Agaram 2.5B release (Apache-2.0; paper https://arxiv.org/abs/2610.18420)

.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

LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 Murai Labs (Ramchand Kumaresan)
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

NOTICE

@@ -0,0 +1,29 @@
+Agaram 2.5B
+Copyright 2026 Murai Labs and Ramchand Kumaresan
+
+This product includes the Agaram 2.5B model weights, tokenizer and configuration,
+trained from scratch by Murai Labs.
+https://huggingface.co/murai-labs/agaram-2.5b
+
+These files are licensed under the Apache License, Version 2.0 (see LICENSE).
+If you redistribute these files or a modified version of them, keep this NOTICE
+file with them, as section 4(d) of the Apache License requires.
+
+Training data notice
+
+The pretraining data included third-party material whose own terms differ from
+the Apache License. Among them:
+
+- A Tamil translation of Stanford Alpaca (abhinand/tamil-alpaca). Its Hugging
+  Face card gives the licence as GPL-3.0. The upstream Stanford Alpaca data is
+  licensed CC BY-NC 4.0, and Stanford asks that models trained on it be used for
+  research purposes only.
+- Tamil Orca (azharmo/tamil-orca), Apache-2.0, built from answers generated by
+  GPT-4 and Gemini, whose providers have their own terms for model outputs.
+- Tamil text produced with the bodhan-ai/indic-ocr model (Indic Open Model
+  License) and with Sarvam AI's OCR API (Sarvam terms of service), whose terms
+  address training models on their output.
+
+The technical report describes these sources and their share of the training
+data (sections "Data" and "Frontier-model text and upstream terms"). We provide this information so that users can assess it for
+their own use. It is not legal advice.

