": 0.645,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": -0.06741360574960709,
+      "rewards/margins": 0.13478204607963562,
+      "rewards/rejected": -0.2021956443786621,
+      "step": 200
+    },
+    {
+      "epoch": 0.4187385501177702,
+      "eval_logits/chosen": -0.3229367136955261,
+      "eval_logits/rejected": -0.34633249044418335,
+      "eval_logps/chosen": -310.3840026855469,
+      "eval_logps/rejected": -274.8875732421875,
+      "eval_loss": 0.6491106748580933,
+      "eval_rewards/accuracies": 0.60317462682724,
+      "eval_rewards/chosen": -0.04975215345621109,
+      "eval_rewards/margins": 0.25227656960487366,
+      "eval_rewards/rejected": -0.30202868580818176,
+      "eval_runtime": 19.4848,
+      "eval_samples_per_second": 102.644,
+      "eval_steps_per_second": 3.233,
+      "step": 200
+    },
+    {
+      "epoch": 0.4396754776236587,
+      "grad_norm": 42.51852593559685,
+      "learning_rate": 9.935251313189563e-07,
+      "logits/chosen": 0.02242279052734375,
+      "logits/rejected": -0.06930123269557953,
+      "logps/chosen": -321.75689697265625,
+      "logps/rejected": -269.69781494140625,
+      "loss": 0.6528,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.06879940629005432,
+      "rewards/margins": 0.2159956693649292,
+      "rewards/rejected": -0.2847950756549835,
+      "step": 210
+    },
+    {
+      "epoch": 0.46061240512954726,
+      "grad_norm": 39.56904653144407,
+      "learning_rate": 9.914204619686312e-07,
+      "logits/chosen": -0.21375274658203125,
+      "logits/rejected": -0.06523901224136353,
+      "logps/chosen": -263.26251220703125,
+      "logps/rejected": -254.17361450195312,
+      "loss": 0.6473,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -0.09464406222105026,
+      "rewards/margins": 0.11034605652093887,
+      "rewards/rejected": -0.20499010384082794,
+      "step": 220
+    },
+    {
+      "epoch": 0.48154933263543576,
+      "grad_norm": 49.49502162405857,
+      "learning_rate": 9.89022990427073e-07,
+      "logits/chosen": -0.1386515200138092,
+      "logits/rejected": -0.027841920033097267,
+      "logps/chosen": -275.6785583496094,
+      "logps/rejected": -282.2340393066406,
+      "loss": 0.651,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0026134043000638485,
+      "rewards/margins": 0.34593018889427185,
+      "rewards/rejected": -0.34854358434677124,
+      "step": 230
+    },
+    {
+      "epoch": 0.5024862601413242,
+      "grad_norm": 42.40141104653558,
+      "learning_rate": 9.86334145175542e-07,
+      "logits/chosen": -0.16665732860565186,
+      "logits/rejected": -0.13909348845481873,
+      "logps/chosen": -282.2431640625,
+      "logps/rejected": -265.6712341308594,
+      "loss": 0.6269,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": -0.03417225927114487,
+      "rewards/margins": 0.2094908207654953,
+      "rewards/rejected": -0.24366307258605957,
+      "step": 240
+    },
+    {
+      "epoch": 0.5234231876472127,
+      "grad_norm": 44.711963555505086,
+      "learning_rate": 9.83355528303984e-07,
+      "logits/chosen": -0.21623122692108154,
+      "logits/rejected": -0.25620579719543457,
+      "logps/chosen": -321.9479064941406,
+      "logps/rejected": -276.6851501464844,
+      "loss": 0.6483,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -0.033338677138090134,
+      "rewards/margins": 0.21741199493408203,
+      "rewards/rejected": -0.25075066089630127,
+      "step": 250
+    },
+    {
+      "epoch": 0.5443601151531012,
+      "grad_norm": 44.08715279117413,
+      "learning_rate": 9.800889145564616e-07,
+      "logits/chosen": -0.023601394146680832,
+      "logits/rejected": -0.08905109018087387,
+      "logps/chosen": -307.87286376953125,
+      "logps/rejected": -257.384765625,
+      "loss": 0.6489,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.03719106689095497,
+      "rewards/margins": 0.2362823486328125,
+      "rewards/rejected": -0.2734734117984772,
+      "step": 260
+    },
+    {
+      "epoch": 0.5652970426589898,
+      "grad_norm": 39.7549215734126,
+      "learning_rate": 9.765362502737097e-07,
+      "logits/chosen": -0.15661786496639252,
+      "logits/rejected": -0.1753096580505371,
+      "logps/chosen": -313.56787109375,
+      "logps/rejected": -277.8585510253906,
+      "loss": 0.6233,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.07719334214925766,
+      "rewards/margins": 0.3890419602394104,
+      "rewards/rejected": -0.31184864044189453,
+      "step": 270
+    },
+    {
+      "epoch": 0.5862339701648783,
+      "grad_norm": 46.953763281596736,
+      "learning_rate": 9.726996522334514e-07,
+      "logits/chosen": -0.22773197293281555,
+      "logits/rejected": -0.25082847476005554,
+      "logps/chosen": -299.99127197265625,
+      "logps/rejected": -249.6244659423828,
+      "loss": 0.6409,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.044135063886642456,
+      "rewards/margins": 0.2800499200820923,
+      "rewards/rejected": -0.32418501377105713,
+      "step": 280
+    },
+    {
+      "epoch": 0.6071708976707668,
+      "grad_norm": 38.382090654521065,
+      "learning_rate": 9.68581406389163e-07,
+      "logits/chosen": -0.14438043534755707,
+      "logits/rejected": -0.17269106209278107,
+      "logps/chosen": -282.90325927734375,
+      "logps/rejected": -274.38079833984375,
+      "loss": 0.6217,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.009734408929944038,
+      "rewards/margins": 0.3537730574607849,
+      "rewards/rejected": -0.363507479429245,
+      "step": 290
+    },
+    {
+      "epoch": 0.6281078251766553,
+      "grad_norm": 41.7647459812509,
+      "learning_rate": 9.641839665080363e-07,
+      "logits/chosen": -0.1018938273191452,
+      "logits/rejected": -0.08353041112422943,
+      "logps/chosen": -316.49810791015625,
+      "logps/rejected": -277.4931640625,
+      "loss": 0.6161,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.03871753811836243,
+      "rewards/margins": 0.39340126514434814,
+      "rewards/rejected": -0.3546837270259857,
+      "step": 300
+    },
+    {
+      "epoch": 0.6281078251766553,
+      "eval_logits/chosen": -0.33166003227233887,
+      "eval_logits/rejected": -0.3552355170249939,
+      "eval_logps/chosen": -310.4118957519531,
+      "eval_logps/rejected": -275.1272277832031,
+      "eval_loss": 0.6316225528717041,
+      "eval_rewards/accuracies": 0.682539701461792,
+      "eval_rewards/chosen": -0.06370855122804642,
+      "eval_rewards/margins": 0.3581322431564331,
+      "eval_rewards/rejected": -0.4218408465385437,
+      "eval_runtime": 19.0833,
+      "eval_samples_per_second": 104.804,
+      "eval_steps_per_second": 3.301,
+      "step": 300
+    },
+    {
+      "epoch": 0.6490447526825438,
+      "grad_norm": 40.9716290999305,
+      "learning_rate": 9.595099527089568e-07,
+      "logits/chosen": -0.11042879521846771,
+      "logits/rejected": -0.18008281290531158,
+      "logps/chosen": -323.7516174316406,
+      "logps/rejected": -247.0631561279297,
+      "loss": 0.6007,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.10534234344959259,
+      "rewards/margins": 0.5895902514457703,
+      "rewards/rejected": -0.48424798250198364,
+      "step": 310
+    },
+    {
+      "epoch": 0.6699816801884323,
+      "grad_norm": 45.14020951986602,
+      "learning_rate": 9.545621499013618e-07,
+      "logits/chosen": -0.024479269981384277,
+      "logits/rejected": -0.038955364376306534,
+      "logps/chosen": -294.4266662597656,
+      "logps/rejected": -252.0673828125,
+      "loss": 0.5995,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": 0.06134527176618576,
+      "rewards/margins": 0.4816582202911377,
+      "rewards/rejected": -0.4203129708766937,
+      "step": 320
+    },
+    {
+      "epoch": 0.6909186076943209,
+      "grad_norm": 38.030813091938484,
+      "learning_rate": 9.493435061259129e-07,
+      "logits/chosen": -0.012772688642144203,
+      "logits/rejected": -0.030413877218961716,
+      "logps/chosen": -291.36761474609375,
+      "logps/rejected": -265.4248046875,
+      "loss": 0.6232,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": 0.001638062298297882,
+      "rewards/margins": 0.40076717734336853,
+      "rewards/rejected": -0.39912909269332886,
+      "step": 330
+    },
+    {
+      "epoch": 0.7118555352002094,
+      "grad_norm": 45.93570520280886,
+      "learning_rate": 9.438571307979704e-07,
+      "logits/chosen": -0.04617694020271301,
+      "logits/rejected": -0.07658630609512329,
+      "logps/chosen": -296.7699890136719,
+      "logps/rejected": -270.9905090332031,
+      "loss": 0.6096,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.05764853209257126,
+      "rewards/margins": 0.28755250573158264,
+      "rewards/rejected": -0.34520095586776733,
+      "step": 340
+    },
+    {
+      "epoch": 0.7327924627060979,
+      "grad_norm": 44.09128545700975,
+      "learning_rate": 9.381062928549151e-07,
+      "logits/chosen": -0.12188152968883514,
+      "logits/rejected": -0.13376249372959137,
+      "logps/chosen": -282.15179443359375,
+      "logps/rejected": -247.45834350585938,
+      "loss": 0.5964,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.09005177021026611,
+      "rewards/margins": 0.41418519616127014,
+      "rewards/rejected": -0.5042369961738586,
+      "step": 350
+    },
+    {
+      "epoch": 0.7537293902119864,
+      "grad_norm": 40.36784058818516,
+      "learning_rate": 9.320944188084241e-07,
+      "logits/chosen": -0.07130910456180573,
+      "logits/rejected": -0.18685731291770935,
+      "logps/chosen": -340.55487060546875,
+      "logps/rejected": -278.08551025390625,
+      "loss": 0.6256,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.1223798543214798,
+      "rewards/margins": 0.4791669249534607,
+      "rewards/rejected": -0.3567870557308197,
+      "step": 360
+    },
+    {
+      "epoch": 0.7746663177178749,
+      "grad_norm": 37.74697617657793,
+      "learning_rate": 9.258250907028572e-07,
+      "logits/chosen": -0.2023000717163086,
+      "logits/rejected": -0.25707709789276123,
+      "logps/chosen": -306.526611328125,
+      "logps/rejected": -248.40023803710938,
+      "loss": 0.603,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.05354640632867813,
+      "rewards/margins": 0.42885661125183105,
+      "rewards/rejected": -0.3753102421760559,
+      "step": 370
+    },
+    {
+      "epoch": 0.7956032452237635,
+      "grad_norm": 36.818935480552355,
+      "learning_rate": 9.193020439809746e-07,
+      "logits/chosen": -0.25006169080734253,
+      "logits/rejected": -0.27239733934402466,
+      "logps/chosen": -306.9652099609375,
+      "logps/rejected": -274.15850830078125,
+      "loss": 0.6108,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.005802587606012821,
+      "rewards/margins": 0.5873938798904419,
+      "rewards/rejected": -0.5931965112686157,
+      "step": 380
+    },
+    {
+      "epoch": 0.816540172729652,
+      "grad_norm": 40.218709775991016,
+      "learning_rate": 9.125291652582547e-07,
+      "logits/chosen": -0.19798080623149872,
+      "logits/rejected": -0.16009968519210815,
+      "logps/chosen": -308.05975341796875,
+      "logps/rejected": -289.6929626464844,
+      "loss": 0.6163,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.09920313209295273,
+      "rewards/margins": 0.47261109948158264,
+      "rewards/rejected": -0.3734079897403717,
+      "step": 390
+    },
+    {
+      "epoch": 0.8374771002355405,
+      "grad_norm": 39.13573330990029,
+      "learning_rate": 9.055104900071375e-07,
+      "logits/chosen": -0.3008892834186554,
+      "logits/rejected": -0.1968032568693161,
+      "logps/chosen": -253.58798217773438,
+      "logps/rejected": -239.539794921875,
+      "loss": 0.5964,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.14220505952835083,
+      "rewards/margins": 0.6365345120429993,
+      "rewards/rejected": -0.49432945251464844,
+      "step": 400
+    },
+    {
+      "epoch": 0.8374771002355405,
+      "eval_logits/chosen": -0.3291381299495697,
+      "eval_logits/rejected": -0.35453417897224426,
+      "eval_logps/chosen": -310.317626953125,
+      "eval_logps/rejected": -275.15972900390625,
+      "eval_loss": 0.6100037693977356,
+      "eval_rewards/accuracies": 0.658730149269104,
+      "eval_rewards/chosen": -0.016582150012254715,
+      "eval_rewards/margins": 0.42150571942329407,
+      "eval_rewards/rejected": -0.43808794021606445,
+      "eval_runtime": 19.1374,
+      "eval_samples_per_second": 104.507,
+      "eval_steps_per_second": 3.292,
+      "step": 400
+    },
+    {
+      "epoch": 0.8584140277414289,
+      "grad_norm": 41.59836395581152,
+      "learning_rate": 8.982502001525777e-07,
+      "logits/chosen": -0.05211949348449707,
+      "logits/rejected": -0.16379515826702118,
+      "logps/chosen": -334.3597717285156,
+      "logps/rejected": -280.1282043457031,
+      "loss": 0.6085,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": 0.10034932941198349,
+      "rewards/margins": 0.5413358211517334,
+      "rewards/rejected": -0.4409864842891693,
+      "step": 410
+    },
+    {
+      "epoch": 0.8793509552473174,
+      "grad_norm": 43.48433949710367,
+      "learning_rate": 8.90752621580335e-07,
+      "logits/chosen": -0.10794836282730103,
+      "logits/rejected": -0.10175999253988266,
+      "logps/chosen": -289.7253723144531,
+      "logps/rejected": -290.4524841308594,
+      "loss": 0.5967,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.03317990154027939,
+      "rewards/margins": 0.5273382663726807,
+      "rewards/rejected": -0.4941583275794983,
+      "step": 420
+    },
+    {
+      "epoch": 0.9002878827532059,
+      "grad_norm": 38.200809011500716,
+      "learning_rate": 8.83022221559489e-07,
+      "logits/chosen": -0.08915697038173676,
+      "logits/rejected": -0.18852075934410095,
+      "logps/chosen": -325.81707763671875,
+      "logps/rejected": -257.1981506347656,
+      "loss": 0.5916,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.053056418895721436,
+      "rewards/margins": 0.5633317232131958,
+      "rewards/rejected": -0.5102753043174744,
+      "step": 430
+    },
+    {
+      "epoch": 0.9212248102590945,
+      "grad_norm": 37.18511688193762,
+      "learning_rate": 8.750636060807145e-07,
+      "logits/chosen": -0.0535130575299263,
+      "logits/rejected": -0.1133667603135109,
+      "logps/chosen": -299.4546813964844,
+      "logps/rejected": -248.877685546875,
+      "loss": 0.6032,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": 0.14839240908622742,
+      "rewards/margins": 0.4819253385066986,
+      "rewards/rejected": -0.3335329294204712,
+      "step": 440
+    },
+    {
+      "epoch": 0.942161737764983,
+      "grad_norm": 35.5219084800687,
+      "learning_rate": 8.668815171119019e-07,
+      "logits/chosen": -0.16904090344905853,
+      "logits/rejected": -0.12047908455133438,
+      "logps/chosen": -297.3148498535156,
+      "logps/rejected": -285.4595031738281,
+      "loss": 0.6148,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.051097553223371506,
+      "rewards/margins": 0.505901038646698,
+      "rewards/rejected": -0.4548035264015198,
+      "step": 450
+    },
+    {
+      "epoch": 0.9630986652708715,
+      "grad_norm": 40.701223424626065,
+      "learning_rate": 8.584808297727591e-07,
+      "logits/chosen": -0.07098624855279922,
+      "logits/rejected": -0.11652670055627823,
+      "logps/chosen": -291.76434326171875,
+      "logps/rejected": -246.9138641357422,
+      "loss": 0.6087,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": 0.10639622062444687,
+      "rewards/margins": 0.5133938193321228,
+      "rewards/rejected": -0.40699753165245056,
+      "step": 460
+    },
+    {
+      "epoch": 0.98403559277676,
+      "grad_norm": 41.789096950373676,
+      "learning_rate": 8.498665494300771e-07,
+      "logits/chosen": -0.19742052257061005,
+      "logits/rejected": -0.16556285321712494,
+      "logps/chosen": -302.3099060058594,
+      "logps/rejected": -277.933837890625,
+      "loss": 0.5919,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.026506727561354637,
+      "rewards/margins": 0.43121033906936646,
+      "rewards/rejected": -0.40470361709594727,
+      "step": 470
+    },
+    {
+      "epoch": 1.0049725202826485,
+      "grad_norm": 34.0036850645019,
+      "learning_rate": 8.410438087153911e-07,
+      "logits/chosen": -0.05161554738879204,
+      "logits/rejected": -0.1036214604973793,
+      "logps/chosen": -300.4537658691406,
+      "logps/rejected": -268.56048583984375,
+      "loss": 0.5867,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.00218582758679986,
+      "rewards/margins": 0.4712887704372406,
+      "rewards/rejected": -0.4734745919704437,
+      "step": 480
+    },
+    {
+      "epoch": 1.025909447788537,
+      "grad_norm": 36.00706898235858,
+      "learning_rate": 8.320178644668141e-07,
+      "logits/chosen": -0.11621465533971786,
+      "logits/rejected": -0.18176773190498352,
+      "logps/chosen": -299.8227233886719,
+      "logps/rejected": -256.3122253417969,
+      "loss": 0.54,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": 0.12694445252418518,
+      "rewards/margins": 0.600742757320404,
+      "rewards/rejected": -0.47379836440086365,
+      "step": 490
+    },
+    {
+      "epoch": 1.0468463752944255,
+      "grad_norm": 36.99527595331141,
+      "learning_rate": 8.22794094596864e-07,
+      "logits/chosen": -0.1485815942287445,
+      "logits/rejected": -0.20506855845451355,
+      "logps/chosen": -305.8435363769531,
+      "logps/rejected": -270.3700256347656,
+      "loss": 0.5394,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.029846886172890663,
+      "rewards/margins": 0.6080519556999207,
+      "rewards/rejected": -0.5782050490379333,
+      "step": 500
+    },
+    {
+      "epoch": 1.0468463752944255,
+      "eval_logits/chosen": -0.33202826976776123,
+      "eval_logits/rejected": -0.35762789845466614,
+      "eval_logps/chosen": -310.303955078125,
+      "eval_logps/rejected": -275.2332458496094,
+      "eval_loss": 0.6065964102745056,
+      "eval_rewards/accuracies": 0.7103174328804016,
+      "eval_rewards/chosen": -0.009755274280905724,
+      "eval_rewards/margins": 0.465115487575531,
+      "eval_rewards/rejected": -0.4748707413673401,
+      "eval_runtime": 18.9724,
+      "eval_samples_per_second": 105.416,
+      "eval_steps_per_second": 3.321,
+      "step": 500
+    },
+    {
+      "epoch": 1.067783302800314,
+      "grad_norm": 33.8725028749791,
+      "learning_rate": 8.133779948881513e-07,
+      "logits/chosen": -0.18220725655555725,
+      "logits/rejected": -0.1532759964466095,
+      "logps/chosen": -269.54156494140625,
+      "logps/rejected": -259.2676086425781,
+      "loss": 0.533,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.12358293682336807,
+      "rewards/margins": 0.7069646120071411,
+      "rewards/rejected": -0.5833816528320312,
+      "step": 510
+    },
+    {
+      "epoch": 1.0887202303062025,
+      "grad_norm": 32.98315936865049,
+      "learning_rate": 8.037751757188367e-07,
+      "logits/chosen": -0.2609891891479492,
+      "logits/rejected": -0.1464134156703949,
+      "logps/chosen": -262.7670593261719,
+      "logps/rejected": -260.43121337890625,
+      "loss": 0.5216,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": 0.10809774696826935,
+      "rewards/margins": 0.6094505786895752,
+      "rewards/rejected": -0.501352846622467,
+      "step": 520
+    },
+    {
+      "epoch": 1.109657157812091,
+      "grad_norm": 36.61105396566652,
+      "learning_rate": 7.939913587198095e-07,
+      "logits/chosen": -0.2482234686613083,
+      "logits/rejected": -0.3337472081184387,
+      "logps/chosen": -288.12847900390625,
+      "logps/rejected": -249.4586181640625,
+      "loss": 0.5439,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": 0.05145542696118355,
+      "rewards/margins": 0.7037261128425598,
+      "rewards/rejected": -0.6522706747055054,
+      "step": 530
+    },
+    {
+      "epoch": 1.1305940853179797,
+      "grad_norm": 38.97199731536176,
+      "learning_rate": 7.840323733655778e-07,
+      "logits/chosen": -0.19931235909461975,
+      "logits/rejected": -0.09966816008090973,
+      "logps/chosen": -259.0215759277344,
+      "logps/rejected": -237.9501495361328,
+      "loss": 0.5534,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": 0.12085232883691788,
+      "rewards/margins": 0.6305907964706421,
+      "rewards/rejected": -0.5097384452819824,
+      "step": 540
+    },
+    {
+      "epoch": 1.151531012823868,
+      "grad_norm": 35.71159175636079,
+      "learning_rate": 7.739041535009041e-07,
+      "logits/chosen": -0.14787928760051727,
+      "logits/rejected": -0.24329273402690887,
+      "logps/chosen": -305.64697265625,
+      "logps/rejected": -244.08755493164062,
+      "loss": 0.5301,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": 0.08590878546237946,
+      "rewards/margins": 0.6826232075691223,
+      "rewards/rejected": -0.596714437007904,
+      "step": 550
+    },
+    {
+      "epoch": 1.1724679403297567,
+      "grad_norm": 40.90404154922072,
+      "learning_rate": 7.636127338052511e-07,
+      "logits/chosen": -0.10747908055782318,
+      "logits/rejected": -0.11999531090259552,
+      "logps/chosen": -294.2823486328125,
+      "logps/rejected": -265.80010986328125,
+      "loss": 0.5442,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": 0.1678900569677353,
+      "rewards/margins": 0.6707212924957275,
+      "rewards/rejected": -0.502831220626831,
+      "step": 560
+    },
+    {
+      "epoch": 1.193404867835645,
+      "grad_norm": 36.957100723172736,
+      "learning_rate": 7.531642461971514e-07,
+      "logits/chosen": -0.24748548865318298,
+      "logits/rejected": -0.13145090639591217,
+      "logps/chosen": -270.5980529785156,
+      "logps/rejected": -276.2918395996094,
+      "loss": 0.5296,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": 0.1355637013912201,
+      "rewards/margins": 0.7545391917228699,
+      "rewards/rejected": -0.6189755797386169,
+      "step": 570
+    },
+    {
+      "epoch": 1.2143417953415336,
+      "grad_norm": 36.431278984426584,
+      "learning_rate": 7.425649161806352e-07,
+      "logits/chosen": -0.22467997670173645,
+      "logits/rejected": -0.20354416966438293,
+      "logps/chosen": -284.69439697265625,
+      "logps/rejected": -250.2189483642578,
+      "loss": 0.5209,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": 0.03756130859255791,
+      "rewards/margins": 0.7149747610092163,
+      "rewards/rejected": -0.6774134635925293,
+      "step": 580
+    },
+    {
+      "epoch": 1.235278722847422,
+      "grad_norm": 34.3971566508403,
+      "learning_rate": 7.318210591359008e-07,
+      "logits/chosen": -0.02497723139822483,
+      "logits/rejected": -0.0003952041151933372,
+      "logps/chosen": -316.3713684082031,
+      "logps/rejected": -274.1322937011719,
+      "loss": 0.5209,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": 0.07891981303691864,
+      "rewards/margins": 0.7175663113594055,
+      "rewards/rejected": -0.6386464834213257,
+      "step": 590
+    },
+    {
+      "epoch": 1.2562156503533106,
+      "grad_norm": 37.248359587110805,
+      "learning_rate": 7.209390765564318e-07,
+      "logits/chosen": -0.15565916895866394,
+      "logits/rejected": -0.10866830497980118,
+      "logps/chosen": -306.9658203125,
+      "logps/rejected": -306.49078369140625,
+      "loss": 0.5099,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": 0.1327342391014099,
+      "rewards/margins": 0.7248380780220032,
+      "rewards/rejected": -0.5921037197113037,
+      "step": 600
+    },
+    {
+      "epoch": 1.2562156503533106,
+      "eval_logits/chosen": -0.33804643154144287,
+      "eval_logits/rejected": -0.3634737432003021,
+      "eval_logps/chosen": -310.3228759765625,
+      "eval_logps/rejected": -275.3493347167969,
+      "eval_loss": 0.6006649136543274,
+      "eval_rewards/accuracies": 0.6785714030265808,
+      "eval_rewards/chosen": -0.019210144877433777,
+      "eval_rewards/margins": 0.5136736631393433,
+      "eval_rewards/rejected": -0.5328837633132935,
+      "eval_runtime": 19.1325,
+      "eval_samples_per_second": 104.534,
+      "eval_steps_per_second": 3.293,
+      "step": 600
+    },
+    {
+      "epoch": 1.2771525778591992,
+      "grad_norm": 39.48209474577774,
