snapshot-2024-03-17

What's Changed

• Make superbooga & superboogav2 functional again by @oobabooga in #5656
• Add AQLM support (experimental) by @oobabooga in #5466
• Bump AutoAWQ to 0.2.3 (Linux only) by @oobabooga in #5658
• Add StreamingLLM for llamacpp & llamacpp_HF (2nd attempt) by @oobabooga in #5669
• Merge dev branch by @oobabooga in #5680
• UI: Add a new "User description" field for user personality/biography by @oobabooga in #5691
• Merge dev branch by @oobabooga in #5716

Full Changelog: snapshot-2024-03-10...snapshot-2024-03-17

snapshot-2024-03-10

What's Changed

• Update PyTorch to 2.2 (also update flash-attn to 2.5.6) by @oobabooga in #5618
• Do not install extensions requirements by default by @oobabooga in #5621
• Create an update wizard by @oobabooga in #5623
• Gradio 4 by @oobabooga in #5522
• Update peft requirement from ==0.8.* to ==0.9.* by @dependabot in #5626
• Save the extensions after Gradio 4 by @oobabooga in #5632
• Revert gradio to 3.50.2 by @oobabooga in #5640
• Merge dev branch by @oobabooga in #5641
• Add cache_4bit option for ExLlamaV2 by @oobabooga in #5645
• Change cache_4bit to cache_q4, improve descriptions by @bartowski1182 in #5649
• Merge dev branch by @oobabooga in #5655

Full Changelog: snapshot-2024-03-03...snapshot-2024-03-10

snapshot-2024-03-03

What's Changed

• Bump transformers to 4.38.1 for gemma compatibility by @bartowski1182 in #5575
• Bump hqq from 0.1.3 to 0.1.3.post1 by @dependabot in #5582
• Update llama2-chat-format.json by @jeffbiocode in #5593
• Cubic sampling w/ curve param by @kalomaze in #5551
• Merge dev branch by @oobabooga in #5617

Full Changelog: snapshot-2024-02-25...snapshot-2024-03-03

snapshot-2024-02-25

What's Changed

• Move chat UI elements to the right on desktop by @oobabooga in #5538
• Remove message.content from openai streaming API by @deoxykev in #5503
• Merge dev branch by @oobabooga in #5549
• Update accelerate requirement from ==0.25.* to ==0.27.* by @dependabot in #5546
• Bump autoawq from 0.1.8 to 0.2.2 by @dependabot in #5547
• Update optimum requirement from ==1.16.* to ==1.17.* by @dependabot in #5548
• Remove -k from curl command to download miniconda by @oobabooga in #5535
• ngrok logging does not use the shared logger module by @dsroark in #5570
• Llama2chatformat by @jeffbiocode in #5553
• Big picture fixes by @TheLounger in #5565
• Merge dev branch by @oobabooga in #5574

New Contributors

• @deoxykev made their first contribution in #5503
• @dsroark made their first contribution in #5570
• @jeffbiocode made their first contribution in #5553

Full Changelog: snapshot-2024-02-18...snapshot-2024-02-25

snapshot-2024-02-18

What's Changed

• Bump exllamav2 to 0.0.13.1 by @oobabooga in #5463
• Update main.css to allow scrolling in code blocks by @tseven in #5495
• Bump hqq from 0.1.2.post1 to 0.1.3 by @dependabot in #5489
• Update numpy requirement from ==1.24.* to ==1.26.* by @dependabot in #5490
• Update peft requirement from ==0.7.* to ==0.8.* by @dependabot in #5446
• Merge dev branch by @oobabooga in #5496
• llamacpp_HF: do not use oobabooga/llama-tokenizer by @oobabooga in #5499
• Merge dev branch by @oobabooga in #5502
• Bump gradio to 4.19 by @DominikKowalczyk in #5419
• Remove "Maximum UI updates/second" parameter by @oobabooga in #5507
• Revert gradio to 3.50.2 by @oobabooga in #5513
• Add a "llamacpp_HF creator" menu by @oobabooga in #5519
• Downloader: start a new session for each file by @oobabooga in #5520
• Add a menu for customizing the instruction template for the model by @oobabooga in #5521
• Add --autosplit flag for ExLlamaV2 by @oobabooga in #5524
• Removed extra spaces from Mistral instruction template that were causing model to misbehave by @fschuh in #5517
• Merge dev branch by @oobabooga in #5530
• Merge dev branch by @oobabooga in #5534

New Contributors

• @tseven made their first contribution in #5495
• @DominikKowalczyk made their first contribution in #5419
• @fschuh made their first contribution in #5517

Full Changelog: snapshot-2024-02-11...snapshot-2024-02-18

snapshot-2024-02-11

What's Changed

• Bump llama-cpp-python to 0.2.38 by @oobabooga in #5420
• Quadratic sampling by @kalomaze in #5403
• Remove non-HF ExLlamaV2 loader by @oobabooga in #5431
• Fix the n_batch slider by @BadisG in #5436
• Split by rows instead of layers for llama.cpp multi-gpu by @Ph0rk0z in #5435
• Improve ChatML template by @BadisG in #5411
• Truncate long chat completions inputs by @oobabooga in #5439
• Add custom sampler order support by @oobabooga in #5443
• Merge dev branch by @oobabooga in #5452
• Merge dev branch by @oobabooga in #5453

