Write a 0.2 changelog

[philpax]

There's been quite a few changes from 0.1. We should document them for people updating their applications.

[danforbes]

Thoughts on getting this into the next release? #224

[philpax]

Fine by me!

[danforbes]

I guess we'll want to look at this in order to figure out what goes into the changelog v0.1.1...main

[danforbes]

Looking at the history for the inference.rs example is another good way to track changes to the public API https://github.com/rustformers/llm/commits/main/crates/llm/examples/inference.rs

[hhamud]

we can do so automatically here
https://docs.github.com/en/repositories/releasing-projects-on-github/automatically-generated-release-notes

[hhamud]

Example here

What's Changed

• refactor: move load to KnownModel trait by @philpax in refactor: move load to KnownModel trait #199
• Refactors from llm-chain integration by @danforbes in Refactors from llm-chain integration #200
• Fine-grain inference feedback by @danforbes in Fine-grain inference feedback #206
• Update upstream GGML dependency by @danforbes in Update upstream GGML dependency #215
• Implement RedPajama as a NeoX Specialization by @danforbes in Implement RedPajama as a NeoX Specialization #204
• Implement LoRA patching via the Loader by @LLukas22 in Implement LoRA patching via the Loader #211
• Dynamically calculate the LoRA patching context size by @LLukas22 in Dynamically calculate the LoRA patching context size #225
• NeoX -> GPT-NeoX by @danforbes in NeoX -> GPT-NeoX #233
• Update GGML dependency by @danforbes in Update GGML dependency #226
• fix: cancel the loading spinner in error cases by @tehmatt in fix: cancel the loading spinner in error cases #236
• Added support for multiple LoRA adapters by @LLukas22 in Added support for multiple LoRA adapters #232
• fix: qnt0 build by @philpax in fix: qnt0 build #242
• Typo fix in extern function name by @tanmaysachan in Typo fix in extern function name #247
• Implement MPT Model by @danforbes in Implement MPT Model #218
• feat: support quantizing with a GGML container by @philpax in feat: support quantizing with a GGML container #245
• Add perplexity calculation and cli-arg by @tanmaysachan in Add perplexity calculation and cli-arg #248
• Update Gpt-NeoX implementation by @LLukas22 in Update Gpt-NeoX implementation #237
• Fix perplexity calculation by @tanmaysachan in Fix perplexity calculation #251
• GGML README by @danforbes in GGML README #224
• Lingering NeoX reference by @danforbes in Lingering NeoX reference #254
• Model architectures should appear in alphabetic order by @danforbes in Model architectures should appear in alphabetic order #255
• More model standardization by @danforbes in More model standardization #256
• README updates for v0.2 by @danforbes in README updates for v0.2 #253
• More CONTRIBUTING updates by @danforbes in More CONTRIBUTING updates #258
• Update headings by @danforbes in Update headings #259
• Context size consistency & fixes by @danforbes in Context size consistency & fixes #260
• Remove LLaMA "old_loader" by @danforbes in Remove LLaMA "old_loader" #262
• refactor: remove model-associated inference params by @philpax in refactor: remove model-associated inference params #272
• Pull upstream changes into GptNeox by @LLukas22 in Pull upstream changes into GptNeox  #270
• Fix embeddings extraction by @skirodev in Fix embeddings extraction #273
• Update GGML Dependency & Bindings by @danforbes in Update GGML Dependency & Bindings #275
• Add scratchbuffers to MPT by @LLukas22 in Add scratchbuffers to MPT #278
• Fixed broken GPT-J quantization by @LLukas22 in Fixed broken GPT-J quantization #283
• Add HuggingFace's Tokenizer by @RedBoxing in Add HuggingFace's Tokenizer #271
• Sampler rework by @philpax in Sampler rework #280
• Add alternative key shortcuts for shift enter by @pdufour in Add alternative key shortcuts for shift enter #287
• Fix error in Usage section of README by @danforbes in Fix error in Usage section of README #290
• Made Sampler and InferenceParameters threadsafe by @LLukas22 in Made Sampler and InferenceParameters threadsafe #292
• Fix embeddings extraction for all models by @skirodev in Fix embeddings extraction for all models #291
• Fix --help panicing in debug mode by @tehmatt in Fix --help panicing in debug mode #300
• Default to f16 model memory k/v in llm CLI and InferenceSessionConfig by @KerfuffleV2 in Default to f16 model memory k/v in llm CLI and InferenceSessionConfig #296
• Switch from GGML to LLama.cpp as GGML source. by @LLukas22 in Switch from GGML to LLama.cpp as GGML source.  #302
• NixOS flake stuff needs to be in root by @danforbes in NixOS flake stuff needs to be in root #306
• Add bindings section to README by @danforbes in Add bindings section to README #309
• adding secondbrain to applications section on Readme by @juliooa in adding secondbrain to applications section on Readme #310
• Added support for cuBLAS and CLBlast in ggml. by @darxkies in Added support for cuBLAS and CLBlast in ggml. #282
• Fix GPT-2 Positional Embeddings by @LLukas22 in Fix GPT-2 Positional Embeddings #322
• Implement Metal support by @pixelspark in Implement Metal support #311
• fix(metal): patch ggml-metal.m with shader code by @philpax in fix(metal): patch ggml-metal.m with shader code #328
• K-quants by @nightscape in K-quants #326
• feat: automatically run checks before commit by @philpax in feat: automatically run checks before commit #329
• Added InferenceSessionRef to exports, for easier serializing and deserializing. by @JuliaMerz in Added InferenceSessionRef to exports, for easier serializing and deserializing. #335
• Add Falcon Support by @LLukas22 in Add Falcon Support #313
• Added tokenizers-remote by @jafioti in Added tokenizers-remote #339
• Basic Integration Testing by @LLukas22 in Basic Integration Testing #337
• feat(test): check hyperparameters can roundtrip by @philpax in feat(test): check hyperparameters can roundtrip #344
• Improve model architecture error message by @viirya in Improve model architecture error message #348
• fix Split up initial and per-message prompts for chat mode #216 - split prompts for chat mode in CLI by @philpax in fix #216 - split prompts for chat mode in CLI #346
• Add placeholder check for message prompt by @viirya in Add placeholder check for message prompt #351
• Fix linux CLBlast build and add CI for accelerated builds by @LLukas22 in Fix linux CLBlast build and add CI for accelerated builds  #358
• Add CI for cuda builds by @LLukas22 in Add CI for cuda builds #360
• Add ability to delete tokens (undo feed) by @steventrouble in Add ability to delete tokens (undo feed) #334
• Update gpt2 to use wte if no lm_head by @steventrouble in Update gpt2 to use wte if no lm_head #362
• Add CI for OpenCL on windows by @LLukas22 in Add CI for OpenCL on windows #366
• allow chat to halt new token generation on stop_sequence by @averypelle in allow chat to halt new token generation on stop_sequence #364
• feat(tracing): add tracing to llm and llm-base crates by @radu-matei in feat(tracing): add tracing to llm and llm-base crates #367
• CUDA/OpenCL Acceleration by @LLukas22 in CUDA/OpenCL Acceleration #325
• Fix MSRV to 1.65.0 by @philpax in Fix MSRV to 1.65.0 #375
• Peg CUDA architecture by @LLukas22 in Peg CUDA architecture #376
• Update README.md by @basraayman in Update README.md #379
• Update README.md by @baskrahmer in Update README.md #380
• Better GPU Documentation by @LLukas22 in Better GPU Documentation #385
• add Falcon 40B model support by @skirodev in add Falcon 40B model support #368
• Custom RoPE Scaling by @LLukas22 in Custom RoPE Scaling #389
• Fix Halting inference works for session.infer but not for session.feed_prompt #393 halting feed_prompt by @clarkmcc in Fix #393 halting feed_prompt #394
• Use llm_samplers crate for sampler backend by @KerfuffleV2 in Use llm_samplers crate for sampler backend #359