+      "learning_rate": 7.099254522348064e-07,
+      "logits/chosen": -0.09819002449512482,
+      "logits/rejected": -0.1692955642938614,
+      "logps/chosen": -290.13140869140625,
+      "logps/rejected": -237.3570098876953,
+      "loss": 0.5202,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.1899166852235794,
+      "rewards/margins": 0.8487440347671509,
+      "rewards/rejected": -0.6588274240493774,
+      "step": 610
+    },
+    {
+      "epoch": 1.2980895053650876,
+      "grad_norm": 35.38064058006504,
+      "learning_rate": 6.987867483994716e-07,
+      "logits/chosen": 0.08088377118110657,
+      "logits/rejected": -0.028685202822089195,
+      "logps/chosen": -286.01324462890625,
+      "logps/rejected": -242.1278076171875,
+      "loss": 0.5113,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": 0.08664991706609726,
+      "rewards/margins": 0.6891879439353943,
+      "rewards/rejected": -0.6025381088256836,
+      "step": 620
+    },
+    {
+      "epoch": 1.3190264328709762,
+      "grad_norm": 36.280579625575726,
+      "learning_rate": 6.875296018047809e-07,
+      "logits/chosen": -0.15825395286083221,
+      "logits/rejected": -0.13115188479423523,
+      "logps/chosen": -291.768798828125,
+      "logps/rejected": -271.0327453613281,
+      "loss": 0.5133,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": 0.25919127464294434,
+      "rewards/margins": 0.8277707099914551,
+      "rewards/rejected": -0.5685793161392212,
+      "step": 630
+    },
+    {
+      "epoch": 1.3399633603768648,
+      "grad_norm": 32.57724385812019,
+      "learning_rate": 6.761607197766296e-07,
+      "logits/chosen": -0.21067364513874054,
+      "logits/rejected": -0.15081565082073212,
+      "logps/chosen": -272.00518798828125,
+      "logps/rejected": -279.40478515625,
+      "loss": 0.5105,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.14864543080329895,
+      "rewards/margins": 0.7687338590621948,
+      "rewards/rejected": -0.6200884580612183,
+      "step": 640
+    },
+    {
+      "epoch": 1.3609002878827532,
+      "grad_norm": 37.48616481446126,
+      "learning_rate": 6.646868762160398e-07,
+      "logits/chosen": -0.19943957030773163,
+      "logits/rejected": -0.161566823720932,
+      "logps/chosen": -268.23638916015625,
+      "logps/rejected": -252.7967987060547,
+      "loss": 0.5372,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.008694097399711609,
+      "rewards/margins": 0.623866617679596,
+      "rewards/rejected": -0.6151725053787231,
+      "step": 650
+    },
+    {
+      "epoch": 1.3818372153886418,
+      "grad_norm": 39.563737783736045,
+      "learning_rate": 6.531149075630796e-07,
+      "logits/chosen": -0.11065585911273956,
+      "logits/rejected": -0.17729897797107697,
+      "logps/chosen": -263.4093933105469,
+      "logps/rejected": -251.0884246826172,
+      "loss": 0.5323,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.026829296723008156,
+      "rewards/margins": 0.4506412148475647,
+      "rewards/rejected": -0.4774704873561859,
+      "step": 660
+    },
+    {
+      "epoch": 1.4027741428945302,
+      "grad_norm": 33.46296192971815,
+      "learning_rate": 6.414517087235185e-07,
+      "logits/chosen": -0.20335140824317932,
+      "logits/rejected": -0.2963979244232178,
+      "logps/chosen": -293.1023864746094,
+      "logps/rejected": -253.64291381835938,
+      "loss": 0.5049,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": 0.00017919539823196828,
+      "rewards/margins": 0.5864929556846619,
+      "rewards/rejected": -0.5863137245178223,
+      "step": 670
+    },
+    {
+      "epoch": 1.4237110704004188,
+      "grad_norm": 40.03881951253335,
+      "learning_rate": 6.297042289606479e-07,
+      "logits/chosen": -0.13722732663154602,
+      "logits/rejected": -0.1250244826078415,
+      "logps/chosen": -295.9792785644531,
+      "logps/rejected": -296.83038330078125,
+      "loss": 0.5133,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.014182251878082752,
+      "rewards/margins": 0.5892963409423828,
+      "rewards/rejected": -0.6034785509109497,
+      "step": 680
+    },
+    {
+      "epoch": 1.4446479979063072,
+      "grad_norm": 29.250271504061402,
+      "learning_rate": 6.178794677547137e-07,
+      "logits/chosen": -0.18004044890403748,
+      "logits/rejected": -0.1459943801164627,
+      "logps/chosen": -298.6792297363281,
+      "logps/rejected": -257.5166931152344,
+      "loss": 0.5157,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.07589882612228394,
+      "rewards/margins": 0.7056849598884583,
+      "rewards/rejected": -0.6297860741615295,
+      "step": 690
+    },
+    {
+      "epoch": 1.4655849254121958,
+      "grad_norm": 32.49512416553128,
+      "learning_rate": 6.059844706324286e-07,
+      "logits/chosen": -0.07275749742984772,
+      "logits/rejected": 0.005568481981754303,
+      "logps/chosen": -314.0689697265625,
+      "logps/rejected": -328.4625244140625,
+      "loss": 0.5056,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.13801348209381104,
+      "rewards/margins": 0.876538872718811,
+      "rewards/rejected": -0.738525390625,
+      "step": 700
+    },
+    {
+      "epoch": 1.4655849254121958,
+      "eval_logits/chosen": -0.3406723737716675,
+      "eval_logits/rejected": -0.36724287271499634,
+      "eval_logps/chosen": -310.4104309082031,
+      "eval_logps/rejected": -275.4717102050781,
+      "eval_loss": 0.5875913500785828,
+      "eval_rewards/accuracies": 0.6904761791229248,
+      "eval_rewards/chosen": -0.06298798322677612,
+      "eval_rewards/margins": 0.5310800671577454,
+      "eval_rewards/rejected": -0.5940679311752319,
+      "eval_runtime": 19.0211,
+      "eval_samples_per_second": 105.146,
+      "eval_steps_per_second": 3.312,
+      "step": 700
+    },
+    {
+      "epoch": 1.4865218529180844,
+      "grad_norm": 35.75327607918745,
+      "learning_rate": 5.940263249690477e-07,
+      "logits/chosen": -0.24695155024528503,
+      "logits/rejected": -0.22096967697143555,
+      "logps/chosen": -295.2511901855469,
+      "logps/rejected": -276.71051025390625,
+      "loss": 0.5132,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": 0.058757662773132324,
+      "rewards/margins": 0.7594215273857117,
+      "rewards/rejected": -0.7006638646125793,
+      "step": 710
+    },
+    {
+      "epoch": 1.5074587804239727,
+      "grad_norm": 41.16638143070179,
+      "learning_rate": 5.820121557655108e-07,
+      "logits/chosen": -0.1299329400062561,
+      "logits/rejected": -0.10874001681804657,
+      "logps/chosen": -303.56732177734375,
+      "logps/rejected": -280.8491516113281,
+      "loss": 0.5135,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": 0.11438952386379242,
+      "rewards/margins": 0.7343131899833679,
+      "rewards/rejected": -0.6199236512184143,
+      "step": 720
+    },
+    {
+      "epoch": 1.5283957079298613,
+      "grad_norm": 41.49005357475057,
+      "learning_rate": 5.699491214031657e-07,
+      "logits/chosen": -0.14048215746879578,
+      "logits/rejected": -0.18900522589683533,
+      "logps/chosen": -280.69598388671875,
+      "logps/rejected": -255.81298828125,
+      "loss": 0.5166,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.11929772049188614,
+      "rewards/margins": 0.7799339294433594,
+      "rewards/rejected": -0.6606361269950867,
+      "step": 730
+    },
+    {
+      "epoch": 1.54933263543575,
+      "grad_norm": 34.97801212344648,
+      "learning_rate": 5.578444093786008e-07,
+      "logits/chosen": 0.06411169469356537,
+      "logits/rejected": 0.15786513686180115,
+      "logps/chosen": -315.06182861328125,
+      "logps/rejected": -282.93115234375,
+      "loss": 0.5128,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": 0.24365051090717316,
+      "rewards/margins": 0.8540836572647095,
+      "rewards/rejected": -0.6104331612586975,
+      "step": 740
+    },
+    {
+      "epoch": 1.5702695629416383,
+      "grad_norm": 32.53803724377542,
+      "learning_rate": 5.457052320211339e-07,
+      "logits/chosen": -0.22511212527751923,
+      "logits/rejected": -0.22842809557914734,
+      "logps/chosen": -298.0352783203125,
+      "logps/rejected": -268.8421936035156,
+      "loss": 0.4961,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.09702634066343307,
+      "rewards/margins": 0.632623016834259,
+      "rewards/rejected": -0.5355967283248901,
+      "step": 750
+    },
+    {
+      "epoch": 1.5912064904475267,
+      "grad_norm": 36.857273577646474,
+      "learning_rate": 5.335388221955012e-07,
+      "logits/chosen": -0.044903479516506195,
+      "logits/rejected": 0.0046501667238771915,
+      "logps/chosen": -358.9960632324219,
+      "logps/rejected": -336.44354248046875,
+      "loss": 0.4997,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": 0.28686124086380005,
+      "rewards/margins": 0.7968252301216125,
+      "rewards/rejected": -0.5099639296531677,
+      "step": 760
+    },
+    {
+      "epoch": 1.6121434179534153,
+      "grad_norm": 31.737574949447513,
+      "learning_rate": 5.213524289923126e-07,
+      "logits/chosen": -0.03366886079311371,
+      "logits/rejected": -0.18763691186904907,
+      "logps/chosen": -327.0910339355469,
+      "logps/rejected": -272.4969787597656,
+      "loss": 0.5114,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": 0.2613201141357422,
+      "rewards/margins": 0.8974519968032837,
+      "rewards/rejected": -0.6361318826675415,
+      "step": 770
+    },
+    {
+      "epoch": 1.633080345459304,
+      "grad_norm": 34.36622996446216,
+      "learning_rate": 5.091533134088387e-07,
+      "logits/chosen": -0.12894697487354279,
+      "logits/rejected": -0.142494797706604,
+      "logps/chosen": -320.5164794921875,
+      "logps/rejected": -280.9849548339844,
+      "loss": 0.5146,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.27436262369155884,
+      "rewards/margins": 0.8554395437240601,
+      "rewards/rejected": -0.581076979637146,
+      "step": 780
+    },
+    {
+      "epoch": 1.6540172729651923,
+      "grad_norm": 38.709599843434866,
+      "learning_rate": 4.969487440227038e-07,
+      "logits/chosen": -0.21234026551246643,
+      "logits/rejected": -0.22163410484790802,
+      "logps/chosen": -300.9823303222656,
+      "logps/rejected": -276.39508056640625,
+      "loss": 0.4852,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.24367956817150116,
+      "rewards/margins": 0.9389246106147766,
+      "rewards/rejected": -0.6952449679374695,
+      "step": 790
+    },
+    {
+      "epoch": 1.674954200471081,
+      "grad_norm": 31.812270366410278,
+      "learning_rate": 4.847459926610619e-07,
+      "logits/chosen": -0.018023919314146042,
+      "logits/rejected": -0.12209578603506088,
+      "logps/chosen": -350.00439453125,
+      "logps/rejected": -305.96954345703125,
+      "loss": 0.4936,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": 0.1853758990764618,
+      "rewards/margins": 0.8932290077209473,
+      "rewards/rejected": -0.7078530788421631,
+      "step": 800
+    },
+    {
+      "epoch": 1.674954200471081,
+      "eval_logits/chosen": -0.3383854031562805,
+      "eval_logits/rejected": -0.36581355333328247,
+      "eval_logps/chosen": -310.3436584472656,
+      "eval_logps/rejected": -275.40155029296875,
+      "eval_loss": 0.5994271636009216,
+      "eval_rewards/accuracies": 0.6746031641960144,
+      "eval_rewards/chosen": -0.029591551050543785,
+      "eval_rewards/margins": 0.5294089317321777,
+      "eval_rewards/rejected": -0.559000551700592,
+      "eval_runtime": 19.1971,
+      "eval_samples_per_second": 104.182,
+      "eval_steps_per_second": 3.282,
+      "step": 800
+    },
+    {
+      "epoch": 1.6958911279769695,
+      "grad_norm": 35.049859279627235,
+      "learning_rate": 4.7255233006783624e-07,
+      "logits/chosen": -0.18587855994701385,
+      "logits/rejected": -0.16283151507377625,
+      "logps/chosen": -307.56439208984375,
+      "logps/rejected": -278.2900390625,
+      "loss": 0.5055,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": 0.10620725154876709,
+      "rewards/margins": 0.892121434211731,
+      "rewards/rejected": -0.7859140634536743,
+      "step": 810
+    },
+    {
+      "epoch": 1.7168280554828579,
+      "grad_norm": 34.134111938828184,
+      "learning_rate": 4.6037502157160567e-07,
+      "logits/chosen": -0.2755785584449768,
+      "logits/rejected": -0.31054019927978516,
+      "logps/chosen": -320.56695556640625,
+      "logps/rejected": -273.4154968261719,
+      "loss": 0.5048,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": 0.20151178538799286,
+      "rewards/margins": 0.9227310419082642,
+      "rewards/rejected": -0.7212191820144653,
+      "step": 820
+    },
+    {
+      "epoch": 1.7377649829887463,
+      "grad_norm": 38.68230830225304,
+      "learning_rate": 4.482213227567161e-07,
+      "logits/chosen": -0.2091369926929474,
+      "logits/rejected": -0.14194951951503754,
+      "logps/chosen": -277.35693359375,
+      "logps/rejected": -281.7249450683594,
+      "loss": 0.484,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": 0.10125979036092758,
+      "rewards/margins": 0.8731620907783508,
+      "rewards/rejected": -0.7719023823738098,
+      "step": 830
+    },
+    {
+      "epoch": 1.7587019104946349,
+      "grad_norm": 33.945256383172804,
+      "learning_rate": 4.3609847514019763e-07,
+      "logits/chosen": -0.30677181482315063,
+      "logits/rejected": -0.1776028573513031,
+      "logps/chosen": -261.38873291015625,
+      "logps/rejected": -267.8268127441406,
+      "loss": 0.5094,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": 0.18778590857982635,
+      "rewards/margins": 0.8258863687515259,
+      "rewards/rejected": -0.6381004452705383,
+      "step": 840
+    },
+    {
+      "epoch": 1.7796388380005235,
+      "grad_norm": 34.20224125304341,
+      "learning_rate": 4.240137018570661e-07,
+      "logits/chosen": -0.11399509757757187,
+      "logits/rejected": -0.1185418963432312,
+      "logps/chosen": -275.3580017089844,
+      "logps/rejected": -282.1163024902344,
+      "loss": 0.5074,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.03538894280791283,
+      "rewards/margins": 0.7777508497238159,
+      "rewards/rejected": -0.813139796257019,
+      "step": 850
+    },
+    {
+      "epoch": 1.8005757655064119,
+      "grad_norm": 31.946083788257877,
+      "learning_rate": 4.1197420335657366e-07,
+      "logits/chosen": -0.09024197608232498,
+      "logits/rejected": -0.20059815049171448,
+      "logps/chosen": -298.12762451171875,
+      "logps/rejected": -242.2689666748047,
+      "loss": 0.5089,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.15691380202770233,
+      "rewards/margins": 0.8090406656265259,
+      "rewards/rejected": -0.6521269083023071,
+      "step": 860
+    },
+    {
+      "epoch": 1.8215126930123005,
+      "grad_norm": 28.583891688236477,
+      "learning_rate": 3.9998715311197783e-07,
+      "logits/chosen": -0.1159423366189003,
+      "logits/rejected": -0.16631600260734558,
+      "logps/chosen": -337.1079406738281,
+      "logps/rejected": -285.5828857421875,
+      "loss": 0.4894,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.11767788976430893,
+      "rewards/margins": 0.6672422289848328,
+      "rewards/rejected": -0.5495643615722656,
+      "step": 870
+    },
+    {
+      "epoch": 1.842449620518189,
+      "grad_norm": 33.76166639382152,
+      "learning_rate": 3.880596933463843e-07,
+      "logits/chosen": -0.07985590398311615,
+      "logits/rejected": -0.03050703927874565,
+      "logps/chosen": -274.9603576660156,
+      "logps/rejected": -266.6165771484375,
+      "loss": 0.5075,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": 0.09320048242807388,
+      "rewards/margins": 0.7023938298225403,
+      "rewards/rejected": -0.6091933250427246,
+      "step": 880
+    },
+    {
+      "epoch": 1.8633865480240774,
+      "grad_norm": 40.29693979290027,
+      "learning_rate": 3.761989307772085e-07,
+      "logits/chosen": -0.15825888514518738,
+      "logits/rejected": -0.1759234219789505,
+      "logps/chosen": -306.653076171875,
+      "logps/rejected": -279.01177978515625,
+      "loss": 0.5055,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": 0.16093602776527405,
+      "rewards/margins": 0.937556266784668,
+      "rewards/rejected": -0.7766203284263611,
+      "step": 890
+    },
+    {
+      "epoch": 1.8843234755299658,
+      "grad_norm": 36.002728955160904,
+      "learning_rate": 3.6441193238179146e-07,
+      "logits/chosen": -0.05288013815879822,
+      "logits/rejected": -0.13031360507011414,
+      "logps/chosen": -334.34417724609375,
+      "logps/rejected": -269.53558349609375,
+      "loss": 0.4904,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": 0.14437799155712128,
+      "rewards/margins": 0.8378474116325378,
+      "rewards/rejected": -0.6934694051742554,
+      "step": 900
+    },
+    {
+      "epoch": 1.8843234755299658,
+      "eval_logits/chosen": -0.3443281650543213,
+      "eval_logits/rejected": -0.3704880475997925,
+      "eval_logps/chosen": -310.4006042480469,
+      "eval_logps/rejected": -275.5133972167969,
+      "eval_loss": 0.5989052653312683,
+      "eval_rewards/accuracies": 0.6944444179534912,
+      "eval_rewards/chosen": -0.05807310715317726,
+      "eval_rewards/margins": 0.55684494972229,
+      "eval_rewards/rejected": -0.614918053150177,
+      "eval_runtime": 18.9036,
+      "eval_samples_per_second": 105.8,
+      "eval_steps_per_second": 3.333,
+      "step": 900
+    },
+    {
+      "epoch": 1.9052604030358546,
+      "grad_norm": 31.165480893885753,
+      "learning_rate": 3.5270572118669715e-07,
+      "logits/chosen": -0.11026148498058319,
+      "logits/rejected": -0.16978046298027039,
+      "logps/chosen": -278.1309509277344,
+      "logps/rejected": -235.88150024414062,
+      "loss": 0.5272,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.02903774380683899,
+      "rewards/margins": 0.8374470472335815,
+      "rewards/rejected": -0.8084093332290649,
+      "step": 910
+    },
+    {
+      "epoch": 1.926197330541743,
+      "grad_norm": 40.996379136014625,
+      "learning_rate": 3.4108727208319314e-07,
+      "logits/chosen": -0.17259056866168976,
+      "logits/rejected": -0.20306821167469025,
+      "logps/chosen": -312.1188049316406,
+      "logps/rejected": -266.6794128417969,
+      "loss": 0.5039,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.16847112774848938,
+      "rewards/margins": 0.8200035095214844,
+      "rewards/rejected": -0.6515323519706726,
+      "step": 920
+    },
+    {
+      "epoch": 1.9471342580476314,
+      "grad_norm": 31.825399091982387,
+      "learning_rate": 3.295635076714144e-07,
+      "logits/chosen": -0.1496247947216034,
+      "logits/rejected": -0.17214402556419373,
+      "logps/chosen": -301.103515625,
+      "logps/rejected": -270.370361328125,
+      "loss": 0.4903,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.16043229401111603,
+      "rewards/margins": 0.7891928553581238,
+      "rewards/rejected": -0.628760576248169,
+      "step": 930
+    },
+    {
+      "epoch": 1.96807118555352,
+      "grad_norm": 39.834235726892466,
+      "learning_rate": 3.181412941356816e-07,
+      "logits/chosen": -0.13259033858776093,
+      "logits/rejected": -0.11601094901561737,
+      "logps/chosen": -266.5997314453125,
+      "logps/rejected": -231.60848999023438,
+      "loss": 0.485,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.19742903113365173,
+      "rewards/margins": 0.8145695924758911,
+      "rewards/rejected": -0.6171405911445618,
+      "step": 940
+    },
+    {
+      "epoch": 1.9890081130594086,
+      "grad_norm": 35.65864774290716,
+      "learning_rate": 3.068274371534356e-07,
+      "logits/chosen": -0.11888673156499863,
+      "logits/rejected": -0.2420310229063034,
+      "logps/chosen": -301.62213134765625,
+      "logps/rejected": -250.2771453857422,
+      "loss": 0.4887,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": 0.002846676157787442,
+      "rewards/margins": 0.7688849568367004,
+      "rewards/rejected": -0.7660382986068726,
+      "step": 950
+    },
+    {
+      "epoch": 2.009945040565297,
+      "grad_norm": 28.814451971980255,
+      "learning_rate": 2.956286778402226e-07,
+      "logits/chosen": -0.10088515281677246,
+      "logits/rejected": -0.12641310691833496,
+      "logps/chosen": -301.3664855957031,
+      "logps/rejected": -283.16461181640625,
+      "loss": 0.4567,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.20988380908966064,
+      "rewards/margins": 0.9322754740715027,
+      "rewards/rejected": -0.7223917245864868,
+      "step": 960
+    },
+    {
+      "epoch": 2.0308819680711854,
+      "grad_norm": 28.79413061525845,
+      "learning_rate": 2.84551688733146e-07,
+      "logits/chosen": -0.07364392280578613,
+      "logits/rejected": -0.1452295333147049,
+      "logps/chosen": -299.0059509277344,
+      "logps/rejected": -238.53231811523438,
+      "loss": 0.4557,
+      "rewards/accuracies": 0.84375,
+      "rewards/chosen": 0.166302889585495,
+      "rewards/margins": 0.9308949708938599,
+      "rewards/rejected": -0.7645919919013977,
+      "step": 970
+    },
+    {
+      "epoch": 2.051818895577074,
+      "grad_norm": 31.23059423273923,
+      "learning_rate": 2.7360306981518147e-07,
+      "logits/chosen": -0.09392131119966507,
+      "logits/rejected": -0.047506559640169144,
+      "logps/chosen": -262.2680969238281,
+      "logps/rejected": -253.03271484375,
+      "loss": 0.4445,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.040274180471897125,
+      "rewards/margins": 0.7928990125656128,
+      "rewards/rejected": -0.7526248097419739,
+      "step": 980
+    },
+    {
+      "epoch": 2.0727558230829626,
+      "grad_norm": 27.855961833782285,
+      "learning_rate": 2.6278934458271996e-07,
+      "logits/chosen": -0.1288485825061798,
+      "logits/rejected": -0.1765173375606537,
+      "logps/chosen": -314.7231750488281,
+      "logps/rejected": -302.44964599609375,
+      "loss": 0.4327,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.18189235031604767,
+      "rewards/margins": 1.0036437511444092,
+      "rewards/rejected": -0.8217514157295227,
+      "step": 990
+    },
+    {
+      "epoch": 2.093692750588851,
+      "grad_norm": 35.285113595842645,
+      "learning_rate": 2.5211695615868456e-07,
+      "logits/chosen": -0.1857554316520691,
+      "logits/rejected": -0.3039974272251129,
+      "logps/chosen": -281.1855163574219,
+      "logps/rejected": -240.62710571289062,
+      "loss": 0.4622,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": 0.15526309609413147,
+      "rewards/margins": 0.9393559694290161,
+      "rewards/rejected": -0.7840928435325623,
+      "step": 1000
+    },
+    {
+      "epoch": 2.093692750588851,
+      "eval_logits/chosen": -0.3450426459312439,
+      "eval_logits/rejected": -0.37242141366004944,
+      "eval_logps/chosen": -310.4169006347656,
+      "eval_logps/rejected": -275.49713134765625,
+      "eval_loss": 0.59389328956604,
+      "eval_rewards/accuracies": 0.6944444179534912,
+      "eval_rewards/chosen": -0.06623569875955582,
+      "eval_rewards/margins": 0.5405489802360535,
+      "eval_rewards/rejected": -0.6067846417427063,
+      "eval_runtime": 19.108,
+      "eval_samples_per_second": 104.668,
+      "eval_steps_per_second": 3.297,
+      "step": 1000
+    },
+    {
+      "epoch": 2.11462967809474,
+      "grad_norm": 32.59720699053956,
+      "learning_rate": 2.4159226345353647e-07,
+      "logits/chosen": -0.30413442850112915,
+      "logits/rejected": -0.24426865577697754,
+      "logps/chosen": -270.49920654296875,
+      "logps/rejected": -255.44723510742188,
+      "loss": 0.4508,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": 0.03533598408102989,
+      "rewards/margins": 0.7992294430732727,
+      "rewards/rejected": -0.7638934254646301,
+      "step": 1010
+    },
+    {
+      "epoch": 2.135566605600628,
+      "grad_norm": 30.8340208692361,
+      "learning_rate": 2.312215373764551e-07,
+      "logits/chosen": -0.07855603843927383,
+      "logits/rejected": -0.094606414437294,
+      "logps/chosen": -289.9374084472656,
+      "logps/rejected": -277.69586181640625,
+      "loss": 0.4353,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.19073505699634552,
+      "rewards/margins": 0.8868581652641296,
+      "rewards/rejected": -0.6961231231689453,
+      "step": 1020
