Upload Chat-Tuning homework models and data

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* 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
+ultrachat_dpo_neg.json filter=lfs diff=lfs merge=lfs -text
+ultrachat_dpo_pos.json filter=lfs diff=lfs merge=lfs -text
+ultrachat_short.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,84 @@
+---
+library_name: pytorch
+pipeline_tag: text-generation
+tags:
+  - text-generation
+  - pytorch
+  - fineweb-edu
+  - ultrachat
+  - homework
+datasets:
+  - HuggingFaceFW/fineweb-edu
+  - HuggingFaceH4/ultrachat_200k
+---
+
+# Chat-Tuning-Homework
+
+This is a course-homework model repo containing both checkpoints and derived data artifacts.
+
+## Contents
+
+- `model_base.pth`: 1.1M-step base model checkpoint in the homework's LLaMA-like single-file format.
+- `model_chat.pth`: chat-tuned checkpoint in the homework model format.
+- `params.json`: model architecture parameters used by the homework `LLM` loader.
+- `ultrachat_short.json`: filtered short-form UltraChat conversations used for chat tuning.
+- `ultrachat_dpo_pos.json`: positive DPO preference data.
+- `ultrachat_dpo_neg.json`: negative DPO preference data.
+
+## Model Card
+
+### Architecture
+
+The checkpoints use the homework transformer architecture with:
+
+- dimension: 1024
+- feed-forward dimension: 4096
+- heads: 16
+- layers: 8
+- maximum sequence length: 1024
+- vocabulary size: 50432
+
+These values are also stored in `params.json`.
+
+### Training Summary
+
+- `model_base.pth` is the pretrained base checkpoint exported from the 1.1M-step FineWebEDU run.
+- `model_chat.pth` is the chat-tuned checkpoint saved after supervised chat tuning in the homework notebook workflow.
+
+These files are intended for loading with the homework `LLM` implementation and the corresponding `load_weights(...)` function.
+
+### Intended Use
+
+- educational experiments
+- homework reproduction
+- lightweight chat fine-tuning exercises
+
+### Limitations
+
+- this is a homework model, not a production model
+- outputs can be repetitive, unstable, or factually incorrect
+- the chat-tuned model was trained on a filtered subset of UltraChat-derived data
+
+## Data Card
+
+### Data Sources
+
+- FineWebEDU for base pretraining
+- UltraChat 200k for chat tuning and preference-style data preparation
+
+### Included Data Files
+
+- `ultrachat_short.json`: shortened chat-tuning corpus
+- `ultrachat_dpo_pos.json`: preferred responses
+- `ultrachat_dpo_neg.json`: dispreferred responses
+
+### Data Notes
+
+These data files are included here for homework reproducibility. They are derived artifacts prepared locally for the assignment workflow rather than canonical upstream dataset exports.
+
+## File Format Notes
+
+- `model_base.pth` and `model_chat.pth` are PyTorch checkpoint dictionaries
+- attention weights are stored in the homework-compatible unpacked format
+- all exported weights are stored as `bfloat16`
+

model_base.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3df23ea539ee489eb5fe48702cd459e6f42b19cace1fb035f761452df8ff178
+size 410006722

model_chat.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc0f260b99b4259d0a69c889d2ed8d6be45c15448675bae9dae0ac7ace947480
+size 410009717

params.json

@@ -0,0 +1,8 @@
+{
+  "dim": 1024,
+  "ffn_dim": 4096,
+  "max_seq_len": 1024,
+  "n_heads": 16,
+  "n_layers": 8,
+  "num_tokens": 50432
+}

ultrachat_dpo_neg.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d9a6473df24ef8cd2429067dccc59e4432ac145656421a9b29ae709fddfb4cd
+size 34866275

ultrachat_dpo_pos.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7592f16b73960e266c7b3ab1ba3fb7dcf601e3a4d9bc6c7824b43c6c8d1f91f2
+size 47060892

ultrachat_short.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7e6efa58b27be6ee8674d31b1e0ee5e75b00a5739583bf9af15dcd84f5b5680
+size 319227374