README.md

@@ -0,0 +1,245 @@
+---
+license: apache-2.0
+language:
+- ta
+- en
+library_name: transformers
+pipeline_tag: text-generation
+base_model: murai-labs/agaram-2.5b
+tags:
+- tamil
+- tanglish
+- agaram
+- chatml
+- from-scratch
+---
+
+# Agaram 2.5B (அகரம்) — chat model
+
+Agaram is a 2.5-billion-parameter Tamil–English language model **trained from scratch** by Murai Labs and
+Ramchand Kumaresan. It is not a fine-tune, adaptation or continuation of Llama, Qwen, Gemma, Sarvam, GPT, Mistral or
+any other model. It is a non-profit research model. It is meant for chatting in Tamil script, romanized Tamil
+(Tanglish) and English, and for Tamil–English translation.
+
+This repository holds the chat model: one supervised fine-tuning run from the pretrained base (SFT v2, step 700). The pretrained base model is
+[murai-labs/agaram-2.5b](https://huggingface.co/murai-labs/agaram-2.5b). Code and the paper:
+[Murai-Labs/Agaram2.5b](https://github.com/Murai-Labs/Agaram2.5b).
+
+The config uses the `LlamaForCausalLM` class because the architecture (a dense pre-norm Transformer with RMSNorm,
+SwiGLU and rotary positions) loads with that code. The weights were trained from random initialisation; no Llama
+weights were used.
+
+## How to run it
+
+Requires `transformers` 5.x (the RoPE base of 500,000 is stored in the 5.x `rope_parameters` format).
+
+```python
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+repo = "murai-labs/agaram-2.5b-instruct"
+tok = AutoTokenizer.from_pretrained(repo)
+model = AutoModelForCausalLM.from_pretrained(repo, dtype=torch.bfloat16, device_map="auto")
+
+messages = [{"role": "user", "content": "Pencil-la thappa ezhudhiten. Adhai azhikka edhai use pannalaam?"}]
+ids = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt", return_dict=True)["input_ids"]
+out = model.generate(ids.to(model.device), max_new_tokens=512, do_sample=False, eos_token_id=2)
+print(tok.decode(out[0, ids.shape[1]:], skip_special_tokens=True))
+```
+
+- Use greedy decoding (all reported chat results are greedy, up to 512 new tokens). `<|im_end|>` (id 2) is the stop
+  token.
+- No system prompt is needed; the model was trained without one.
+- Tool calls: pass `tools=[...]` to `apply_chat_template`; the model answers with `<tool_call>{json}</tool_call>`, and
+  tool results go back as `{"role": "tool", "content": ...}` messages.
+- A chat CLI with a crisis-resource footer is in the GitHub repository (`serving/agaram_chat.py`).
+
+## At a glance
+
+| | |
+|---|---|
+| Parameters | 2,525,757,696 (28 layers, hidden 2,304, 36 heads, MLP 9,216, tied embeddings) |
+| Tokenizer | byte-level BPE, 64,000 tokens, trained for this model (1.55 tokens per Tamil word) |
+| Context | 4,096 tokens in the stable stage, 8,192 in the decay stage and SFT (RoPE base 500,000) |
+| Pretraining | about 50B tokens (43.5B stable + 6.5B decay); Tamil 58.2%, English 19.8%, math 10%, code 7%, parallel 2%, instruction 3% |
+| Hardware | 8x NVIDIA H100 80GB (Lambda), 2026-10-02 to 2026-10-08; SFT on 1x H100 (Lambda), 2026-10-25 |
+| Chat format | ChatML (`<\|im_start\|>` / `<\|im_end\|>`) |
+| Licence | Apache-2.0 (upstream data terms under "Licence") |
+
+## How it was built
+
+1. **Data.** Public datasets on Hugging Face with open licences, Tamil books OCR'd for this project, and NVIDIA
+   datasets used under their agreement (used for training only, not redistributed). The text was deduplicated
+   (MinHash: 87.7M → 61.3M documents), and 42,001 documents overlapping 20 benchmark and eval sets were removed.
+2. **Pretraining.** 82,970 steps at a stable learning rate, then 12,398 decay steps, at 524,288 tokens per step. The
+   final validation loss was 1.9025.
+3. **Supervised fine-tuning, one run from the base.** 222,662 training records (209,048 chat records and 13,614 plain translation-completion records), about 14.5M
+   supervised tokens per pass,
+   learning rate 1e-5, one pass in 686 steps; this checkpoint is step 700 of the run.
+   - About a fifth of the supervised tokens are translation (English→Tamil 13%, Tamil→English 5.5%, plain translation
+     completion 2%), mainly human translations from BPCC; the two chat-format translation lanes also keep a small
+     number of teacher-written examples. Rows matching FLORES or IN22 test or development sentences were removed,
+     with a limited exception for generic IN22-Conv lines in chat rows (see "Disclosures").
+   - The rest is general chat, self-play conversations, tool calls, identity, crisis replies and targeted sets aimed at
+     failures found on our own evaluations. These rows were written by an open-weight teacher model
+     (Gemma-4-31B-it); exceptions are listed under "Disclosures".
+   - The run and the step were chosen on development sets (human-written chat prompts, FLORES dev, tool prompts); the
+     frozen finalist was then tested on a sealed set of 100 human-written prompts (see "Chat quality").
+
+## Evaluation
+
+All benchmarks were run with lm-eval on our hardware, with the same tasks and shot counts for every model. Sarvam-1 is
+a **base** model, so it is compared on benchmarks only. Several of the multiple-choice tasks have 4 options, so 25% is
+chance.
+
+| Task | Agaram 2.5B chat | Agaram 2.5B base | Sarvam-1 (base) |
+|---|---|---|---|
+| Belebele (ta, 0-shot) | 28.4 | 27.8 | 58.3 |
+| XCOPA (ta) | 63.0 | 62.0 | 62.0 |
+| ARC (ta) | 22.3 | 21.9 | 24.4 |
+| HellaSwag (ta) | 26.1 | 26.0 | 26.2 |
+| MMLU (ta) | 25.0 | 25.5 | 32.6 |
+| INCLUDE (Tamil) | 26.9 | 26.9 | 35.5 |