+    },
+    {
+      "epoch": 2.1565035331065165,
+      "grad_norm": 29.34387598294554,
+      "learning_rate": 2.2101095709895512e-07,
+      "logits/chosen": 0.009572875685989857,
+      "logits/rejected": -0.09117701649665833,
+      "logps/chosen": -282.7410583496094,
+      "logps/rejected": -239.8011932373047,
+      "loss": 0.438,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": 0.12627363204956055,
+      "rewards/margins": 0.8894168138504028,
+      "rewards/rejected": -0.7631431221961975,
+      "step": 1030
+    },
+    {
+      "epoch": 2.177440460612405,
+      "grad_norm": 29.466746389871066,
+      "learning_rate": 2.1096660637315932e-07,
+      "logits/chosen": -0.0853545218706131,
+      "logits/rejected": -0.1613493412733078,
+      "logps/chosen": -309.2115783691406,
+      "logps/rejected": -279.7373352050781,
+      "loss": 0.4374,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": 0.2535983920097351,
+      "rewards/margins": 1.087397575378418,
+      "rewards/rejected": -0.8337991833686829,
+      "step": 1040
+    },
+    {
+      "epoch": 2.1983773881182938,
+      "grad_norm": 31.3324256474218,
+      "learning_rate": 2.0109446990692963e-07,
+      "logits/chosen": -0.17987683415412903,
+      "logits/rejected": -0.24151673913002014,
+      "logps/chosen": -323.58477783203125,
+      "logps/rejected": -268.08111572265625,
+      "loss": 0.4365,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.09657944738864899,
+      "rewards/margins": 0.8588522672653198,
+      "rewards/rejected": -0.7622728943824768,
+      "step": 1050
+    },
+    {
+      "epoch": 2.219314315624182,
+      "grad_norm": 33.25666630785023,
+      "learning_rate": 1.9140042979800737e-07,
+      "logits/chosen": -0.13123273849487305,
+      "logits/rejected": -0.12296972423791885,
+      "logps/chosen": -260.5819396972656,
+      "logps/rejected": -232.4111328125,
+      "loss": 0.4355,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": 0.05866159126162529,
+      "rewards/margins": 0.7927514314651489,
+      "rewards/rejected": -0.7340899109840393,
+      "step": 1060
+    },
+    {
+      "epoch": 2.2402512431300705,
+      "grad_norm": 27.925843344985296,
+      "learning_rate": 1.8189026202929391e-07,
+      "logits/chosen": -0.1511206328868866,
+      "logits/rejected": -0.17038318514823914,
+      "logps/chosen": -290.31951904296875,
+      "logps/rejected": -260.115966796875,
+      "loss": 0.4534,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": 0.11966502666473389,
+      "rewards/margins": 0.9545691609382629,
+      "rewards/rejected": -0.8349041938781738,
+      "step": 1070
+    },
+    {
+      "epoch": 2.2611881706359593,
+      "grad_norm": 26.475722050618995,
+      "learning_rate": 1.725696330273575e-07,
+      "logits/chosen": -0.08606094121932983,
+      "logits/rejected": -0.1871510148048401,
+      "logps/chosen": -288.9709777832031,
+      "logps/rejected": -249.9591064453125,
+      "loss": 0.4441,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": 0.19028839468955994,
+      "rewards/margins": 1.0648086071014404,
+      "rewards/rejected": -0.8745201826095581,
+      "step": 1080
+    },
+    {
+      "epoch": 2.2821250981418477,
+      "grad_norm": 28.63389247775027,
+      "learning_rate": 1.634440962862148e-07,
+      "logits/chosen": -0.05144480615854263,
+      "logits/rejected": -0.04060221463441849,
+      "logps/chosen": -295.44122314453125,
+      "logps/rejected": -262.346435546875,
+      "loss": 0.4318,
+      "rewards/accuracies": 0.8062499761581421,
+      "rewards/chosen": 0.20048430562019348,
+      "rewards/margins": 0.9130541086196899,
+      "rewards/rejected": -0.7125697135925293,
+      "step": 1090
+    },
+    {
+      "epoch": 2.303062025647736,
+      "grad_norm": 28.577071054830697,
+      "learning_rate": 1.545190890584042e-07,
+      "logits/chosen": -0.23821565508842468,
+      "logits/rejected": -0.23792441189289093,
+      "logps/chosen": -287.6651306152344,
+      "logps/rejected": -261.96875,
+      "loss": 0.4458,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.24947874248027802,
+      "rewards/margins": 0.945774257183075,
+      "rewards/rejected": -0.696295440196991,
+      "step": 1100
+    },
+    {
+      "epoch": 2.303062025647736,
+      "eval_logits/chosen": -0.3449550271034241,
+      "eval_logits/rejected": -0.3728056252002716,
+      "eval_logps/chosen": -310.3917541503906,
+      "eval_logps/rejected": -275.5622253417969,
+      "eval_loss": 0.5923458933830261,
+      "eval_rewards/accuracies": 0.6944444179534912,
+      "eval_rewards/chosen": -0.05364745110273361,
+      "eval_rewards/margins": 0.5857000946998596,
+      "eval_rewards/rejected": -0.639347493648529,
+      "eval_runtime": 19.0051,
+      "eval_samples_per_second": 105.235,
+      "eval_steps_per_second": 3.315,
+      "step": 1100
+    },
+    {
+      "epoch": 2.323998953153625,
+      "grad_norm": 30.512738174296523,
+      "learning_rate": 1.4579992911531496e-07,
+      "logits/chosen": -0.2525126338005066,
+      "logits/rejected": -0.3065794110298157,
+      "logps/chosen": -325.3319091796875,
+      "logps/rejected": -288.85089111328125,
+      "loss": 0.4406,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": 0.28674450516700745,
+      "rewards/margins": 1.1190677881240845,
+      "rewards/rejected": -0.8323231935501099,
+      "step": 1110
+    },
+    {
+      "epoch": 2.3449358806595133,
+      "grad_norm": 28.019655950524385,
+      "learning_rate": 1.372918115787112e-07,
+      "logits/chosen": -0.19494856894016266,
+      "logits/rejected": -0.16709737479686737,
+      "logps/chosen": -269.15802001953125,
+      "logps/rejected": -281.6461486816406,
+      "loss": 0.4203,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": 0.2475571632385254,
+      "rewards/margins": 1.078798770904541,
+      "rewards/rejected": -0.8312416076660156,
+      "step": 1120
+    },
+    {
+      "epoch": 2.3658728081654017,
+      "grad_norm": 30.412163422048742,
+      "learning_rate": 1.289998058253297e-07,
+      "logits/chosen": -0.30280619859695435,
+      "logits/rejected": -0.23964910209178925,
+      "logps/chosen": -300.67291259765625,
+      "logps/rejected": -286.5229187011719,
+      "loss": 0.4429,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.10387909412384033,
+      "rewards/margins": 0.9766238331794739,
+      "rewards/rejected": -0.8727447390556335,
+      "step": 1130
+    },
+    {
+      "epoch": 2.38680973567129,
+      "grad_norm": 26.079069616215104,
+      "learning_rate": 1.209288524664029e-07,
+      "logits/chosen": -0.09895231574773788,
+      "logits/rejected": -0.04007222130894661,
+      "logps/chosen": -295.33795166015625,
+      "logps/rejected": -295.7159118652344,
+      "loss": 0.4524,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.3174125850200653,
+      "rewards/margins": 1.068551778793335,
+      "rewards/rejected": -0.7511391639709473,
+      "step": 1140
+    },
+    {
+      "epoch": 2.407746663177179,
+      "grad_norm": 35.04317303338736,
+      "learning_rate": 1.1308376040390344e-07,
+      "logits/chosen": -0.2176963984966278,
+      "logits/rejected": -0.2429191619157791,
+      "logps/chosen": -268.51568603515625,
+      "logps/rejected": -243.83602905273438,
+      "loss": 0.4648,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": 0.1225685253739357,
+      "rewards/margins": 0.8287593722343445,
+      "rewards/rejected": -0.7061909437179565,
+      "step": 1150
+    },
+    {
+      "epoch": 2.4286835906830673,
+      "grad_norm": 27.071258876733747,
+      "learning_rate": 1.0546920396526382e-07,
+      "logits/chosen": -0.05287874490022659,
+      "logits/rejected": -0.08571253716945648,
+      "logps/chosen": -345.3739013671875,
+      "logps/rejected": -296.2100524902344,
+      "loss": 0.4248,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.31707775592803955,
+      "rewards/margins": 1.1236298084259033,
+      "rewards/rejected": -0.8065520524978638,
+      "step": 1160
+    },
+    {
+      "epoch": 2.4496205181889557,
+      "grad_norm": 30.813022693471254,
+      "learning_rate": 9.808972011828054e-08,
+      "logits/chosen": -0.07085756957530975,
+      "logits/rejected": -0.1990344077348709,
+      "logps/chosen": -342.7474060058594,
+      "logps/rejected": -266.599609375,
+      "loss": 0.4366,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": 0.2785649299621582,
+      "rewards/margins": 1.1454359292984009,
+      "rewards/rejected": -0.8668708801269531,
+      "step": 1170
+    },
+    {
+      "epoch": 2.470557445694844,
+      "grad_norm": 28.23769321193845,
+      "learning_rate": 9.094970576786032e-08,
+      "logits/chosen": 0.02408856526017189,
+      "logits/rejected": -0.0395381897687912,
+      "logps/chosen": -267.3003234863281,
+      "logps/rejected": -249.65158081054688,
+      "loss": 0.4323,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.17695440351963043,
+      "rewards/margins": 1.0320888757705688,
+      "rewards/rejected": -0.8551343679428101,
+      "step": 1180
+    },
+    {
+      "epoch": 2.491494373200733,
+      "grad_norm": 30.361270294186127,
+      "learning_rate": 8.405341513622055e-08,
+      "logits/chosen": -0.16236011683940887,
+      "logits/rejected": -0.20971974730491638,
+      "logps/chosen": -324.1960754394531,
+      "logps/rejected": -260.82647705078125,
+      "loss": 0.4368,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": 0.3019522428512573,
+      "rewards/margins": 0.9704391360282898,
+      "rewards/rejected": -0.6684868931770325,
+      "step": 1190
+    },
+    {
+      "epoch": 2.5124313007066212,
+      "grad_norm": 29.87892753836197,
+      "learning_rate": 7.740495722810269e-08,
+      "logits/chosen": -0.1393619179725647,
+      "logits/rejected": -0.17410680651664734,
+      "logps/chosen": -275.013427734375,
+      "logps/rejected": -218.1080322265625,
+      "loss": 0.4462,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": 0.1310921609401703,
+      "rewards/margins": 0.8609299659729004,
+      "rewards/rejected": -0.7298377156257629,
+      "step": 1200
+    },
+    {
+      "epoch": 2.5124313007066212,
+      "eval_logits/chosen": -0.34315019845962524,
+      "eval_logits/rejected": -0.37096092104911804,
+      "eval_logps/chosen": -310.3815612792969,
+      "eval_logps/rejected": -275.54351806640625,
+      "eval_loss": 0.589375913143158,
+      "eval_rewards/accuracies": 0.7023809552192688,
+      "eval_rewards/chosen": -0.04855651780962944,
+      "eval_rewards/margins": 0.5814378261566162,
+      "eval_rewards/rejected": -0.6299943923950195,
+      "eval_runtime": 19.1944,
+      "eval_samples_per_second": 104.197,
+      "eval_steps_per_second": 3.282,
+      "step": 1200
+    },
+    {
+      "epoch": 2.5333682282125096,
+      "grad_norm": 30.713812818544117,
+      "learning_rate": 7.100829338251146e-08,
+      "logits/chosen": -0.10057993978261948,
+      "logits/rejected": -0.16287049651145935,
+      "logps/chosen": -328.2908935546875,
+      "logps/rejected": -289.701904296875,
+      "loss": 0.4416,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": 0.3149394094944,
+      "rewards/margins": 1.1269749402999878,
+      "rewards/rejected": -0.8120356798171997,
+      "step": 1210
+    },
+    {
+      "epoch": 2.5543051557183984,
+      "grad_norm": 32.721460368023834,
+      "learning_rate": 6.486723491243778e-08,
+      "logits/chosen": -0.14739738404750824,
+      "logits/rejected": -0.1756991297006607,
+      "logps/chosen": -287.8624572753906,
+      "logps/rejected": -264.6856994628906,
+      "loss": 0.4523,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": 0.20502778887748718,
+      "rewards/margins": 0.9769600033760071,
+      "rewards/rejected": -0.7719322443008423,
+      "step": 1220
+    },
+    {
+      "epoch": 2.575242083224287,
+      "grad_norm": 32.50553549993069,
+      "learning_rate": 5.898544083397e-08,
+      "logits/chosen": -0.12704332172870636,
+      "logits/rejected": -0.1398163139820099,
+      "logps/chosen": -319.8119201660156,
+      "logps/rejected": -287.4822082519531,
+      "loss": 0.4467,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.2838557958602905,
+      "rewards/margins": 1.0057270526885986,
+      "rewards/rejected": -0.7218713164329529,
+      "step": 1230
+    },
+    {
+      "epoch": 2.596179010730175,
+      "grad_norm": 30.599102169827287,
+      "learning_rate": 5.3366415686149137e-08,
+      "logits/chosen": -0.1681247055530548,
+      "logits/rejected": -0.15492555499076843,
+      "logps/chosen": -248.78054809570312,
+      "logps/rejected": -238.68106079101562,
+      "loss": 0.4368,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.10712571442127228,
+      "rewards/margins": 0.8596469759941101,
+      "rewards/rejected": -0.7525211572647095,
+      "step": 1240
+    },
+    {
+      "epoch": 2.617115938236064,
+      "grad_norm": 31.156979603329955,
+      "learning_rate": 4.8013507442865585e-08,
+      "logits/chosen": -0.19022394716739655,
+      "logits/rejected": -0.10595899820327759,
+      "logps/chosen": -271.9454345703125,
+      "logps/rejected": -274.1131286621094,
+      "loss": 0.4319,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.17164357006549835,
+      "rewards/margins": 0.7940610647201538,
+      "rewards/rejected": -0.6224175095558167,
+      "step": 1250
+    },
+    {
+      "epoch": 2.6380528657419524,
+      "grad_norm": 28.881054799580607,
+      "learning_rate": 4.292990551804171e-08,
+      "logits/chosen": -0.08757440745830536,
+      "logits/rejected": -0.08349265158176422,
+      "logps/chosen": -287.8446350097656,
+      "logps/rejected": -251.55636596679688,
+      "loss": 0.4229,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": 0.27148956060409546,
+      "rewards/margins": 1.0949037075042725,
+      "rewards/rejected": -0.8234142065048218,
+      "step": 1260
+    },
+    {
+      "epoch": 2.658989793247841,
+      "grad_norm": 27.816379754479705,
+      "learning_rate": 3.811863886528882e-08,
+      "logits/chosen": -0.05129946395754814,
+      "logits/rejected": -0.07520854473114014,
+      "logps/chosen": -320.88092041015625,
+      "logps/rejected": -259.69488525390625,
+      "loss": 0.4403,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.1268620789051056,
+      "rewards/margins": 0.9729422330856323,
+      "rewards/rejected": -0.8460801839828491,
+      "step": 1270
+    },
+    {
+      "epoch": 2.6799267207537296,
+      "grad_norm": 31.45496593116063,
+      "learning_rate": 3.358257417317095e-08,
+      "logits/chosen": -0.08449853211641312,
+      "logits/rejected": -0.1029936671257019,
+      "logps/chosen": -288.6929016113281,
+      "logps/rejected": -251.7803192138672,
+      "loss": 0.4427,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.18886741995811462,
+      "rewards/margins": 0.9686278104782104,
+      "rewards/rejected": -0.779760479927063,
+      "step": 1280
+    },
+    {
+      "epoch": 2.700863648259618,
+      "grad_norm": 29.618711306226224,
+      "learning_rate": 2.9324414157151367e-08,
+      "logits/chosen": -0.1431390345096588,
+      "logits/rejected": -0.10319207608699799,
+      "logps/chosen": -265.52508544921875,
+      "logps/rejected": -260.0169677734375,
+      "loss": 0.4661,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.10996869951486588,
+      "rewards/margins": 0.7998045682907104,
+      "rewards/rejected": -0.6898358464241028,
+      "step": 1290
+    },
+    {
+      "epoch": 2.7218005757655064,
+      "grad_norm": 30.57983222127276,
+      "learning_rate": 2.5346695949237717e-08,
+      "logits/chosen": -0.1215812936425209,
+      "logits/rejected": -0.1575727015733719,
+      "logps/chosen": -285.28839111328125,
+      "logps/rejected": -258.9991455078125,
+      "loss": 0.4312,
+      "rewards/accuracies": 0.8062499761581421,
+      "rewards/chosen": 0.16850288212299347,
+      "rewards/margins": 0.923494815826416,
+      "rewards/rejected": -0.7549919486045837,
+      "step": 1300
+    },
+    {
+      "epoch": 2.7218005757655064,
+      "eval_logits/chosen": -0.3442156910896301,
+      "eval_logits/rejected": -0.37236616015434265,
+      "eval_logps/chosen": -310.4347229003906,
+      "eval_logps/rejected": -275.5621337890625,
+      "eval_loss": 0.5860841870307922,
+      "eval_rewards/accuracies": 0.6666666865348816,
+      "eval_rewards/chosen": -0.07512283325195312,
+      "eval_rewards/margins": 0.5641648769378662,
+      "eval_rewards/rejected": -0.6392877101898193,
+      "eval_runtime": 18.9961,
+      "eval_samples_per_second": 105.285,
+      "eval_steps_per_second": 3.316,
+      "step": 1300
+    },
+    {
+      "epoch": 2.7427375032713948,
+      "grad_norm": 33.466317771235126,
+      "learning_rate": 2.165178958628744e-08,
+      "logits/chosen": -0.2976987659931183,
+      "logits/rejected": -0.2995850145816803,
+      "logps/chosen": -273.274658203125,
+      "logps/rejected": -265.376953125,
+      "loss": 0.4557,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": 0.248153418302536,
+      "rewards/margins": 0.9688236117362976,
+      "rewards/rejected": -0.720670223236084,
+      "step": 1310
+    },
+    {
+      "epoch": 2.7636744307772836,
+      "grad_norm": 34.66631251738303,
+      "learning_rate": 1.824189659787284e-08,
+      "logits/chosen": -0.10840437561273575,
+      "logits/rejected": -0.2128714621067047,
+      "logps/chosen": -305.4066162109375,
+      "logps/rejected": -265.4433288574219,
+      "loss": 0.4569,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.2427886724472046,
+      "rewards/margins": 0.9924157857894897,
+      "rewards/rejected": -0.7496271133422852,
+      "step": 1320
+    },
+    {
+      "epoch": 2.784611358283172,
+      "grad_norm": 32.12579125137935,
+      "learning_rate": 1.511904869454772e-08,
+      "logits/chosen": -0.2719363570213318,
+      "logits/rejected": -0.282664954662323,
+      "logps/chosen": -258.68927001953125,
+      "logps/rejected": -247.0824737548828,
+      "loss": 0.4442,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.14507904648780823,
+      "rewards/margins": 0.9792217016220093,
+      "rewards/rejected": -0.8341425657272339,
+      "step": 1330
+    },
+    {
+      "epoch": 2.8055482857890603,
+      "grad_norm": 30.777443115008786,
+      "learning_rate": 1.2285106557296476e-08,
+      "logits/chosen": -0.005404007621109486,
+      "logits/rejected": -0.11788828670978546,
+      "logps/chosen": -291.2290954589844,
+      "logps/rejected": -259.48468017578125,
+      "loss": 0.4343,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": 0.14346763491630554,
+      "rewards/margins": 0.9186245203018188,
+      "rewards/rejected": -0.7751568555831909,
+      "step": 1340
+    },
+    {
+      "epoch": 2.8264852132949487,
+      "grad_norm": 30.795896785948266,
+      "learning_rate": 9.741758728888217e-09,
+      "logits/chosen": -0.21444204449653625,
+      "logits/rejected": -0.22242781519889832,
+      "logps/chosen": -273.7785949707031,
+      "logps/rejected": -255.1972198486328,
+      "loss": 0.4474,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.1873357743024826,
+      "rewards/margins": 0.925756573677063,
+      "rewards/rejected": -0.7384207844734192,
+      "step": 1350
+    },
+    {
+      "epoch": 2.8474221408008376,
+      "grad_norm": 32.2751606467674,
+      "learning_rate": 7.490520607794981e-09,
+      "logits/chosen": -0.23268666863441467,
+      "logits/rejected": -0.24138864874839783,
+      "logps/chosen": -321.114990234375,
+      "logps/rejected": -286.67596435546875,
+      "loss": 0.4185,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.2826289236545563,
+      "rewards/margins": 1.1657564640045166,
+      "rewards/rejected": -0.8831275701522827,
+      "step": 1360
+    },
+    {
+      "epoch": 2.868359068306726,
+      "grad_norm": 35.16124310393445,
+      "learning_rate": 5.532733545274781e-09,
+      "logits/chosen": -0.15783634781837463,
+      "logits/rejected": -0.04948469251394272,
+      "logps/chosen": -257.70880126953125,
+      "logps/rejected": -257.0361633300781,
+      "loss": 0.4388,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.09304230660200119,
+      "rewards/margins": 0.8344671130180359,
+      "rewards/rejected": -0.7414248585700989,
+      "step": 1370
+    },
+    {
+      "epoch": 2.8892959958126143,
+      "grad_norm": 28.606615842220055,
+      "learning_rate": 3.869564046156459e-09,
+      "logits/chosen": -0.187531977891922,
+      "logits/rejected": -0.3003791868686676,
+      "logps/chosen": -326.945068359375,
+      "logps/rejected": -275.7688903808594,
+      "loss": 0.4474,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.3150181770324707,
+      "rewards/margins": 0.9270998239517212,
+      "rewards/rejected": -0.6120817065238953,
+      "step": 1380
+    },
+    {
+      "epoch": 2.910232923318503,
+      "grad_norm": 29.689707753715854,
+      "learning_rate": 2.5020030738031052e-09,
+      "logits/chosen": -0.06661088764667511,
+      "logits/rejected": -0.03964465484023094,
+      "logps/chosen": -326.2409973144531,
+      "logps/rejected": -333.43048095703125,
+      "loss": 0.4429,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.26736724376678467,
+      "rewards/margins": 1.0932358503341675,
+      "rewards/rejected": -0.8258686065673828,
+      "step": 1390
+    },
+    {
+      "epoch": 2.9311698508243915,
+      "grad_norm": 28.977571013449776,
+      "learning_rate": 1.4308654596684177e-09,
+      "logits/chosen": -0.11069446802139282,
+      "logits/rejected": -0.20247487723827362,
+      "logps/chosen": -306.44610595703125,
+      "logps/rejected": -266.9939270019531,
+      "loss": 0.4454,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": 0.33073872327804565,
+      "rewards/margins": 1.0625050067901611,
+      "rewards/rejected": -0.7317661046981812,
+      "step": 1400
+    },
+    {
+      "epoch": 2.9311698508243915,
+      "eval_logits/chosen": -0.34011754393577576,
+      "eval_logits/rejected": -0.36805158853530884,
+      "eval_logps/chosen": -310.2955627441406,
+      "eval_logps/rejected": -275.4775085449219,
+      "eval_loss": 0.5941523313522339,
+      "eval_rewards/accuracies": 0.6944444179534912,
+      "eval_rewards/chosen": -0.005565390922129154,
+      "eval_rewards/margins": 0.591437041759491,
+      "eval_rewards/rejected": -0.5970024466514587,
+      "eval_runtime": 19.2489,
+      "eval_samples_per_second": 103.902,
+      "eval_steps_per_second": 3.273,
+      "step": 1400
+    },
+    {
+      "epoch": 2.95210677833028,
+      "grad_norm": 29.74874723420356,
+      "learning_rate": 6.567894177967325e-10,
+      "logits/chosen": -0.046352408826351166,
+      "logits/rejected": -0.054122112691402435,
+      "logps/chosen": -287.88287353515625,
+      "logps/rejected": -262.4837646484375,
+      "loss": 0.4161,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.250956654548645,
+      "rewards/margins": 1.026626706123352,
+      "rewards/rejected": -0.775670051574707,
+      "step": 1410
+    },
+    {
+      "epoch": 2.9730437058361687,
+      "grad_norm": 29.04099364011046,
+      "learning_rate": 1.802361645573125e-10,
+      "logits/chosen": -0.15342023968696594,
+      "logits/rejected": -0.1559501588344574,
+      "logps/chosen": -272.7152404785156,
+      "logps/rejected": -269.2229919433594,
+      "loss": 0.4192,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.05591464787721634,
+      "rewards/margins": 1.0562889575958252,
+      "rewards/rejected": -1.000374436378479,
+      "step": 1420
+    },
+    {
+      "epoch": 2.993980633342057,
+      "grad_norm": 30.18988447303673,
+      "learning_rate": 1.4896438384481846e-12,
+      "logits/chosen": -0.20836491882801056,
+      "logits/rejected": -0.19784407317638397,
+      "logps/chosen": -317.96209716796875,
+      "logps/rejected": -302.5794982910156,
+      "loss": 0.4315,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.22425825893878937,
+      "rewards/margins": 0.8794215321540833,
+      "rewards/rejected": -0.6551632285118103,
+      "step": 1430
+    },
+    {
+      "epoch": 2.996074326092646,
+      "step": 1431,
+      "total_flos": 0.0,
+      "train_loss": 0.5334697115221363,
+      "train_runtime": 7355.3343,
+      "train_samples_per_second": 24.935,
+      "train_steps_per_second": 0.195
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1431,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