Full Changelog: snapshot-2024-02-04...snapshot-2024-02-11

snapshot-2024-02-04

What's Changed

• Swap to huggingface_hub get_token function by @Anthonyg5005 in #5381
• Ability to run the Coqui TTS extension on the CPU by @Hukasx0 in #5365
• Support running as a git submodule. by @acdvorak in #5227
• сhanged concatenation of strings to f-strings to improve readability by @eukub in #5068
• [FEAT] Add docker options for CPU only, Intel GPU, AMD GPU by @smCloudInTheSky in #5380
• Add roleplay.gbnf by @BadisG in #5368
• Remove tiktoken from logit bias to fix API logit bias by @Ph0rk0z in #5391
• Bump llama-cpp-python to 0.2.36 by @oobabooga in #5397
• Merge dev branch by @oobabooga in #5404

New Contributors

• @Anthonyg5005 made their first contribution in #5381
• @Hukasx0 made their first contribution in #5365
• @acdvorak made their first contribution in #5227
• @eukub made their first contribution in #5068
• @smCloudInTheSky made their first contribution in #5380

Full Changelog: snapshot-2024-01-28...snapshot-2024-02-04

snapshot-2024-01-28

What's Changed

• Add prompt_lookup_num_tokens parameter by @oobabooga in #5296
• Fixed whisper readme Typo Hyperlink by @Katehuuh in #5281
• Fix error "Attempting to unscale FP16 gradients" when training LoRA by @optozorax in #5268
• Update n_gpu_layers to 256 for larger models. by @Ph0rk0z in #5262
• Update position of show-controls label to avoid line breaks due to font size by @Yiximail in #5256
• do_sample false when temperature 0 by @StefanDanielSchwarz in #5275
• Bump llama-cpp-python to 0.2.29 by @oobabooga in #5307
• Bump llama-cpp-python to 0.2.31 by @oobabooga in #5345
• Fix logprobs tokens in OpenAI API by @lmg-anon in #5339
• Properly handle Images with RGBA color format by @ercanozer in #5332
• Allow content arrays for multimodal OpenAI requests by @Cohee1207 in #5277
• Merge dev branch by @oobabooga in #5347
• Merge dev branch by @oobabooga in #5348
• Bump exllamav2 to 0.0.12 by @oobabooga in #5352
• Bump hqq from 0.1.2 to 0.1.2.post1 by @dependabot in #5349
• Added trust_remote_code to deepspeed when loading model by @sam-ngu in #5237
• Merge dev branch by @oobabooga in #5379

New Contributors

• @Katehuuh made their first contribution in #5281
• @optozorax made their first contribution in #5268
• @lmg-anon made their first contribution in #5339
• @ercanozer made their first contribution in #5332
• @sam-ngu made their first contribution in #5237

Full Changelog: snapshot-2024-01-14...snapshot-2024-01-28

snapshot-2024-01-21

What's Changed

• Bump hqq from 0.1.1.post1 to 0.1.2 by @dependabot in #5204
• Press Tab to switch between current tab and Parameters tab by @oobabooga in #5210
• dynatemp_low, dynatemp_high, dynatemp_exponent parameters by @oobabooga in #5209
• Past chat histories in a side bar on desktop by @oobabooga in #5098
• Reduce the number of built-in presets by @oobabooga in #5217
• Follow internal representation in gradio.components.dropdown.Dropdown when updating instruction template choices by @mamei16 in #5197
• Improve Mistral/Mixtral instruct prompt by @BadisG in #5214
• Merge dev branch by @oobabooga in #5220
• Add requirement jinja2==3.1.* to fix error as described in issue #5240 by @Ahrimdon in #5249
• Merge dev branch by @oobabooga in #5253
• Correct field alias types for OpenAI extension by @samuel-weinhardt in #5257
• Merge dev branch (#5257) by @oobabooga in #5266

New Contributors

• @mamei16 made their first contribution in #5197
• @Ahrimdon made their first contribution in #5249
• @samuel-weinhardt made their first contribution in #5257

Full Changelog: snapshot-2024-01-07...snapshot-2024-01-21

snapshot-2024-01-14

What's Changed

• Bump hqq from 0.1.1.post1 to 0.1.2 by @dependabot in #5204
• Press Tab to switch between current tab and Parameters tab by @oobabooga in #5210
• dynatemp_low, dynatemp_high, dynatemp_exponent parameters by @oobabooga in #5209
• Past chat histories in a side bar on desktop by @oobabooga in #5098
• Reduce the number of built-in presets by @oobabooga in #5217
• Follow internal representation in gradio.components.dropdown.Dropdown when updating instruction template choices by @mamei16 in #5197
• Improve Mistral/Mixtral instruct prompt by @BadisG in #5214
• Merge dev branch by @oobabooga in #5220
• Add requirement jinja2==3.1.* to fix error as described in issue #5240 by @Ahrimdon in #5249
• Merge dev branch by @oobabooga in #5253
• Correct field alias types for OpenAI extension by @samuel-weinhardt in #5257
• Merge dev branch (#5257) by @oobabooga in #5266

New Contributors

• @mamei16 made their first contribution in #5197
• @Ahrimdon made their first contribution in #5249
• @samuel-weinhardt made their first contribution in #5257

Full Changelog: snapshot-2024-01-07...snapshot-2024-01-14