joseneto023dev
/

Dyck

@@ -1,208 +1,99 @@
 ---
-base_model: unsloth/deepseek-r1-distill-qwen-1.5b-unsloth-bnb-4bit
-library_name: peft
-pipeline_tag: text-generation
 tags:
-- base_model:adapter:unsloth/deepseek-r1-distill-qwen-1.5b-unsloth-bnb-4bit
-- lora
-- transformers
-- unsloth
 ---
-# Model Card for Model ID
-<!-- Provide a quick summary of what the model is/does. -->
-## Model Details
-### Model Description
-<!-- Provide a longer summary of what this model is. -->
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-### Model Sources [optional]
-<!-- Provide the basic links for the model. -->
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-## Uses
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-### Direct Use
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-[More Information Needed]
-### Downstream Use [optional]
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-[More Information Needed]
-### Out-of-Scope Use
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-[More Information Needed]
-## Bias, Risks, and Limitations
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-[More Information Needed]
-### Recommendations
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-## How to Get Started with the Model
-Use the code below to get started with the model.
-[More Information Needed]
 ## Training Details
-### Training Data
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-[More Information Needed]
-### Training Procedure
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-#### Preprocessing [optional]
-[More Information Needed]
-#### Training Hyperparameters
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-#### Speeds, Sizes, Times [optional]
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-[More Information Needed]
-## Evaluation
-<!-- This section describes the evaluation protocols and provides the results. -->
-### Testing Data, Factors & Metrics
-#### Testing Data
-<!-- This should link to a Dataset Card if possible. -->
-[More Information Needed]
-#### Factors
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-[More Information Needed]
-#### Metrics
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-[More Information Needed]
-### Results
-[More Information Needed]
-#### Summary
-## Model Examination [optional]
-<!-- Relevant interpretability work for the model goes here -->
-[More Information Needed]
-## Environmental Impact
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-## Technical Specifications [optional]
-### Model Architecture and Objective
-[More Information Needed]
-### Compute Infrastructure
-[More Information Needed]
-#### Hardware
-[More Information Needed]
-#### Software
-[More Information Needed]
-## Citation [optional]
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-**BibTeX:**
-[More Information Needed]
-**APA:**
-[More Information Needed]
-## Glossary [optional]
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-[More Information Needed]
-## More Information [optional]
-[More Information Needed]
-## Model Card Authors [optional]
-[More Information Needed]
-## Model Card Contact
-[More Information Needed]
-### Framework versions
-- PEFT 0.18.1

 ---
+license: apache-2.0
+base_model: unsloth/DeepSeek-R1-Distill-Qwen-1.5B
 tags:
+  - dyck
+  - reasoning
+  - brackets
+  - fine-tuning
+  - lora
+  - unsloth
+language:
+  - en
+datasets:
+  - conversation.jsonl
+pipeline_tag: text-generation
 ---
+# Dyck Completion Model (Reasoning)
+This model is fine-tuned to **complete Dyck sequences** (balanced bracket sequences) with **step-by-step reasoning**. Given a prefix of opening brackets, it outputs the minimal closing brackets so the full sequence is a valid Dyck word.
+**Response style:** Output follows the **dataset format only** (structured `# Thought N:`, `# Step k: add 'X'.`, then `FINAL ANSWER: <sequence>`). It is not intended to mimic Qwen/DeepSeek-style prose (e.g. no "Wait...", "Let me recount", or conversational commentary). Training and inference prompts enforce this dataset style.
+## Task
+- **Input:** A prefix of opening brackets (e.g. `[ < (`).
+- **Output:** Step-by-step reasoning, then the **complete valid Dyck sequence** (e.g. `) > ]` appended).
+- **Bracket pairs:** `()`, `[]`, `{}`, `<>`
+## Base Model
+- **Architecture:** [DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/unsloth/DeepSeek-R1-Distill-Qwen-1.5B) (Unsloth)
+- **Fine-tuning:** LoRA (r=64, alpha=128, dropout=0.05) on q/k/v/o and MLP projections
+- **Training:** Causal LM; loss on assistant tokens only; format: `{reasoning}\n\nFINAL ANSWER: {full_sequence}`
+## Intended Use
+- Research and education on formal language (Dyck) and chain-of-thought reasoning.
+- Benchmarking reasoning models on bracket completion.
+## How to Use
+**Inference:** Use the **merged model** (single load, base+LoRA already merged) or load base + adapter via PEFT. Merged model = one `AutoModelForCausalLM`; computation is equivalent to base+adapter at every layer.
+### With merged model (this repo, if uploaded as merged)
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+model_id = "YOUR_USERNAME/YOUR_REPO"  # e.g. akashdutta1030/dyck-deepseek-r1-lora
+tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True)
+prompt = """Complete the following Dyck language sequence by adding the minimal necessary closing brackets.
+Sequence: [ < (
+Rules:
+- Add only the closing brackets needed to match all unmatched opening brackets
+- Response format (dataset style only): Use "# Thought N: ..." for each step, then "# Step k: add 'X'.", then "FINAL ANSWER: " followed by the complete Dyck sequence. Do not add Qwen/DeepSeek-style prose or conversational commentary."""
+inputs = tokenizer(prompt, return_tensors="pt")
+outputs = model.generate(**inputs, max_new_tokens=256, do_sample=True, temperature=0.05)
+response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+# Parse "FINAL ANSWER: ..." from response for the completed sequence
+```
+### With LoRA adapter (load base + adapter)
+```python
+from unsloth import FastLanguageModel
+model, tokenizer = FastLanguageModel.from_pretrained(
+    model_name="unsloth/DeepSeek-R1-Distill-Qwen-1.5B",
+    max_seq_length=768,
+)
+model, tokenizer = FastLanguageModel.from_pretrained(
+    "YOUR_USERNAME/YOUR_REPO",  # adapter repo
+    max_seq_length=768,
+)
+# Then generate as above
+```
 ## Training Details
+- **Data:** JSONL conversations (user question → assistant reasoning + final answer). Dataset size configurable (e.g. 60k).
+- **Split:** ~95% train, ~5% eval.
+- **Sequence length:** 768 tokens (run `check_dataset_seq_len.py` to confirm max).
+- **Optimization:** AdamW, cosine LR 6e-6, warmup 25%, max_grad_norm=0.5. 2 epochs typical.
+- **Weighted loss:** Tokens from "FINAL ANSWER: " onward get weight 5.0; reasoning tokens 1.0 (stronger signal on the answer).
+## Limitations
+- Trained on synthetic Dyck data; may not generalize to arbitrary bracket-like tasks.
+- Performance depends on prefix length and bracket vocabulary.
+## Citation
+If you use this model, please cite the base model (DeepSeek-R1-Distill-Qwen) and this fine-tuning setup as appropriate.