ComfyUI/models/smol/SmolLM2-1.7B-Instruct/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7649586c424c337f6c403fdb617ac9d954daf9a7192f3afe5b6318f37e9bb19e
+size 6520

ComfyUI/models/smol/SmolLM2-135M-Instruct/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

ComfyUI/models/smol/SmolLM2-135M-Instruct/README.md

@@ -0,0 +1,133 @@
+---
+library_name: transformers
+license: apache-2.0
+language:
+- en
+pipeline_tag: text-generation
+tags:
+- safetensors
+- onnx
+- transformers.js
+base_model:
+- HuggingFaceTB/SmolLM2-135M
+---
+
+
+# SmolLM2
+
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/3ntM63zkmxY2cNRhgY_Kl.png)
+
+##  Table of Contents
+
+1. [Model Summary](##model-summary)
+2. [Limitations](##limitations)
+3. [Training](##training)
+4. [License](##license)
+5. [Citation](##citation)
+
+## Model Summary
+
+SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters. They are capable of solving a wide range of tasks while being lightweight enough to run on-device.
+
+SmolLM2 demonstrates significant advances over its predecessor SmolLM1, particularly in instruction following, knowledge, reasoning. The 135M model was trained on 2 trillion tokens using a diverse dataset combination: FineWeb-Edu, DCLM, The Stack, along with new filtered datasets we curated and will release soon.  We developed the instruct version through supervised fine-tuning (SFT) using a combination of public datasets and our own curated datasets. We then applied Direct Preference Optimization (DPO) using [UltraFeedback](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized).
+
+The instruct model additionally supports tasks such as text rewriting, summarization and function calling (for the 1.7B) thanks to datasets developed by [Argilla](https://huggingface.co/argilla) such as [Synth-APIGen-v0.1](https://huggingface.co/datasets/argilla/Synth-APIGen-v0.1).
+You can find the SFT dataset here: https://huggingface.co/datasets/HuggingFaceTB/smol-smoltalk and finetuning code at https://github.com/huggingface/alignment-handbook/tree/main/recipes/smollm2
+
+### How to use
+
+### Transformers
+```bash
+pip install transformers
+```
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+checkpoint = "HuggingFaceTB/SmolLM2-135M-Instruct"
+
+device = "cuda" # for GPU usage or "cpu" for CPU usage
+tokenizer = AutoTokenizer.from_pretrained(checkpoint)
+# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
+model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
+
+messages = [{"role": "user", "content": "What is gravity?"}]
+input_text=tokenizer.apply_chat_template(messages, tokenize=False)
+print(input_text)
+inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
+print(tokenizer.decode(outputs[0]))
+```
+
+### Chat in TRL
+You can also use the TRL CLI to chat with the model from the terminal:
+```bash
+pip install trl
+trl chat --model_name_or_path HuggingFaceTB/SmolLM2-135M-Instruct --device cpu
+```
+
+## Evaluation
+
+In this section, we report the evaluation results of SmolLM2. All evaluations are zero-shot unless stated otherwise, and we use [lighteval](https://github.com/huggingface/lighteval) to run them.
+
+## Base pre-trained model
+
+| Metrics            | SmolLM2-135M-8k | SmolLM-135M  |
+|:-------------------|:----------------:|:------------:|
+| HellaSwag         | **42.1**         | 41.2         |
+| ARC (Average)     | **43.9**         | 42.4         |
+| PIQA              | 68.4             | 68.4         |
+| MMLU (cloze)      | **31.5**         | 30.2         |
+| CommonsenseQA     | **33.9**         | 32.7         |
+| TriviaQA          | 4.1              | **4.3**      |
+| Winogrande        | 51.3             | 51.3         |
+| OpenBookQA        | **34.6**         | 34.0         |
+| GSM8K (5-shot)    | **1.4**          | 1.0          |
+
+
+## Instruction model
+
+| Metric                       | SmolLM2-135M-Instruct | SmolLM-135M-Instruct |
+|:-----------------------------|:---------------------:|:--------------------:|
+| IFEval (Average prompt/inst) | **29.9**                 | 17.2                |
+| MT-Bench                     | **19.8**                 | 16.8                |
+| HellaSwag                    | **40.9**                 | 38.9                |
+| ARC (Average)                | **37.3**                 | 33.9                |
+| PIQA                         | **66.3**                 | 64.0                |
+| MMLU (cloze)                 | **29.3**                 | 28.3                |
+| BBH (3-shot)                 | **28.2**                 | 25.2                |
+| GSM8K (5-shot)               | 1.4                  | 1.4                 |
+
+
+
+## Limitations
+
+SmolLM2 models primarily understand and generate content in English. They can produce text on a variety of topics, but the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data. These models should be used as assistive tools rather than definitive sources of information. Users should always verify important information and critically evaluate any generated content.
+
+## Training
+
+### Model
+
+- **Architecture:** Transformer decoder
+- **Pretraining tokens:** 2T
+- **Precision:** bfloat16
+
+### Hardware
+
+- **GPUs:** 64 H100
+
+### Software
+
+- **Training Framework:** [nanotron](https://github.com/huggingface/nanotron/tree/main)
+
+## License
+
+[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
+
+## Citation
+```bash
+@misc{allal2024SmolLM2,
+      title={SmolLM2 - with great data, comes great performance}, 
+      author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Gabriel Martín Blázquez and Lewis Tunstall and Agustín Piqueres and Andres Marafioti and Cyril Zakka and Leandro von Werra and Thomas Wolf},
+      year={2024},
+}
+```

ComfyUI/models/smol/SmolLM2-135M-Instruct/all_results.json

@@ -0,0 +1,22 @@
+{
+    "epoch": 1.9973828840617638,
+    "eval_logits/chosen": 4.600930213928223,
+    "eval_logits/rejected": 4.9520039558410645,
+    "eval_logps/chosen": -443.648193359375,
+    "eval_logps/rejected": -378.15826416015625,
+    "eval_loss": 0.6740825176239014,
+    "eval_rewards/accuracies": 0.6150793433189392,
+    "eval_rewards/chosen": -0.07192634046077728,
+    "eval_rewards/margins": 0.26874542236328125,
+    "eval_rewards/rejected": -0.34067174792289734,
+    "eval_runtime": 20.4479,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 97.81,
+    "eval_steps_per_second": 3.081,
+    "total_flos": 0.0,
+    "train_loss": 0.675485389037702,
+    "train_runtime": 5897.7907,
+    "train_samples": 61134,
+    "train_samples_per_second": 20.731,
+    "train_steps_per_second": 0.162
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/config.json

@@ -0,0 +1,37 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 576,
+  "initializer_range": 0.041666666666666664,
+  "intermediate_size": 1536,
+  "is_llama_config": true,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 9,
+  "num_hidden_layers": 30,
+  "num_key_value_heads": 3,
+  "pad_token_id": 2,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_interleaved": false,
+  "rope_scaling": null,
+  "rope_theta": 100000,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.42.3",
+  "transformers.js_config": {
+    "kv_cache_dtype": {
+      "q4f16": "float16",
+      "fp16": "float16"
+    }
+  },
+  "use_cache": true,
+  "vocab_size": 49152
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 1.9973828840617638,
+    "eval_logits/chosen": 4.600930213928223,
+    "eval_logits/rejected": 4.9520039558410645,
+    "eval_logps/chosen": -443.648193359375,
+    "eval_logps/rejected": -378.15826416015625,
+    "eval_loss": 0.6740825176239014,
+    "eval_rewards/accuracies": 0.6150793433189392,
+    "eval_rewards/chosen": -0.07192634046077728,
+    "eval_rewards/margins": 0.26874542236328125,
+    "eval_rewards/rejected": -0.34067174792289734,
+    "eval_runtime": 20.4479,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 97.81,
+    "eval_steps_per_second": 3.081
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 2,
+  "transformers_version": "4.42.3"
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5af571cbf074e6d21a03528d2330792e532ca608f24ac70a143f6b369968ab8c
+size 269060552

ComfyUI/models/smol/SmolLM2-135M-Instruct/runs/Oct31_10-14-22_ip-26-0-174-36/events.out.tfevents.1730370128.ip-26-0-174-36.3239327.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:774faae98f36baa1ee525b594023c5fc25b4ad9249e04ccdf4cba9ee4f4585fd
+size 78157

ComfyUI/models/smol/SmolLM2-135M-Instruct/runs/Oct31_10-14-22_ip-26-0-174-36/events.out.tfevents.1730376056.ip-26-0-174-36.3239327.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:043a35029ecb401ef0df75148304075329cd444faf2097b0316e9200018fea5c
+size 828

ComfyUI/models/smol/SmolLM2-135M-Instruct/special_tokens_map.json

@@ -0,0 +1,34 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": {
+    "content": "<|im_start|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/tokenizer_config.json

@@ -0,0 +1,154 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<repo_name>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<file_sep>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<jupyter_script>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": "<|im_start|>",
+  "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful AI assistant named SmolLM, trained by Hugging Face<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "model_max_length": 2048,
+  "pad_token": "<|im_end|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 1.9973828840617638,
+    "total_flos": 0.0,
+    "train_loss": 0.675485389037702,
+    "train_runtime": 5897.7907,
+    "train_samples": 61134,
+    "train_samples_per_second": 20.731,
+    "train_steps_per_second": 0.162
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/trainer_state.json