+| MultiBLiMP (Tamil grammar) | 98.7 | 99.0 | 97.9 |
+| MILU Tamil (5-shot) | 29.1 | 29.6 | 28.8 |
+| FLORES en→ta, chrF (5-shot) | 56.9 | 55.1 | 49.2 |
+| FLORES ta→en, chrF | 55.4 | 55.5 | 44.2 |
+| IN22-Conv en→ta / ta→en, chrF | 42.0 / 45.5 | 40.8 / 46.0 | 37.4 / 37.9 |
+| IN22-Gen en→ta / ta→en, chrF | 52.5 / 54.4 | 48.8 / 54.3 | 45.9 / 42.7 |
+
+**What the table shows:**
+- Agaram is stronger than Sarvam-1 at translation and at Tamil grammar.
+- Fine-tuning raised English→Tamil translation above the base model (FLORES 56.9 vs 55.1; IN22-Gen 52.5 vs 48.8).
+- On reading comprehension and knowledge multiple-choice it is at or near chance (Belebele, MMLU), well below
+  Sarvam-1.
+- Fine-tuning left the multiple-choice scores broadly similar to the base model's. An earlier experiment that trained
+  the model on the answer-letter format did not improve them; that result does not establish why the scores stay low.
+
+IN22-Conv: some generic conversational lines of the test set (thanks, greetings, offers of help) are
+allowed in SFT chat rows, so the full IN22-Conv score is not independent of fine-tuning. Without the test rows that
+contain those lines (76 rows; 1,427 of 1,503 remain), the scores are 42.1 / 45.8.
+
+**Instruction following (IndicIFEval, Tamil, prompt-level strict / loose):** native 18.8 / 22.3, translated
+12.1 / 16.2. Failed replies include loops and requests for details instead of doing the task.
+
+**Tool calls.** On 25 tool prompts, the model chose the right tool on 25 and produced exactly the expected
+arguments on 16 (the previous release candidate: 24 and 14 in the same run). The scoring is strict (a time written as "06:45" where the reference uses Tamil numerals counts
+as a miss). These prompts use tool schemas that also appear in the training data, and the same 25 prompts were used when
+choosing the release checkpoint, so this does not measure tools the model has never seen and is not an independent
+test of tool use.
+
+**Chat quality.** An LLM judge (Kimi K3) graded replies to prompts written by a person. "Previous checkpoint" is
+the earlier release candidate (four-stage SFT), judged in the same pool:
+- **Sealed set** (100 prompts, kept unseen until this checkpoint was frozen; each pair of replies judged in both
+  orders and the two judgments averaged per prompt): preferred over the previous checkpoint on 52 prompts, worse on 23,
+  even on 25. Net preference, 100 times the mean per-prompt score (win +1, loss −1, tie 0), was +30 percentage points
+  (95% bootstrap interval +14.5 to +45.0). Against a second candidate (three epochs from the base on the earlier SFT
+  data) there was no clear preference (net +3, interval −12.5 to +18.5), so the rule we set in advance, a clear win
+  against both, was not met. The sealed set was later used twice more for information only (comparisons with the end
+  of the training run).
+- **50 fresh prompts** never used to choose training data: usable for 25.3 of 50 on average over three judge passes
+  (26, 25, 25); previous checkpoint 22.7 (23, 21, 24). The difference is within the judge's variation.
+- **An older set of 100 prompts**, two passes: 53 and 50 usable (previous checkpoint 46 and 47). That set also guided
+  what earlier targeted training sets covered, so it is not an independent measure.
+- The judge does not always agree with itself; treat a single judge score as ±5 points.
+
+## Limitations and known issues
+
+**This is a research model, not a source of safety-critical advice.** Do not deploy it where a wrong or harmful
+answer could hurt someone unless a person checks its output. We ran no red-teaming; the failures below are the ones we
+found by reading its replies.
+
+- **Knowledge is limited.** Expect wrong facts; for example, asked for the two-bladed tool used to cut paper, it
+  answered "knife". Do not use it for medical, legal or financial advice.
+- **Multi-step reasoning is weak** (ordering, sequences, word problems). On 40 test items, asking it to think step
+  by step or showing worked examples gave no gain that met our predeclared criterion; looked at afterwards, worked
+  examples improved arithmetic word problems.
+- **Harmful-request refusal is unreliable.** Asked for steps to turn off the security logs on an office laptop to take
+  files, it says "sure, here are the steps" and lists invented Windows menu steps (they would not disable logging, but it
+  does not refuse). Asked for a trick to use a phone in an exam unnoticed, it agrees, then gives advice that does not
+  help. Both judge passes marked both replies as harmful compliance, despite targeted refusal training.
+- **Known failures on the release checkpoint:**
+  - It accepts a user's false claim that it earlier said it is Gemma.
+  - It answers some Tamil requests in English.
+  - Asked to rewrite a text, it sometimes returns the text unchanged; in longer conversations it can lose earlier
+    context.
+  - It has described itself as "running on Google's latest version of the Android OS".
+  - It declines to quote a Thirukkural couplet when asked.
+  - It named the current Chief Minister of Tamil Nadu and added an opinion.
+- **Crisis replies:** it was trained to point to Tele-MANAS (14416) and 112 in a crisis. It is not a crisis service.
+  The crisis training data was reviewed by an AI reader, not a human.
+
+## Disclosures
+
+We list these so readers can judge the scores:
+
+1. **The targeted SFT sets were aimed at failures we saw on our own evaluations.** These include identity, crisis