New Contributors

• @skirodev made their first contribution in Fix embeddings extraction #273
• @RedBoxing made their first contribution in Add HuggingFace's Tokenizer #271
• @pdufour made their first contribution in Add alternative key shortcuts for shift enter #287
• @juliooa made their first contribution in adding secondbrain to applications section on Readme #310
• @darxkies made their first contribution in Added support for cuBLAS and CLBlast in ggml. #282
• @pixelspark made their first contribution in Implement Metal support #311
• @nightscape made their first contribution in K-quants #326
• @JuliaMerz made their first contribution in Added InferenceSessionRef to exports, for easier serializing and deserializing. #335
• @jafioti made their first contribution in Added tokenizers-remote #339
• @viirya made their first contribution in Improve model architecture error message #348
• @steventrouble made their first contribution in Add ability to delete tokens (undo feed) #334
• @averypelle made their first contribution in allow chat to halt new token generation on stop_sequence #364
• @radu-matei made their first contribution in feat(tracing): add tracing to llm and llm-base crates #367
• @basraayman made their first contribution in Update README.md #379
• @baskrahmer made their first contribution in Update README.md #380
• @clarkmcc made their first contribution in Fix #393 halting feed_prompt #394

Full Changelog: v0.1.1...v0.2

[hhamud]

Though I think this ci add on looks interesting
https://intuit.github.io/auto/docs

[philpax]

Aye, that's a good start; we'll likely need to edit it to remove the redundancies and focus on the things that are relevant to users, but I really like the new contributors list!