@@ -0,0 +1,1626 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9973828840617638,
+  "eval_steps": 100,
+  "global_step": 954,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.002093692750588851,
+      "grad_norm": 38.333577570579074,
+      "learning_rate": 1.0416666666666666e-08,
+      "logits/chosen": 5.468747138977051,
+      "logits/rejected": 5.353150367736816,
+      "logps/chosen": -399.0700988769531,
+      "logps/rejected": -414.2703857421875,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.02093692750588851,
+      "grad_norm": 36.45097456473781,
+      "learning_rate": 1.0416666666666667e-07,
+      "logits/chosen": 4.634159088134766,
+      "logits/rejected": 4.8650617599487305,
+      "logps/chosen": -481.9865417480469,
+      "logps/rejected": -402.9172668457031,
+      "loss": 0.7192,
+      "rewards/accuracies": 0.4236111044883728,
+      "rewards/chosen": 0.036201052367687225,
+      "rewards/margins": 0.05521820858120918,
+      "rewards/rejected": -0.01901715248823166,
+      "step": 10
+    },
+    {
+      "epoch": 0.04187385501177702,
+      "grad_norm": 38.31461130932718,
+      "learning_rate": 2.0833333333333333e-07,
+      "logits/chosen": 4.8017449378967285,
+      "logits/rejected": 5.193596363067627,
+      "logps/chosen": -428.74591064453125,
+      "logps/rejected": -379.7098693847656,
+      "loss": 0.7525,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.014457901008427143,
+      "rewards/margins": -0.03727109357714653,
+      "rewards/rejected": 0.0517289862036705,
+      "step": 20
+    },
+    {
+      "epoch": 0.06281078251766553,
+      "grad_norm": 38.39033659648525,
+      "learning_rate": 3.1249999999999997e-07,
+      "logits/chosen": 4.625308513641357,
+      "logits/rejected": 4.913487434387207,
+      "logps/chosen": -459.8934631347656,
+      "logps/rejected": -365.87176513671875,
+      "loss": 0.7389,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": 0.032367587089538574,
+      "rewards/margins": 0.07106685638427734,
+      "rewards/rejected": -0.03869926929473877,
+      "step": 30
+    },
+    {
+      "epoch": 0.08374771002355404,
+      "grad_norm": 34.959297165636315,
+      "learning_rate": 4.1666666666666667e-07,
+      "logits/chosen": 5.135643005371094,
+      "logits/rejected": 5.29467248916626,
+      "logps/chosen": -388.5003662109375,
+      "logps/rejected": -341.11138916015625,
+      "loss": 0.7521,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.061884332448244095,
+      "rewards/margins": -0.051958512514829636,
+      "rewards/rejected": -0.009925814345479012,
+      "step": 40
+    },
+    {
+      "epoch": 0.10468463752944256,
+      "grad_norm": 40.59746593571697,
+      "learning_rate": 5.208333333333334e-07,
+      "logits/chosen": 4.794947147369385,
+      "logits/rejected": 5.206262111663818,
+      "logps/chosen": -418.7637634277344,
+      "logps/rejected": -366.21783447265625,
+      "loss": 0.7539,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -0.0046131848357617855,
+      "rewards/margins": 0.009517465718090534,
+      "rewards/rejected": -0.014130651950836182,
+      "step": 50
+    },
+    {
+      "epoch": 0.12562156503533106,
+      "grad_norm": 35.116262927070615,
+      "learning_rate": 6.249999999999999e-07,
+      "logits/chosen": 4.984349250793457,
+      "logits/rejected": 5.210784435272217,
+      "logps/chosen": -389.5479431152344,
+      "logps/rejected": -355.3258361816406,
+      "loss": 0.7337,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": 0.09027661383152008,
+      "rewards/margins": 0.13924987614154816,
+      "rewards/rejected": -0.048973266035318375,
+      "step": 60
+    },
+    {
+      "epoch": 0.14655849254121958,
+      "grad_norm": 37.65630163274615,
+      "learning_rate": 7.291666666666666e-07,
+      "logits/chosen": 5.079476356506348,
+      "logits/rejected": 5.1062331199646,
+      "logps/chosen": -472.6788635253906,
+      "logps/rejected": -410.6566467285156,
+      "loss": 0.7532,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": 0.01137494295835495,
+      "rewards/margins": -0.019245151430368423,
+      "rewards/rejected": 0.030620098114013672,
+      "step": 70
+    },
+    {
+      "epoch": 0.16749542004710807,
+      "grad_norm": 36.35209638887961,
+      "learning_rate": 8.333333333333333e-07,
+      "logits/chosen": 4.831971645355225,
+      "logits/rejected": 5.179555892944336,
+      "logps/chosen": -465.8661193847656,
+      "logps/rejected": -352.46063232421875,
+      "loss": 0.7337,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.029347699135541916,
+      "rewards/margins": 0.011580700054764748,
+      "rewards/rejected": 0.017767000943422318,
+      "step": 80
+    },
+    {
+      "epoch": 0.1884323475529966,
+      "grad_norm": 40.247342074541066,
+      "learning_rate": 9.374999999999999e-07,
+      "logits/chosen": 4.667853355407715,
+      "logits/rejected": 5.083367347717285,
+      "logps/chosen": -410.145263671875,
+      "logps/rejected": -347.24871826171875,
+      "loss": 0.7325,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.04706032946705818,
+      "rewards/margins": 0.05596155673265457,
+      "rewards/rejected": -0.008901228196918964,
+      "step": 90
+    },
+    {
+      "epoch": 0.2093692750588851,
+      "grad_norm": 36.88448509348576,
+      "learning_rate": 9.999463737538052e-07,
+      "logits/chosen": 5.017066955566406,
+      "logits/rejected": 5.157826900482178,
+      "logps/chosen": -453.6114196777344,
+      "logps/rejected": -376.13214111328125,
+      "loss": 0.7296,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": 0.03123999759554863,
+      "rewards/margins": 0.038888636976480484,
+      "rewards/rejected": -0.007648637983947992,
+      "step": 100
+    },
+    {
+      "epoch": 0.2093692750588851,
+      "eval_logits/chosen": 4.755386829376221,
+      "eval_logits/rejected": 5.127224445343018,
+      "eval_logps/chosen": -443.48101806640625,
+      "eval_logps/rejected": -377.5273742675781,
+      "eval_loss": 0.7357296347618103,
+      "eval_rewards/accuracies": 0.5515872836112976,
+      "eval_rewards/chosen": 0.01168334111571312,
+      "eval_rewards/margins": 0.03692733868956566,
+      "eval_rewards/rejected": -0.025244001299142838,
+      "eval_runtime": 21.3186,
+      "eval_samples_per_second": 93.815,
+      "eval_steps_per_second": 2.955,
+      "step": 100
+    },
+    {
+      "epoch": 0.23030620256477363,
+      "grad_norm": 36.21924654761057,
+      "learning_rate": 9.993432105822034e-07,
+      "logits/chosen": 4.768385887145996,
+      "logits/rejected": 5.076653957366943,
+      "logps/chosen": -449.16375732421875,
+      "logps/rejected": -369.2919006347656,
+      "loss": 0.7211,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": -0.04003235697746277,
+      "rewards/margins": -0.059906214475631714,
+      "rewards/rejected": 0.019873863086104393,
+      "step": 110
+    },
+    {
+      "epoch": 0.2512431300706621,
+      "grad_norm": 36.12599439727971,
+      "learning_rate": 9.980706626858607e-07,
+      "logits/chosen": 5.0697174072265625,
+      "logits/rejected": 5.350961208343506,
+      "logps/chosen": -392.9084777832031,
+      "logps/rejected": -342.9964294433594,
+      "loss": 0.7213,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": 0.005513651762157679,
+      "rewards/margins": 0.07320307195186615,
+      "rewards/rejected": -0.06768941879272461,
+      "step": 120
+    },
+    {
+      "epoch": 0.2721800575765506,
+      "grad_norm": 43.81008737879554,
+      "learning_rate": 9.961304359538434e-07,
+      "logits/chosen": 4.7396440505981445,
+      "logits/rejected": 5.110291957855225,
+      "logps/chosen": -445.08209228515625,
+      "logps/rejected": -356.9689636230469,
+      "loss": 0.7319,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": 0.03245898336172104,
+      "rewards/margins": 0.0837341919541359,
+      "rewards/rejected": -0.05127520486712456,
+      "step": 130
+    },
+    {
+      "epoch": 0.29311698508243916,
+      "grad_norm": 37.12091478913465,
+      "learning_rate": 9.935251313189563e-07,
+      "logits/chosen": 4.5339274406433105,
+      "logits/rejected": 5.020459175109863,
+      "logps/chosen": -473.4126892089844,
+      "logps/rejected": -364.12939453125,
+      "loss": 0.7193,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": 0.008073748089373112,
+      "rewards/margins": 0.035515300929546356,
+      "rewards/rejected": -0.02744155190885067,
+      "step": 140
+    },
+    {
+      "epoch": 0.31405391258832765,
+      "grad_norm": 36.18564480374988,
+      "learning_rate": 9.902582412711118e-07,
+      "logits/chosen": 4.540812969207764,
+      "logits/rejected": 4.964258193969727,
+      "logps/chosen": -426.5033264160156,
+      "logps/rejected": -353.1463317871094,
+      "loss": 0.7232,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.01480414904654026,
+      "rewards/margins": 0.06241898611187935,
+      "rewards/rejected": -0.07722313702106476,
+      "step": 150
+    },
+    {
+      "epoch": 0.33499084009421615,
+      "grad_norm": 32.90233228487631,
+      "learning_rate": 9.86334145175542e-07,
+      "logits/chosen": 4.807779788970947,
+      "logits/rejected": 5.042156219482422,
+      "logps/chosen": -396.0440673828125,
+      "logps/rejected": -360.52886962890625,
+      "loss": 0.7013,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.010227044112980366,
+      "rewards/margins": 0.055318038910627365,
+      "rewards/rejected": -0.045090995728969574,
+      "step": 160
+    },
+    {
+      "epoch": 0.3559277676001047,
+      "grad_norm": 37.311964741290105,
+      "learning_rate": 9.817581034021272e-07,
+      "logits/chosen": 4.897703170776367,
+      "logits/rejected": 5.062272071838379,
+      "logps/chosen": -389.55810546875,
+      "logps/rejected": -329.748779296875,
+      "loss": 0.7043,
+      "rewards/accuracies": 0.518750011920929,
+      "rewards/chosen": -0.008063828572630882,
+      "rewards/margins": 0.049024712294340134,
+      "rewards/rejected": -0.05708853527903557,
+      "step": 170
+    },
+    {
+      "epoch": 0.3768646951059932,
+      "grad_norm": 34.14102924438106,
+      "learning_rate": 9.765362502737097e-07,
+      "logits/chosen": 5.039429187774658,
+      "logits/rejected": 5.049492835998535,
+      "logps/chosen": -384.9471130371094,
+      "logps/rejected": -381.6601257324219,
+      "loss": 0.7091,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": 0.004380516707897186,
+      "rewards/margins": 0.038515396416187286,
+      "rewards/rejected": -0.0341348834335804,
+      "step": 180
+    },
+    {
+      "epoch": 0.39780162261188173,
+      "grad_norm": 35.88568083270778,
+      "learning_rate": 9.706755858428485e-07,
+      "logits/chosen": 5.025214195251465,
+      "logits/rejected": 5.097342014312744,
+      "logps/chosen": -397.56402587890625,
+      "logps/rejected": -396.12799072265625,
+      "loss": 0.7161,
+      "rewards/accuracies": 0.518750011920929,
+      "rewards/chosen": -0.056682147085666656,
+      "rewards/margins": 0.023842817172408104,
+      "rewards/rejected": -0.08052496612071991,
+      "step": 190
+    },
+    {
+      "epoch": 0.4187385501177702,
+      "grad_norm": 38.60485416145746,
+      "learning_rate": 9.641839665080363e-07,
+      "logits/chosen": 5.1590471267700195,
+      "logits/rejected": 5.290652275085449,
+      "logps/chosen": -399.9522399902344,
+      "logps/rejected": -363.53936767578125,
+      "loss": 0.7062,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": 0.04297986626625061,
+      "rewards/margins": 0.06266864389181137,
+      "rewards/rejected": -0.019688773900270462,
+      "step": 200
+    },
+    {
+      "epoch": 0.4187385501177702,
+      "eval_logits/chosen": 4.725487232208252,
+      "eval_logits/rejected": 5.087867736816406,
+      "eval_logps/chosen": -443.55450439453125,
+      "eval_logps/rejected": -377.6705627441406,
+      "eval_loss": 0.6988219022750854,
+      "eval_rewards/accuracies": 0.567460298538208,
+      "eval_rewards/chosen": -0.025081120431423187,
+      "eval_rewards/margins": 0.07174728065729141,
+      "eval_rewards/rejected": -0.0968284010887146,
+      "eval_runtime": 21.5315,
+      "eval_samples_per_second": 92.887,
+      "eval_steps_per_second": 2.926,
+      "step": 200
+    },
+    {
+      "epoch": 0.4396754776236587,
+      "grad_norm": 36.10684145537435,
+      "learning_rate": 9.570700944819582e-07,
+      "logits/chosen": 4.827897548675537,
+      "logits/rejected": 5.154609680175781,
+      "logps/chosen": -451.2969665527344,
+      "logps/rejected": -372.116455078125,
+      "loss": 0.7141,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": -0.07117662578821182,
+      "rewards/margins": 0.02444976009428501,
+      "rewards/rejected": -0.09562637656927109,
+      "step": 210
+    },
+    {
+      "epoch": 0.46061240512954726,
+      "grad_norm": 33.13953697631782,
+      "learning_rate": 9.493435061259129e-07,
+      "logits/chosen": 5.24191427230835,
+      "logits/rejected": 5.477172374725342,
+      "logps/chosen": -365.3572692871094,
+      "logps/rejected": -345.34814453125,
+      "loss": 0.7138,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": -0.07684006541967392,
+      "rewards/margins": -0.033464811742305756,
+      "rewards/rejected": -0.043375253677368164,
+      "step": 220
+    },
+    {
+      "epoch": 0.48154933263543576,
+      "grad_norm": 36.46869252662201,
+      "learning_rate": 9.4101455916603e-07,
+      "logits/chosen": 4.996638298034668,
+      "logits/rejected": 5.203185081481934,
+      "logps/chosen": -390.2169494628906,
+      "logps/rejected": -381.74090576171875,
+      "loss": 0.7027,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -0.006761978380382061,
+      "rewards/margins": 0.11298926174640656,
+      "rewards/rejected": -0.11975125223398209,
+      "step": 230
+    },
+    {
+      "epoch": 0.5024862601413242,
+      "grad_norm": 32.465417090707724,
+      "learning_rate": 9.320944188084241e-07,
+      "logits/chosen": 4.961588382720947,
+      "logits/rejected": 5.104936122894287,
+      "logps/chosen": -405.73651123046875,
+      "logps/rejected": -368.39312744140625,
+      "loss": 0.6966,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.04590672254562378,
+      "rewards/margins": 0.21882851421833038,
+      "rewards/rejected": -0.1729217916727066,
+      "step": 240
+    },
+    {
+      "epoch": 0.5234231876472127,
+      "grad_norm": 32.56446548910322,
+      "learning_rate": 9.225950427718974e-07,
+      "logits/chosen": 4.295259475708008,
+      "logits/rejected": 4.731950759887695,
+      "logps/chosen": -457.6085510253906,
+      "logps/rejected": -378.92083740234375,
+      "loss": 0.6895,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -0.08796132355928421,
+      "rewards/margins": 0.07854396849870682,
+      "rewards/rejected": -0.16650527715682983,
+      "step": 250
+    },
+    {
+      "epoch": 0.5443601151531012,
+      "grad_norm": 37.3555609307497,
+      "learning_rate": 9.125291652582547e-07,
+      "logits/chosen": 4.772681713104248,
+      "logits/rejected": 4.774602890014648,
+      "logps/chosen": -429.554931640625,
+      "logps/rejected": -350.5638122558594,
+      "loss": 0.697,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.017600687220692635,
+      "rewards/margins": 0.15435068309307098,
+      "rewards/rejected": -0.1367500126361847,
+      "step": 260
+    },
+    {
+      "epoch": 0.5652970426589898,
+      "grad_norm": 33.42195670184768,
+      "learning_rate": 9.019102798817195e-07,
+      "logits/chosen": 4.580355644226074,
+      "logits/rejected": 5.011557102203369,
+      "logps/chosen": -446.68438720703125,
+      "logps/rejected": -380.5400390625,
+      "loss": 0.6778,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.019351882860064507,
+      "rewards/margins": 0.15824225544929504,
+      "rewards/rejected": -0.13889038562774658,
+      "step": 270
+    },
+    {
+      "epoch": 0.5862339701648783,
+      "grad_norm": 39.119947805234894,
+      "learning_rate": 8.90752621580335e-07,
+      "logits/chosen": 5.025314807891846,
+      "logits/rejected": 5.18468713760376,
+      "logps/chosen": -424.27191162109375,
+      "logps/rejected": -344.4115295410156,
+      "loss": 0.7073,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.023305395618081093,
+      "rewards/margins": 0.14927226305007935,
+      "rewards/rejected": -0.125966876745224,
+      "step": 280
+    },
+    {
+      "epoch": 0.6071708976707668,
+      "grad_norm": 33.472133866310024,
+      "learning_rate": 8.79071147533597e-07,
+      "logits/chosen": 4.961835861206055,
+      "logits/rejected": 5.123082637786865,
+      "logps/chosen": -400.4245300292969,
+      "logps/rejected": -388.8963623046875,
+      "loss": 0.6687,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": 0.08449111878871918,
+      "rewards/margins": 0.2874522805213928,
+      "rewards/rejected": -0.20296116173267365,
+      "step": 290
+    },
+    {
+      "epoch": 0.6281078251766553,
+      "grad_norm": 38.29458361274397,
+      "learning_rate": 8.668815171119019e-07,
+      "logits/chosen": 4.6071085929870605,
+      "logits/rejected": 4.85768985748291,
+      "logps/chosen": -445.59393310546875,
+      "logps/rejected": -373.83636474609375,
+      "loss": 0.6782,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": 0.043323811143636703,
+      "rewards/margins": 0.2540794312953949,
+      "rewards/rejected": -0.2107556313276291,
+      "step": 300
+    },
+    {
+      "epoch": 0.6281078251766553,
+      "eval_logits/chosen": 4.662118434906006,
+      "eval_logits/rejected": 5.016141414642334,
+      "eval_logps/chosen": -443.56884765625,
+      "eval_logps/rejected": -377.883056640625,
+      "eval_loss": 0.6942777037620544,
+      "eval_rewards/accuracies": 0.567460298538208,
+      "eval_rewards/chosen": -0.03225937858223915,
+      "eval_rewards/margins": 0.17080551385879517,
+      "eval_rewards/rejected": -0.20306488871574402,
+      "eval_runtime": 21.6344,
+      "eval_samples_per_second": 92.445,
+      "eval_steps_per_second": 2.912,
+      "step": 300
+    },
+    {
+      "epoch": 0.6490447526825438,
+      "grad_norm": 35.277636776310345,
+      "learning_rate": 8.54200070884685e-07,
+      "logits/chosen": 4.7398271560668945,
+      "logits/rejected": 5.0438690185546875,
+      "logps/chosen": -455.08074951171875,
+      "logps/rejected": -346.21905517578125,
+      "loss": 0.6648,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.00532907247543335,
+      "rewards/margins": 0.22818481922149658,
+      "rewards/rejected": -0.23351387679576874,
+      "step": 310
+    },
+    {