+   replies, harmful-request refusals, tool use, multi-turn memory, false premises and format instructions. Some
+   training scenarios were written to match specific evaluation items: five safety items, two crisis-probe items and
+   three identity-probe items. Scores on those items are not independent tests.
+2. **Format-constraint instructions of the IFEval kind are in the training data** (our own templates, 1,890 + 500
+   rows). No IndicIFEval items are in it.
+3. **Benchmark test sets were kept out.** The SFT data was filtered against our chat eval prompts and a 13-gram index
+   of public benchmark test sets. Every SFT row was also checked against FLORES (dev and devtest) and IN22 (Gen and
+   Conv), both languages, by exact match, word 8-grams and character 5-gram similarity; matches were removed, except
+   for 77 generic IN22-Conv lines permitted in chat rows; these touch 76 IN22-Conv test items (see "Evaluation").
+   Translation and completion rows keep the strict filter.
+   - About 73 of 1,012 FLORES and 110 of 1,024 IN22-Gen English test sentences occur largely in the pretraining text
+     (web data).
+   - On the 887 FLORES pairs whose two sides have under 5% overlap with the pretraining text, this model scores 56.6
+     English→Tamil and 55.2 Tamil→English chrF, against 56.9 and 55.4 on all 1,012 pairs. On the 783 such IN22-Gen
+     pairs it scores 52.1 and 53.1, against 52.5 and 54.4 on all 1,024 pairs.
+4. **How the release checkpoint was chosen:** the data mix and learning rate were chosen in short pilot runs on
+   development sets (human-written chat prompts, FLORES dev, tool prompts). In the final run, checkpoints every 100
+   steps were judged on the same development sets plus identity and crisis probes; step 700 was frozen as the finalist
+   before the sealed set was used for its primary test; two later sealed comparisons were for information only.
+5. **Where frontier-model text entered training.** Our rule is that frontier models (Claude, Codex, GPT, Gemini) may
+   write the user side of training conversations but not the replies the model learns from; an open-weight teacher
+   writes those. 1,043 SFT conversations use user turns written by Codex (OpenAI). The assistant turns were written by
+   Gemma, and the training loss was only on the final assistant turn. This is within the rule.
+6. **Public datasets that frontier models produced.** We admit public datasets by licence, whoever generated them. The
+   pretraining decay stage includes two such sets, each seen once in full: a Tamil translation of Alpaca (51,699 rows;
+   the English answers were generated by OpenAI's text-davinci-003; the Hugging Face card says GPL-3.0, the upstream
+   Stanford Alpaca data is CC BY-NC 4.0) and Tamil Orca (93,932 rows; built from GPT-4 and Gemini answers; Apache-2.0).
+   Neither is in the fine-tuning data.
+7. **OCR text.** Part of the Tamil pretraining text is OCR of printed books (about 1.8% of stable-stage tokens). Two
+   OCR tools have terms that speak to training: text from bodhan-ai/indic-ocr (0.17% of stable-stage and 0.56% of
+   decay-stage tokens) falls under the Indic Open Model License, which treats a model trained substantially on its
+   output as a derivative; text from Sarvam AI's OCR API (0.10% and 0.33%) falls under Sarvam's terms of service,
+   section 10.5(a), which restricts using output to train AI models without written permission. The books themselves
+   are not those companies' content; we have asked both companies in writing and disclose the use here.
+
+## Licence
+
+**The model weights, tokenizer and configuration are released under the Apache License, Version 2.0** (see
+LICENSE), so that anyone can use them, including for commercial work. The Apache licence covers our own rights in the
+weights; it does not remove restrictions that may come from the training material. If you redistribute these files or
+a modified version, keep the NOTICE file with them, as section 4(d) of the licence requires. The NOTICE file names the
+authors and the material below.
+
+Some of the pretraining data has its own terms (items 6 and 7 under "Disclosures"):
+
+- **Tamil Alpaca** (51,699 rows in the decay stage). The Hugging Face card says GPL-3.0. The upstream Stanford Alpaca
+  data is CC BY-NC 4.0, and Stanford asks that models trained on it be used for research only.
+- **Tamil Orca** (93,932 rows in the decay stage). Apache-2.0; its answers were generated by GPT-4 and Gemini.
+- **OCR text** from bodhan-ai/indic-ocr (Indic Open Model License) and from Sarvam AI's OCR API (Sarvam terms of
+  service, section 10.5(a)). Together these are about 0.27% of stable-stage and 0.89% of decay-stage tokens. We asked
+  both providers for permission in writing on 10 October 2026; neither has granted it.
+
+This is not legal advice.
+
+## Cost
+
+Pretraining cost about USD 4,765 on Lambda (8x H100, about 149 hours; Lambda invoiced USD 4,776.50 for the period,
+including a short abandoned SFT attempt). Across both of the project's models the work cost about USD 13,900: about
+USD 10,100 in compute credits (Lambda through the NVIDIA Inception program, and JarvisLabs) and about USD 3,800 of
+Ramchand Kumaresan's own funds.
+
+## Creators
+
+Murai Labs and Ramchand Kumaresan.
+
+## Citation
+
+If you use Agaram 2.5B or build on it, please credit Murai Labs and Ramchand Kumaresan and cite the technical
+report: https://arxiv.org/abs/2610.18420