+      "epoch": 0.6699816801884323,
+      "grad_norm": 36.1242464612453,
+      "learning_rate": 8.410438087153911e-07,
+      "logits/chosen": 4.823008060455322,
+      "logits/rejected": 4.949624538421631,
+      "logps/chosen": -420.04150390625,
+      "logps/rejected": -346.31134033203125,
+      "loss": 0.6633,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.02057427167892456,
+      "rewards/margins": 0.1925923228263855,
+      "rewards/rejected": -0.21316656470298767,
+      "step": 320
+    },
+    {
+      "epoch": 0.6909186076943209,
+      "grad_norm": 34.485635067716444,
+      "learning_rate": 8.274303669726426e-07,
+      "logits/chosen": 4.866278171539307,
+      "logits/rejected": 5.084838390350342,
+      "logps/chosen": -413.07647705078125,
+      "logps/rejected": -359.72637939453125,
+      "loss": 0.6964,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": -0.0661710649728775,
+      "rewards/margins": 0.11275775730609894,
+      "rewards/rejected": -0.17892882227897644,
+      "step": 330
+    },
+    {
+      "epoch": 0.7118555352002094,
+      "grad_norm": 36.021204090397475,
+      "learning_rate": 8.133779948881513e-07,
+      "logits/chosen": 4.962647914886475,
+      "logits/rejected": 5.274256229400635,
+      "logps/chosen": -423.34796142578125,
+      "logps/rejected": -374.83831787109375,
+      "loss": 0.6843,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.06308840215206146,
+      "rewards/margins": 0.14286582171916962,
+      "rewards/rejected": -0.20595422387123108,
+      "step": 340
+    },
+    {
+      "epoch": 0.7327924627060979,
+      "grad_norm": 40.80633953486743,
+      "learning_rate": 7.989055300930704e-07,
+      "logits/chosen": 4.9410552978515625,
+      "logits/rejected": 5.171365737915039,
+      "logps/chosen": -401.3800048828125,
+      "logps/rejected": -339.8207702636719,
+      "loss": 0.6799,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.05132729932665825,
+      "rewards/margins": 0.14917483925819397,
+      "rewards/rejected": -0.20050212740898132,
+      "step": 350
+    },
+    {
+      "epoch": 0.7537293902119864,
+      "grad_norm": 32.290367261199215,
+      "learning_rate": 7.840323733655778e-07,
+      "logits/chosen": 4.760105609893799,
+      "logits/rejected": 4.936800956726074,
+      "logps/chosen": -475.94305419921875,
+      "logps/rejected": -373.4317626953125,
+      "loss": 0.6723,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.040363796055316925,
+      "rewards/margins": 0.26596716046333313,
+      "rewards/rejected": -0.2256033718585968,
+      "step": 360
+    },
+    {
+      "epoch": 0.7746663177178749,
+      "grad_norm": 33.44911142948228,
+      "learning_rate": 7.687784626235447e-07,
+      "logits/chosen": 4.649796485900879,
+      "logits/rejected": 4.882054328918457,
+      "logps/chosen": -437.54791259765625,
+      "logps/rejected": -343.1330871582031,
+      "loss": 0.6722,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.03584844991564751,
+      "rewards/margins": 0.2080194056034088,
+      "rewards/rejected": -0.2438678741455078,
+      "step": 370
+    },
+    {
+      "epoch": 0.7956032452237635,
+      "grad_norm": 32.64019199720913,
+      "learning_rate": 7.531642461971514e-07,
+      "logits/chosen": 4.7331953048706055,
+      "logits/rejected": 5.047934532165527,
+      "logps/chosen": -434.0751953125,
+      "logps/rejected": -363.1179504394531,
+      "loss": 0.673,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.03724004700779915,
+      "rewards/margins": 0.23209133744239807,
+      "rewards/rejected": -0.2693313956260681,
+      "step": 380
+    },
+    {
+      "epoch": 0.816540172729652,
+      "grad_norm": 33.38515796622506,
+      "learning_rate": 7.372106554172801e-07,
+      "logits/chosen": 4.660643577575684,
+      "logits/rejected": 4.7719621658325195,
+      "logps/chosen": -434.41015625,
+      "logps/rejected": -394.9471130371094,
+      "loss": 0.6778,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": 0.021881069988012314,
+      "rewards/margins": 0.2415298968553543,
+      "rewards/rejected": -0.2196488082408905,
+      "step": 390
+    },
+    {
+      "epoch": 0.8374771002355405,
+      "grad_norm": 32.60957530709497,
+      "learning_rate": 7.209390765564318e-07,
+      "logits/chosen": 4.807684421539307,
+      "logits/rejected": 5.217709541320801,
+      "logps/chosen": -368.07122802734375,
+      "logps/rejected": -328.12066650390625,
+      "loss": 0.6863,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -0.03271085396409035,
+      "rewards/margins": 0.1988353729248047,
+      "rewards/rejected": -0.23154623806476593,
+      "step": 400
+    },
+    {
+      "epoch": 0.8374771002355405,
+      "eval_logits/chosen": 4.645900249481201,
+      "eval_logits/rejected": 4.999230861663818,
+      "eval_logps/chosen": -443.68084716796875,
+      "eval_logps/rejected": -378.0348205566406,
+      "eval_loss": 0.6756832003593445,
+      "eval_rewards/accuracies": 0.5992063283920288,
+      "eval_rewards/chosen": -0.08822782337665558,
+      "eval_rewards/margins": 0.19070643186569214,
+      "eval_rewards/rejected": -0.2789342403411865,
+      "eval_runtime": 21.4973,
+      "eval_samples_per_second": 93.035,
+      "eval_steps_per_second": 2.931,
+      "step": 400
+    },
+    {
+      "epoch": 0.8584140277414289,
+      "grad_norm": 80.32344495768098,
+      "learning_rate": 7.043713221597773e-07,
+      "logits/chosen": 4.9558234214782715,
+      "logits/rejected": 5.171336650848389,
+      "logps/chosen": -464.4634704589844,
+      "logps/rejected": -378.52130126953125,
+      "loss": 0.6691,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": -0.028427015990018845,
+      "rewards/margins": 0.16248683631420135,
+      "rewards/rejected": -0.1909138560295105,
+      "step": 410
+    },
+    {
+      "epoch": 0.8793509552473174,
+      "grad_norm": 38.003974257278905,
+      "learning_rate": 6.875296018047809e-07,
+      "logits/chosen": 5.062918663024902,
+      "logits/rejected": 5.093894958496094,
+      "logps/chosen": -414.597900390625,
+      "logps/rejected": -392.76422119140625,
+      "loss": 0.6778,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.05205658823251724,
+      "rewards/margins": 0.166357159614563,
+      "rewards/rejected": -0.21841374039649963,
+      "step": 420
+    },
+    {
+      "epoch": 0.9002878827532059,
+      "grad_norm": 33.39326538286459,
+      "learning_rate": 6.704364923285857e-07,
+      "logits/chosen": 4.783626556396484,
+      "logits/rejected": 5.061443328857422,
+      "logps/chosen": -454.7694396972656,
+      "logps/rejected": -349.71099853515625,
+      "loss": 0.6613,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.03129550814628601,
+      "rewards/margins": 0.2672887146472931,
+      "rewards/rejected": -0.2985842227935791,
+      "step": 430
+    },
+    {
+      "epoch": 0.9212248102590945,
+      "grad_norm": 33.88594593881104,
+      "learning_rate": 6.531149075630796e-07,
+      "logits/chosen": 4.762629985809326,
+      "logits/rejected": 4.992688179016113,
+      "logps/chosen": -422.49639892578125,
+      "logps/rejected": -342.6626892089844,
+      "loss": 0.6829,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -0.08267354220151901,
+      "rewards/margins": 0.18921074271202087,
+      "rewards/rejected": -0.2718842923641205,
+      "step": 440
+    },
+    {
+      "epoch": 0.942161737764983,
+      "grad_norm": 34.194378360359096,
+      "learning_rate": 6.355880676182085e-07,
+      "logits/chosen": 4.86130952835083,
+      "logits/rejected": 5.088041305541992,
+      "logps/chosen": -423.82366943359375,
+      "logps/rejected": -386.20172119140625,
+      "loss": 0.6777,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.029518108814954758,
+      "rewards/margins": 0.2812921106815338,
+      "rewards/rejected": -0.3108102083206177,
+      "step": 450
+    },
+    {
+      "epoch": 0.9630986652708715,
+      "grad_norm": 35.220161431379815,
+      "learning_rate": 6.178794677547137e-07,
+      "logits/chosen": 4.96859073638916,
+      "logits/rejected": 5.295912265777588,
+      "logps/chosen": -408.28228759765625,
+      "logps/rejected": -337.6819763183594,
+      "loss": 0.6573,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.02989841438829899,
+      "rewards/margins": 0.3131854832172394,
+      "rewards/rejected": -0.343083918094635,
+      "step": 460
+    },
+    {
+      "epoch": 0.98403559277676,
+      "grad_norm": 36.11741005068747,
+      "learning_rate": 6.000128468880222e-07,
+      "logits/chosen": 4.616504669189453,
+      "logits/rejected": 4.935946464538574,
+      "logps/chosen": -435.3017578125,
+      "logps/rejected": -375.13800048828125,
+      "loss": 0.6647,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.11524273455142975,
+      "rewards/margins": 0.2546766698360443,
+      "rewards/rejected": -0.3699193596839905,
+      "step": 470
+    },
+    {
+      "epoch": 1.0049725202826485,
+      "grad_norm": 32.471857091487834,
+      "learning_rate": 5.820121557655108e-07,
+      "logits/chosen": 4.9493536949157715,
+      "logits/rejected": 5.226868152618408,
+      "logps/chosen": -423.6285095214844,
+      "logps/rejected": -362.1949768066406,
+      "loss": 0.6629,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.05261852219700813,
+      "rewards/margins": 0.3280298113822937,
+      "rewards/rejected": -0.27541130781173706,
+      "step": 480
+    },
+    {
+      "epoch": 1.025909447788537,
+      "grad_norm": 39.51652905408664,
+      "learning_rate": 5.639015248598023e-07,
+      "logits/chosen": 4.762259006500244,
+      "logits/rejected": 5.021244525909424,
+      "logps/chosen": -424.96697998046875,
+      "logps/rejected": -342.76666259765625,
+      "loss": 0.6644,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.03716667741537094,
+      "rewards/margins": 0.2011403739452362,
+      "rewards/rejected": -0.23830704391002655,
+      "step": 490
+    },
+    {
+      "epoch": 1.0468463752944255,
+      "grad_norm": 34.43579926142672,
+      "learning_rate": 5.457052320211339e-07,
+      "logits/chosen": 4.543593406677246,
+      "logits/rejected": 4.786489009857178,
+      "logps/chosen": -434.46746826171875,
+      "logps/rejected": -367.75689697265625,
+      "loss": 0.6836,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.08059108257293701,
+      "rewards/margins": 0.23346829414367676,
+      "rewards/rejected": -0.3140593469142914,
+      "step": 500
+    },
+    {
+      "epoch": 1.0468463752944255,
+      "eval_logits/chosen": 4.617003440856934,
+      "eval_logits/rejected": 4.9695563316345215,
+      "eval_logps/chosen": -443.69580078125,
+      "eval_logps/rejected": -378.1418762207031,
+      "eval_loss": 0.6708300113677979,
+      "eval_rewards/accuracies": 0.6349206566810608,
+      "eval_rewards/chosen": -0.09571509808301926,
+      "eval_rewards/margins": 0.23677198588848114,
+      "eval_rewards/rejected": -0.3324871063232422,
+      "eval_runtime": 20.9626,
+      "eval_samples_per_second": 95.408,
+      "eval_steps_per_second": 3.005,
+      "step": 500
+    },
+    {
+      "epoch": 1.067783302800314,
+      "grad_norm": 33.3201987416808,
+      "learning_rate": 5.274476699321637e-07,
+      "logits/chosen": 4.583409786224365,
+      "logits/rejected": 4.803020477294922,
+      "logps/chosen": -390.48565673828125,
+      "logps/rejected": -351.6776123046875,
+      "loss": 0.6779,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.0327589213848114,
+      "rewards/margins": 0.32544368505477905,
+      "rewards/rejected": -0.35820263624191284,
+      "step": 510
+    },
+    {
+      "epoch": 1.0887202303062025,
+      "grad_norm": 30.742730814185435,
+      "learning_rate": 5.091533134088387e-07,
+      "logits/chosen": 4.493949890136719,
+      "logits/rejected": 4.9839911460876465,
+      "logps/chosen": -383.7958984375,
+      "logps/rejected": -354.36480712890625,
+      "loss": 0.656,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -0.07341745495796204,
+      "rewards/margins": 0.19709812104701996,
+      "rewards/rejected": -0.2705155909061432,
+      "step": 520
+    },
+    {
+      "epoch": 1.109657157812091,
+      "grad_norm": 34.05900047947194,
+      "learning_rate": 4.908466865911614e-07,
+      "logits/chosen": 4.793222904205322,
+      "logits/rejected": 5.078155517578125,
+      "logps/chosen": -401.0002746582031,
+      "logps/rejected": -340.4061279296875,
+      "loss": 0.6618,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.05051114410161972,
+      "rewards/margins": 0.27152642607688904,
+      "rewards/rejected": -0.32203757762908936,
+      "step": 530
+    },
+    {
+      "epoch": 1.1305940853179797,
+      "grad_norm": 30.483486401054424,
+      "learning_rate": 4.7255233006783624e-07,
+      "logits/chosen": 4.857717990875244,
+      "logits/rejected": 5.0497636795043945,
+      "logps/chosen": -375.65362548828125,
+      "logps/rejected": -330.26165771484375,
+      "loss": 0.6544,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.02352207899093628,
+      "rewards/margins": 0.35729408264160156,
+      "rewards/rejected": -0.38081610202789307,
+      "step": 540
+    },
+    {
+      "epoch": 1.151531012823868,
+      "grad_norm": 35.09603470685652,
+      "learning_rate": 4.5429476797886617e-07,
+      "logits/chosen": 4.932369232177734,
+      "logits/rejected": 5.050224781036377,
+      "logps/chosen": -430.0126953125,
+      "logps/rejected": -331.1691589355469,
+      "loss": 0.6599,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.0060789333656430244,
+      "rewards/margins": 0.25288811326026917,
+      "rewards/rejected": -0.25896701216697693,
+      "step": 550
+    },
+    {
+      "epoch": 1.1724679403297567,
+      "grad_norm": 40.64422646125966,
+      "learning_rate": 4.3609847514019763e-07,
+      "logits/chosen": 4.637743949890137,
+      "logits/rejected": 5.000674724578857,
+      "logps/chosen": -420.3258361816406,
+      "logps/rejected": -362.2751159667969,
+      "loss": 0.6718,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -0.0094971414655447,
+      "rewards/margins": 0.22678783535957336,
+      "rewards/rejected": -0.23628497123718262,
+      "step": 560
+    },
+    {
+      "epoch": 1.193404867835645,
+      "grad_norm": 32.638009640148645,
+      "learning_rate": 4.179878442344892e-07,
+      "logits/chosen": 4.855754375457764,
+      "logits/rejected": 4.871184349060059,
+      "logps/chosen": -384.08660888671875,
+      "logps/rejected": -371.4262390136719,
+      "loss": 0.6766,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.05095939710736275,
+      "rewards/margins": 0.28148993849754333,
+      "rewards/rejected": -0.332449346780777,
+      "step": 570
+    },
+    {
+      "epoch": 1.2143417953415336,
+      "grad_norm": 35.519971577107064,
+      "learning_rate": 3.9998715311197783e-07,
+      "logits/chosen": 4.73850679397583,
+      "logits/rejected": 5.173120021820068,
+      "logps/chosen": -414.8775329589844,
+      "logps/rejected": -341.5818786621094,
+      "loss": 0.6508,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.09668377041816711,
+      "rewards/margins": 0.25211262702941895,
+      "rewards/rejected": -0.34879642724990845,
+      "step": 580
+    },
+    {
+      "epoch": 1.235278722847422,
+      "grad_norm": 34.20580765627037,
+      "learning_rate": 3.821205322452863e-07,
+      "logits/chosen": 4.916988372802734,
+      "logits/rejected": 5.1998610496521,
+      "logps/chosen": -448.5626525878906,
+      "logps/rejected": -367.84027099609375,
+      "loss": 0.644,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.07886572182178497,
+      "rewards/margins": 0.3578983247280121,
+      "rewards/rejected": -0.43676406145095825,
+      "step": 590
+    },
+    {
+      "epoch": 1.2562156503533106,
+      "grad_norm": 33.854286929995176,
+      "learning_rate": 3.6441193238179146e-07,
+      "logits/chosen": 4.852269649505615,
+      "logits/rejected": 4.903324127197266,
+      "logps/chosen": -446.4149475097656,
+      "logps/rejected": -423.3356018066406,
+      "loss": 0.6349,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.14010193943977356,
+      "rewards/margins": 0.15067996084690094,
+      "rewards/rejected": -0.2907818853855133,
+      "step": 600
+    },
+    {
+      "epoch": 1.2562156503533106,
+      "eval_logits/chosen": 4.62031364440918,
+      "eval_logits/rejected": 4.9707465171813965,
+      "eval_logps/chosen": -443.61212158203125,
+      "eval_logps/rejected": -378.1197204589844,
+      "eval_loss": 0.6720485091209412,
+      "eval_rewards/accuracies": 0.5992063283920288,
+      "eval_rewards/chosen": -0.053870752453804016,
+      "eval_rewards/margins": 0.267531156539917,
+      "eval_rewards/rejected": -0.321401983499527,
+      "eval_runtime": 20.8046,
+      "eval_samples_per_second": 96.133,
+      "eval_steps_per_second": 3.028,
+      "step": 600
+    },
+    {
+      "epoch": 1.2771525778591992,
+      "grad_norm": 36.085842973391074,
+      "learning_rate": 3.4688509243692034e-07,
+      "logits/chosen": 4.767918586730957,
+      "logits/rejected": 4.757430553436279,
+      "logps/chosen": -407.41668701171875,
+      "logps/rejected": -317.3873596191406,
+      "loss": 0.6402,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.08183420449495316,
+      "rewards/margins": 0.33883604407310486,
+      "rewards/rejected": -0.42067021131515503,
+      "step": 610
+    },
+    {
+      "epoch": 1.2980895053650876,
+      "grad_norm": 29.698333183198105,
+      "learning_rate": 3.295635076714144e-07,
+      "logits/chosen": 5.085806846618652,
+      "logits/rejected": 5.415268898010254,
+      "logps/chosen": -395.627685546875,
+      "logps/rejected": -331.7653503417969,
+      "loss": 0.6266,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.09060301631689072,
+      "rewards/margins": 0.3094441294670105,
+      "rewards/rejected": -0.4000471234321594,
+      "step": 620
+    },
+    {
+      "epoch": 1.3190264328709762,
+      "grad_norm": 35.208773349468885,
+      "learning_rate": 3.12470398195219e-07,
+      "logits/chosen": 4.828533172607422,
+      "logits/rejected": 4.925856113433838,
+      "logps/chosen": -418.5848083496094,
+      "logps/rejected": -376.3353576660156,
+      "loss": 0.6486,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": 0.06391973793506622,
+      "rewards/margins": 0.44674786925315857,
+      "rewards/rejected": -0.3828281760215759,
+      "step": 630
+    },
+    {
+      "epoch": 1.3399633603768648,
+      "grad_norm": 29.673309842493335,
+      "learning_rate": 2.956286778402226e-07,
+      "logits/chosen": 4.896113872528076,
+      "logits/rejected": 5.183098793029785,
+      "logps/chosen": -394.4980773925781,
+      "logps/rejected": -374.76422119140625,
+      "loss": 0.6394,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.04267222806811333,
+      "rewards/margins": 0.2913575768470764,
+      "rewards/rejected": -0.33402982354164124,
+      "step": 640
+    },
+    {
+      "epoch": 1.3609002878827532,
+      "grad_norm": 35.03684415648848,
+      "learning_rate": 2.7906092344356826e-07,
+      "logits/chosen": 4.610795021057129,
+      "logits/rejected": 4.8373188972473145,
+      "logps/chosen": -379.4288024902344,
+      "logps/rejected": -345.05596923828125,
+      "loss": 0.6646,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.1311950385570526,
+      "rewards/margins": 0.2357216328382492,
+      "rewards/rejected": -0.366916686296463,
+      "step": 650
+    },
+    {
+      "epoch": 1.3818372153886418,
+      "grad_norm": 33.06984951084542,
+      "learning_rate": 2.6278934458271996e-07,
+      "logits/chosen": 4.830328941345215,
+      "logits/rejected": 5.017812252044678,
+      "logps/chosen": -377.4278564453125,
+      "logps/rejected": -343.86529541015625,
+      "loss": 0.6613,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.10503290593624115,
+      "rewards/margins": 0.11762680858373642,
+      "rewards/rejected": -0.22265975177288055,
+      "step": 660
+    },
+    {
+      "epoch": 1.4027741428945302,
+      "grad_norm": 31.761556922593446,
+      "learning_rate": 2.468357538028487e-07,
+      "logits/chosen": 4.728631496429443,
+      "logits/rejected": 4.90619421005249,
+      "logps/chosen": -413.2724609375,
+      "logps/rejected": -346.9877624511719,
+      "loss": 0.6393,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.03663766756653786,
+      "rewards/margins": 0.2855125069618225,
+      "rewards/rejected": -0.32215017080307007,
+      "step": 670
+    },
+    {
+      "epoch": 1.4237110704004188,
+      "grad_norm": 34.93162849349177,
+      "learning_rate": 2.312215373764551e-07,
+      "logits/chosen": 4.728277206420898,
+      "logits/rejected": 5.018845558166504,
+      "logps/chosen": -421.8961486816406,
+      "logps/rejected": -403.57354736328125,
+      "loss": 0.6533,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.1260446161031723,
+      "rewards/margins": 0.2200162708759308,
+      "rewards/rejected": -0.3460609018802643,
+      "step": 680
+    },
+    {
+      "epoch": 1.4446479979063072,
+      "grad_norm": 33.66523822793528,
+      "learning_rate": 2.1596762663442213e-07,
+      "logits/chosen": 4.863284111022949,
+      "logits/rejected": 4.840500354766846,
+      "logps/chosen": -422.4331970214844,
+      "logps/rejected": -355.96868896484375,
+      "loss": 0.6477,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.10065089166164398,
+      "rewards/margins": 0.24511468410491943,
+      "rewards/rejected": -0.3457655906677246,
+      "step": 690
+    },
+    {
+      "epoch": 1.4655849254121958,
+      "grad_norm": 34.48257400044076,
+      "learning_rate": 2.0109446990692963e-07,
+      "logits/chosen": 4.709015846252441,
+      "logits/rejected": 4.914425849914551,
+      "logps/chosen": -452.9461364746094,
+      "logps/rejected": -442.56658935546875,
+      "loss": 0.6427,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.08032918721437454,
+      "rewards/margins": 0.39584842324256897,
+      "rewards/rejected": -0.3155192732810974,
+      "step": 700
+    },
+    {
+      "epoch": 1.4655849254121958,
+      "eval_logits/chosen": 4.592012882232666,
+      "eval_logits/rejected": 4.943046569824219,
+      "eval_logps/chosen": -443.6796875,
+      "eval_logps/rejected": -378.1680908203125,
+      "eval_loss": 0.6795812845230103,
+      "eval_rewards/accuracies": 0.60317462682724,
+      "eval_rewards/chosen": -0.08766676485538483,
+      "eval_rewards/margins": 0.25792673230171204,
+      "eval_rewards/rejected": -0.34559354186058044,
+      "eval_runtime": 21.1978,
+      "eval_samples_per_second": 94.349,
+      "eval_steps_per_second": 2.972,
+      "step": 700
+    },
+    {
+      "epoch": 1.4865218529180844,
+      "grad_norm": 36.350524448045036,
+      "learning_rate": 1.8662200511184872e-07,
+      "logits/chosen": 4.871232509613037,
+      "logits/rejected": 4.886293411254883,
+      "logps/chosen": -417.8133850097656,
+      "logps/rejected": -384.177490234375,
+      "loss": 0.6669,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.12399878352880478,
+      "rewards/margins": 0.27855515480041504,
+      "rewards/rejected": -0.40255388617515564,
+      "step": 710
+    },
+    {
+      "epoch": 1.5074587804239727,
+      "grad_norm": 34.52058371975813,
+      "learning_rate": 1.725696330273575e-07,
+      "logits/chosen": 4.8079633712768555,
+      "logits/rejected": 5.118483543395996,
+      "logps/chosen": -433.02032470703125,
+      "logps/rejected": -383.21539306640625,
+      "loss": 0.6234,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.049599818885326385,
+      "rewards/margins": 0.36388009786605835,
+      "rewards/rejected": -0.31428030133247375,
+      "step": 720
+    },
+    {
+      "epoch": 1.5283957079298613,
+      "grad_norm": 36.62094520000859,
+      "learning_rate": 1.589561912846089e-07,
+      "logits/chosen": 4.67967414855957,
+      "logits/rejected": 4.974714756011963,
+      "logps/chosen": -402.2828063964844,
+      "logps/rejected": -343.87939453125,
+      "loss": 0.6419,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.02172028087079525,
+      "rewards/margins": 0.3966042995452881,
+      "rewards/rejected": -0.4183245599269867,
+      "step": 730
+    },
+    {
+      "epoch": 1.54933263543575,
+      "grad_norm": 34.85140828972076,
+      "learning_rate": 1.4579992911531496e-07,
+      "logits/chosen": 4.999066352844238,
+      "logits/rejected": 5.089913845062256,
+      "logps/chosen": -442.08538818359375,
+      "logps/rejected": -387.76953125,
+      "loss": 0.6641,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.024145543575286865,
+      "rewards/margins": 0.3119828999042511,
+      "rewards/rejected": -0.28783735632896423,
+      "step": 740
+    },
+    {
+      "epoch": 1.5702695629416383,
+      "grad_norm": 33.55559408410901,
+      "learning_rate": 1.3311848288809813e-07,
+      "logits/chosen": 4.944571018218994,
+      "logits/rejected": 4.949963569641113,
+      "logps/chosen": -422.9165954589844,
+      "logps/rejected": -378.2356262207031,
+      "loss": 0.6431,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.10791780799627304,
+      "rewards/margins": 0.16808216273784637,
+      "rewards/rejected": -0.2759999632835388,
+      "step": 750
+    },
+    {
+      "epoch": 1.5912064904475267,
+      "grad_norm": 33.284252993746314,
+      "learning_rate": 1.209288524664029e-07,
+      "logits/chosen": 4.269396781921387,
+      "logits/rejected": 4.640176296234131,
+      "logps/chosen": -513.432861328125,
+      "logps/rejected": -464.742431640625,
+      "loss": 0.6505,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": 0.00313050439581275,
+      "rewards/margins": 0.3427557051181793,
+      "rewards/rejected": -0.33962517976760864,
+      "step": 760
+    },
+    {
+      "epoch": 1.6121434179534153,
+      "grad_norm": 33.301123590813035,
+      "learning_rate": 1.0924737841966497e-07,
+      "logits/chosen": 4.588865756988525,
+      "logits/rejected": 4.75103235244751,
+      "logps/chosen": -465.42059326171875,
+      "logps/rejected": -370.064697265625,
+      "loss": 0.653,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.01498096901923418,
+      "rewards/margins": 0.34728002548217773,
+      "rewards/rejected": -0.3622610569000244,
+      "step": 770
+    },
+    {
+      "epoch": 1.633080345459304,
+      "grad_norm": 33.707974100314466,
+      "learning_rate": 9.808972011828054e-08,
+      "logits/chosen": 4.657374382019043,
+      "logits/rejected": 5.004950523376465,
+      "logps/chosen": -452.0787048339844,
+      "logps/rejected": -383.26824951171875,
+      "loss": 0.6419,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.06604544818401337,
+      "rewards/margins": 0.4590230882167816,
+      "rewards/rejected": -0.39297762513160706,
+      "step": 780
+    },
+    {
+      "epoch": 1.6540172729651923,
+      "grad_norm": 36.400512256730096,
+      "learning_rate": 8.747083474174527e-08,
+      "logits/chosen": 4.775164604187012,
+      "logits/rejected": 5.237417221069336,
+      "logps/chosen": -431.0052185058594,
+      "logps/rejected": -372.1168212890625,
+      "loss": 0.6398,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.017582783475518227,
+      "rewards/margins": 0.35258156061172485,
+      "rewards/rejected": -0.37016433477401733,
+      "step": 790
+    },
+    {
+      "epoch": 1.674954200471081,
+      "grad_norm": 29.96252260731642,
+      "learning_rate": 7.740495722810269e-08,
+      "logits/chosen": 4.998331546783447,
+      "logits/rejected": 4.909043312072754,
+      "logps/chosen": -489.783447265625,
+      "logps/rejected": -415.0606384277344,
+      "loss": 0.6128,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": 0.04877934604883194,
+      "rewards/margins": 0.3542923033237457,
+      "rewards/rejected": -0.3055129647254944,
+      "step": 800
+    },
+    {
+      "epoch": 1.674954200471081,
+      "eval_logits/chosen": 4.6106181144714355,
+      "eval_logits/rejected": 4.968925476074219,
+      "eval_logps/chosen": -443.625244140625,
+      "eval_logps/rejected": -378.2127990722656,
+      "eval_loss": 0.6703739166259766,
+      "eval_rewards/accuracies": 0.6071428656578064,
+      "eval_rewards/chosen": -0.06042463704943657,
+      "eval_rewards/margins": 0.30752548575401306,
+      "eval_rewards/rejected": -0.3679501414299011,
+      "eval_runtime": 21.1621,
+      "eval_samples_per_second": 94.509,
+      "eval_steps_per_second": 2.977,
+      "step": 800
+    },
+    {
+      "epoch": 1.6958911279769695,
+      "grad_norm": 34.07376933070953,
+      "learning_rate": 6.790558119157597e-08,
+      "logits/chosen": 4.842529773712158,
+      "logits/rejected": 4.945174217224121,
+      "logps/chosen": -446.68682861328125,
+      "logps/rejected": -379.9209899902344,
+      "loss": 0.6409,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": 0.010895573534071445,
+      "rewards/margins": 0.47363200783729553,
+      "rewards/rejected": -0.46273642778396606,
+      "step": 810
+    },
+    {
+      "epoch": 1.7168280554828579,
+      "grad_norm": 31.41564508164701,
+      "learning_rate": 5.898544083397e-08,
+      "logits/chosen": 4.57013463973999,
+      "logits/rejected": 4.8762030601501465,
+      "logps/chosen": -459.298583984375,
+      "logps/rejected": -376.189208984375,
+      "loss": 0.6381,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.04928427189588547,
+      "rewards/margins": 0.33450883626937866,
+      "rewards/rejected": -0.38379308581352234,
+      "step": 820
+    },
+    {
+      "epoch": 1.7377649829887463,
+      "grad_norm": 38.55984096337612,
+      "learning_rate": 5.065649387408705e-08,
+      "logits/chosen": 4.863150596618652,
+      "logits/rejected": 4.996617317199707,
+      "logps/chosen": -405.2935485839844,
+      "logps/rejected": -383.06756591796875,
+      "loss": 0.6587,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.14060600101947784,
+      "rewards/margins": 0.1646648645401001,
+      "rewards/rejected": -0.30527088046073914,
+      "step": 830
+    },
+    {
+      "epoch": 1.7587019104946349,
+      "grad_norm": 32.69891650352482,
+      "learning_rate": 4.292990551804171e-08,
+      "logits/chosen": 4.561503887176514,
+      "logits/rejected": 4.661375522613525,
+      "logps/chosen": -374.9468688964844,
+      "logps/rejected": -359.5188293457031,
+      "loss": 0.6426,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.05613694339990616,
+      "rewards/margins": 0.3448534607887268,
+      "rewards/rejected": -0.40099042654037476,
+      "step": 840
+    },
+    {
+      "epoch": 1.7796388380005235,
+      "grad_norm": 32.82316724445512,
+      "learning_rate": 3.581603349196371e-08,
+      "logits/chosen": 4.668177604675293,
+      "logits/rejected": 5.044764518737793,
+      "logps/chosen": -391.29534912109375,
+      "logps/rejected": -374.1195068359375,
+      "loss": 0.6501,
+      "rewards/accuracies": 0.543749988079071,
+      "rewards/chosen": -0.14754648506641388,
+      "rewards/margins": 0.16757197678089142,
+      "rewards/rejected": -0.3151184618473053,
+      "step": 850
+    },
+    {
+      "epoch": 1.8005757655064119,
+      "grad_norm": 32.36442235611696,
+      "learning_rate": 2.9324414157151367e-08,
+      "logits/chosen": 4.706895351409912,
+      "logits/rejected": 5.021437644958496,
+      "logps/chosen": -417.41021728515625,
+      "logps/rejected": -335.3275451660156,
+      "loss": 0.6534,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.06291428953409195,
+      "rewards/margins": 0.3059840798377991,
+      "rewards/rejected": -0.36889833211898804,
+      "step": 860
+    },
+    {
+      "epoch": 1.8215126930123005,
+      "grad_norm": 29.740377909388123,
+      "learning_rate": 2.3463749726290284e-08,
+      "logits/chosen": 4.696743965148926,
+      "logits/rejected": 4.8797287940979,
+      "logps/chosen": -477.77783203125,
+      "logps/rejected": -390.98175048828125,
+      "loss": 0.6614,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": -0.07511474192142487,
+      "rewards/margins": 0.17798468470573425,
+      "rewards/rejected": -0.2530994415283203,
+      "step": 870
+    },
+    {
+      "epoch": 1.842449620518189,
+      "grad_norm": 30.952090476967147,
+      "learning_rate": 1.824189659787284e-08,
+      "logits/chosen": 4.781184196472168,
+      "logits/rejected": 5.032862663269043,
+      "logps/chosen": -387.22906494140625,
+      "logps/rejected": -360.9486389160156,
+      "loss": 0.6618,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.11400938034057617,
+      "rewards/margins": 0.21471650898456573,
+      "rewards/rejected": -0.3287258744239807,
+      "step": 880
+    },
+    {
+      "epoch": 1.8633865480240774,
+      "grad_norm": 31.64887361264221,
+      "learning_rate": 1.3665854824458035e-08,
+      "logits/chosen": 4.322469234466553,
+      "logits/rejected": 4.672883033752441,
+      "logps/chosen": -445.35699462890625,
+      "logps/rejected": -390.5237731933594,
+      "loss": 0.624,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.032880861312150955,
+      "rewards/margins": 0.3543739914894104,
+      "rewards/rejected": -0.38725486397743225,
+      "step": 890
+    },
+    {
+      "epoch": 1.8843234755299658,
+      "grad_norm": 33.85502008551422,
+      "learning_rate": 9.741758728888217e-09,
+      "logits/chosen": 4.4365644454956055,
+      "logits/rejected": 4.837357997894287,
+      "logps/chosen": -472.887451171875,
+      "logps/rejected": -367.82611083984375,
+      "loss": 0.6474,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.05196143314242363,
+      "rewards/margins": 0.3476230800151825,
+      "rewards/rejected": -0.39958447217941284,
+      "step": 900
+    },
+    {
+      "epoch": 1.8843234755299658,
+      "eval_logits/chosen": 4.5737175941467285,
+      "eval_logits/rejected": 4.921082496643066,
+      "eval_logps/chosen": -443.622314453125,
+      "eval_logps/rejected": -378.2174377441406,
+      "eval_loss": 0.6692253351211548,
+      "eval_rewards/accuracies": 0.6269841194152832,
+      "eval_rewards/chosen": -0.05897674709558487,
+      "eval_rewards/margins": 0.31128397583961487,
+      "eval_rewards/rejected": -0.37026071548461914,
+      "eval_runtime": 21.3225,
+      "eval_samples_per_second": 93.797,
+      "eval_steps_per_second": 2.955,
+      "step": 900
+    },
+    {
+      "epoch": 1.9052604030358546,
+      "grad_norm": 33.28147635399462,
+      "learning_rate": 6.474868681043577e-09,
+      "logits/chosen": 4.713411808013916,
+      "logits/rejected": 4.913935661315918,
+      "logps/chosen": -384.9287109375,
+      "logps/rejected": -316.16265869140625,
+      "loss": 0.6491,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.02645047940313816,
+      "rewards/margins": 0.40163594484329224,
+      "rewards/rejected": -0.42808642983436584,
+      "step": 910
+    },
+    {
+      "epoch": 1.926197330541743,
+      "grad_norm": 35.875215811609834,
+      "learning_rate": 3.869564046156459e-09,
+      "logits/chosen": 4.6749348640441895,
+      "logits/rejected": 4.898279190063477,
+      "logps/chosen": -441.083740234375,
+      "logps/rejected": -361.4406433105469,
+      "loss": 0.6389,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.0031513571739196777,
+      "rewards/margins": 0.41674357652664185,
+      "rewards/rejected": -0.4198949337005615,
+      "step": 920
+    },
+    {
+      "epoch": 1.9471342580476314,
+      "grad_norm": 32.946064523302205,
+      "learning_rate": 1.929337314139412e-09,
+      "logits/chosen": 4.862700462341309,
+      "logits/rejected": 4.817538261413574,
+      "logps/chosen": -429.21051025390625,
+      "logps/rejected": -370.45745849609375,
+      "loss": 0.6312,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0764947384595871,
+      "rewards/margins": 0.20215356349945068,
+      "rewards/rejected": -0.2786482870578766,
+      "step": 930
+    },
+    {
+      "epoch": 1.96807118555352,
+      "grad_norm": 37.53766060677335,
+      "learning_rate": 6.567894177967325e-10,
+      "logits/chosen": 5.056074142456055,
+      "logits/rejected": 5.200203895568848,
+      "logps/chosen": -382.3914489746094,
+      "logps/rejected": -319.7542419433594,
+      "loss": 0.6475,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.012996235862374306,
+      "rewards/margins": 0.2734270989894867,
+      "rewards/rejected": -0.28642335534095764,
+      "step": 940
+    },
+    {
+      "epoch": 1.9890081130594086,
+      "grad_norm": 32.67422145978211,
+      "learning_rate": 5.3626246194704575e-11,
+      "logits/chosen": 4.634739875793457,
+      "logits/rejected": 4.890820503234863,
+      "logps/chosen": -425.7994689941406,
+      "logps/rejected": -344.5509033203125,
+      "loss": 0.6372,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": -0.18660762906074524,
+      "rewards/margins": 0.23620739579200745,
+      "rewards/rejected": -0.4228149950504303,
+      "step": 950
+    },
+    {
+      "epoch": 1.9973828840617638,
+      "step": 954,
+      "total_flos": 0.0,
+      "train_loss": 0.675485389037702,
+      "train_runtime": 5897.7907,
+      "train_samples_per_second": 20.731,
+      "train_steps_per_second": 0.162
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 954,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