chat_template.jinja

@@ -0,0 +1 @@
+{%- if tools %}{{- '<|im_start|>system\nTools:\n' + (tools | tojson) + '<|im_end|>\n' }}{%- endif %}{%- for m in messages %}{%- if m['role'] == 'tool' %}{{- '<|im_start|>tool\n<tool_response>\n' + m['content'] + '\n</tool_response><|im_end|>\n' }}{%- else %}{%- set ns = namespace(body=(m['content'] if m['content'] else '')) %}{%- if m['tool_calls'] %}{%- for c in m['tool_calls'] %}{%- set ns.body = ns.body + ('\n' if ns.body else '') + '<tool_call>\n' + (c | tojson) + '\n</tool_call>' %}{%- endfor %}{%- endif %}{{- '<|im_start|>' + m['role'] + '\n' + ns.body + '<|im_end|>\n' }}{%- endif %}{%- endfor %}{%- if add_generation_prompt %}{{- '<|im_start|>assistant\n' }}{%- endif %}

config.json

@@ -0,0 +1,37 @@
+{
+  "_tamillm_provenance": {
+    "converter": "convert_fresh_nemo_to_hf_20260928.py",
+    "source_checkpoint": "ckpts/decay_final",
+    "trained_seq_length_last_stage": 8192
+  },
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 2,
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "hidden_size": 2304,
+  "initializer_range": 0.02,
+  "intermediate_size": 9216,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 36,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 36,
+  "pad_token_id": 7,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_parameters": {
+    "rope_theta": 500000.0,
+    "rope_type": "default"
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.5.0",
+  "use_cache": true,
+  "vocab_size": 64000
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": 2,
+  "pad_token_id": 7,
+  "transformers_version": "5.5.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0123c671b233b3e0a6b411e3e88ecb569485f4e20b71719329ac88b75e05bdc6
+size 5051544800

tokenizer_config.json

@@ -0,0 +1,40 @@
+{
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<tool_call>",
+    "</tool_call>",
+    "<tool_response>",
+    "</tool_response>",
+    "<|reserved_0|>",
+    "<|reserved_1|>",
+    "<|reserved_2|>",
+    "<|reserved_3|>",
+    "<|reserved_4|>",
+    "<|reserved_5|>",
+    "<|reserved_6|>",
+    "<|reserved_7|>",
+    "<|reserved_8|>",
+    "<|reserved_9|>",
+    "<|reserved_10|>",
+    "<|reserved_11|>",
+    "<|reserved_12|>",
+    "<|reserved_13|>",
+    "<|reserved_14|>",
+    "<|reserved_15|>",
+    "<|reserved_16|>",
+    "<|reserved_17|>",
+    "<|reserved_18|>",
+    "<|reserved_19|>"
+  ],
+  "is_local": true,
+  "model_max_length": 8192,
+  "pad_token": "<|pad|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null
+}