ComfyUI/models/smol/SmolLM2-135M-Instruct/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90f508e5769b31070c2c8f82e7ecdce816d610763df89479b5258bc66ee8b357
+size 6520

ComfyUI/models/smol/SmolLM2-360M-Instruct/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

ComfyUI/models/smol/SmolLM2-360M-Instruct/README.md

@@ -0,0 +1,133 @@
+---
+library_name: transformers
+license: apache-2.0
+language:
+- en
+pipeline_tag: text-generation
+tags:
+- safetensors
+- onnx
+- transformers.js
+---
+
+
+# SmolLM2
+
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/oWWfzW4RbWkVIo7f-5444.png)
+
+##  Table of Contents
+
+1. [Model Summary](##model-summary)
+2. [Limitations](##limitations)
+3. [Training](##training)
+4. [License](##license)
+5. [Citation](##citation)
+
+## Model Summary
+
+SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters. They are capable of solving a wide range of tasks while being lightweight enough to run on-device.
+
+SmolLM2 demonstrates significant advances over its predecessor SmolLM1, particularly in instruction following, knowledge, reasoning. The 360M model was trained on 4 trillion tokens using a diverse dataset combination: FineWeb-Edu, DCLM, The Stack, along with new filtered datasets we curated and will release soon.  We developed the instruct version through supervised fine-tuning (SFT) using a combination of public datasets and our own curated datasets. We then applied Direct Preference Optimization (DPO) using [UltraFeedback](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized).
+
+The instruct model additionally supports tasks such as text rewriting, summarization and function calling (for the 1.7B) thanks to datasets developed by [Argilla](https://huggingface.co/argilla) such as [Synth-APIGen-v0.1](https://huggingface.co/datasets/argilla/Synth-APIGen-v0.1).
+You can find the SFT dataset here: https://huggingface.co/datasets/HuggingFaceTB/smol-smoltalk and finetuning code in the [ali] https://github.com/huggingface/alignment-handbook/tree/main/recipes/smollm2
+
+For more details refer to: https://github.com/huggingface/smollm. You will find pre-training, post-training, evaluation and local inference code.
+
+
+### How to use
+
+### Transformers
+```bash
+pip install transformers
+```
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+checkpoint = "HuggingFaceTB/SmolLM2-360M-Instruct"
+
+device = "cuda" # for GPU usage or "cpu" for CPU usage
+tokenizer = AutoTokenizer.from_pretrained(checkpoint)
+# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
+model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
+
+messages = [{"role": "user", "content": "What is the capital of France."}]
+input_text=tokenizer.apply_chat_template(messages, tokenize=False)
+print(input_text)
+inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
+print(tokenizer.decode(outputs[0]))
+```
+
+### Chat in TRL
+You can also use the TRL CLI to chat with the model from the terminal:
+```bash
+pip install trl
+trl chat --model_name_or_path HuggingFaceTB/SmolLM2-360M-Instruct --device cpu
+```
+
+## Evaluation
+
+In this section, we report the evaluation results of SmolLM2. All evaluations are zero-shot unless stated otherwise, and we use [lighteval](https://github.com/huggingface/lighteval) to run them.
+
+## Base Pre-Trained Model
+
+| Metrics            | SmolLM2-360M | Qwen2.5-0.5B | SmolLM-360M  |
+|:-------------------|:------------:|:------------:|:------------:|
+| HellaSwag          | **54.5**     | 51.2         | 51.8         |
+| ARC (Average)      | **53.0**     | 45.4         | 50.1         |
+| PIQA               | **71.7**     | 69.9         | 71.6         |
+| MMLU (cloze)       | **35.8**     | 33.7         | 34.4         |
+| CommonsenseQA      | **38.0**     | 31.6         | 35.3         |
+| TriviaQA           | **16.9**     | 4.3          | 9.1          |
+| Winogrande         | 52.5         | **54.1**     | 52.8         |
+| OpenBookQA         | **37.4**     | **37.4**     | 37.2         |
+| GSM8K (5-shot)     | 3.2          | **33.4**     | 1.6          |
+
+
+## Instruction Model
+
+| Metric                       | SmolLM2-360M-Instruct | Qwen2.5-0.5B-Instruct | SmolLM-360M-Instruct |
+|:-----------------------------|:---------------------:|:---------------------:|:---------------------:|
+| IFEval (Average prompt/inst) | **41.0**             | 31.6                 | 19.8                 |
+| MT-Bench                     | 3.66                 | **4.16**             | 3.37                 |
+| HellaSwag                    | **52.1**             | 48.0                 | 47.9                 |
+| ARC (Average)                | **43.7**             | 37.3                 | 38.8                 |
+| PIQA                         | **70.8**             | 67.2                 | 69.4                 |
+| MMLU (cloze)                 | **32.8**             | 31.7                 | 30.6                 |
+| BBH (3-shot)                 | 27.3                 | **30.7**             | 24.4                 |
+| GSM8K (5-shot)               | 7.43                 | **26.8**             | 1.36                 |
+
+
+## Limitations
+
+SmolLM2 models primarily understand and generate content in English. They can produce text on a variety of topics, but the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data. These models should be used as assistive tools rather than definitive sources of information. Users should always verify important information and critically evaluate any generated content.
+
+## Training
+
+### Model
+
+- **Architecture:** Transformer decoder
+- **Pretraining tokens:** 4T
+- **Precision:** bfloat16
+
+### Hardware
+
+- **GPUs:** 64 H100
+
+### Software
+
+- **Training Framework:** [nanotron](https://github.com/huggingface/nanotron/tree/main)
+
+## License
+
+[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
+
+## Citation
+```bash
+@misc{allal2024SmolLM2,
+      title={SmolLM2 - with great data, comes great performance}, 
+      author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Gabriel Martín Blázquez and Lewis Tunstall and Agustín Piqueres and Andres Marafioti and Cyril Zakka and Leandro von Werra and Thomas Wolf},
+      year={2024},
+}
+```

ComfyUI/models/smol/SmolLM2-360M-Instruct/all_results.json

@@ -0,0 +1,22 @@
+{
+    "epoch": 1.9973828840617638,
+    "eval_logits/chosen": -1.6407532691955566,
+    "eval_logits/rejected": -1.6968854665756226,
+    "eval_logps/chosen": -375.6463623046875,
+    "eval_logps/rejected": -323.7197570800781,
+    "eval_loss": 0.6348475217819214,
+    "eval_rewards/accuracies": 0.6190476417541504,
+    "eval_rewards/chosen": -0.034213583916425705,
+    "eval_rewards/margins": 0.3567626178264618,
+    "eval_rewards/rejected": -0.3909761905670166,
+    "eval_runtime": 22.3598,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 89.446,
+    "eval_steps_per_second": 2.818,
+    "total_flos": 0.0,
+    "train_loss": 0.6516540072998911,
+    "train_runtime": 5944.7081,
+    "train_samples": 61134,
+    "train_samples_per_second": 20.568,
+    "train_steps_per_second": 0.16
+}

ComfyUI/models/smol/SmolLM2-360M-Instruct/config.json

@@ -0,0 +1,37 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 960,
+  "initializer_range": 0.02,
+  "intermediate_size": 2560,
+  "is_llama_config": true,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 15,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 5,
+  "pad_token_id": 2,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_interleaved": false,
+  "rope_scaling": null,
+  "rope_theta": 100000,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.42.3",
+  "transformers.js_config": {
+    "kv_cache_dtype": {
+      "q4f16": "float16",
+      "fp16": "float16"
+    }
+  },
+  "use_cache": true,
+  "vocab_size": 49152
+}

ComfyUI/models/smol/SmolLM2-360M-Instruct/eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 1.9973828840617638,
+    "eval_logits/chosen": -1.6407532691955566,
+    "eval_logits/rejected": -1.6968854665756226,
+    "eval_logps/chosen": -375.6463623046875,
+    "eval_logps/rejected": -323.7197570800781,
+    "eval_loss": 0.6348475217819214,
+    "eval_rewards/accuracies": 0.6190476417541504,
+    "eval_rewards/chosen": -0.034213583916425705,
+    "eval_rewards/margins": 0.3567626178264618,
+    "eval_rewards/rejected": -0.3909761905670166,
+    "eval_runtime": 22.3598,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 89.446,
+    "eval_steps_per_second": 2.818
+}

ComfyUI/models/smol/SmolLM2-360M-Instruct/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 2,
+  "transformers_version": "4.42.3"
+}

ComfyUI/models/smol/SmolLM2-360M-Instruct/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6bffe7435d7ddc10fd3b9a9efd429dafbacb1cb17015fb5562664e7532bf86e
+size 723674912

ComfyUI/models/smol/SmolLM2-360M-Instruct/runs/Oct31_09-01-58_ip-26-0-172-142/events.out.tfevents.1730365788.ip-26-0-172-142.451351.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a59625dee7fb61068e6a6ea53682ee0d91da3f92189019d5ceed597d21ebf27
+size 78139

ComfyUI/models/smol/SmolLM2-360M-Instruct/runs/Oct31_09-01-58_ip-26-0-172-142/events.out.tfevents.1730371773.ip-26-0-172-142.451351.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac5691a0cdfdd383d29b69e85c4aaf632c695ce8fd31683bb9512f310fbf1250
+size 828

ComfyUI/models/smol/SmolLM2-360M-Instruct/runs/Oct31_09-19-57_ip-26-0-174-36/events.out.tfevents.1730366818.ip-26-0-174-36.3233632.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44f4f653d8d22e5db793f6743dc486f1d3f0919a66d5bbd69ebfd22b9e1f598b
+size 38996

ComfyUI/models/smol/SmolLM2-360M-Instruct/runs/Oct31_09-20-43_ip-26-0-161-142/events.out.tfevents.1730366856.ip-26-0-161-142.1301887.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:684365d127b952fdfd215e337f4f8dc5bc4d37ef88fb608632d244ec6f86a449
+size 37621

ComfyUI/models/smol/SmolLM2-360M-Instruct/special_tokens_map.json

@@ -0,0 +1,34 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": {
+    "content": "<|im_start|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}